CN118229684B

CN118229684B - Method and system for identifying adrenal pheochromocytoma

Info

Publication number: CN118229684B
Application number: CN202410647405.8A
Authority: CN
Inventors: 丘雅维; 冀宾; 叶婷婷; 朱恒梁; 苏国强; 杨智钧; 蒋澍
Original assignee: Shenzhen University General Hospital
Current assignee: Shenzhen University General Hospital
Priority date: 2024-05-23
Filing date: 2024-05-23
Publication date: 2024-08-13
Anticipated expiration: 2044-05-23
Also published as: CN118229684A

Abstract

The embodiment of the invention relates to the technical field of medicine, and discloses a method and a system for identifying adrenal pheochromocytoma, wherein the method comprises the following steps: determining a first judgment score according to the difference change of the focus texture parameters on different texture image features; determining a second judgment score after obtaining a focus surface area to volume ratio, a focus voxel volume and a 60 th percent of CT value from the three-dimensional volume interested region to determine the current focus form; after the regression prediction model calculates the prediction probability, determining a third judgment score; adding the first judgment score, the second judgment score and the third judgment score according to a certain proportion relation to calculate the total current focus judgment score; if the total score of the current focus judgment exceeds a first specified threshold, determining that the current focus is adrenal pheochromocytoma, and sending prompt information. By implementing the embodiment of the invention, the diagnosis efficiency of adrenal pheochromocytoma can be improved.

Description

Method and system for identifying adrenal pheochromocytoma

Technical Field

The invention relates to the technical field of medicine, in particular to a method and a system for identifying adrenal chromaffin cell tumor.

Background

Adrenal pheochromocytoma (Pheochromocytoma, abbreviated PCC) and adrenal cortical carcinoma (Adrenocortical Carcinoma, abbreviated ACC) are two different types of tumors of the adrenal gland that differ significantly in clinical manifestations, pathological characteristics, treatment methods, prognosis, etc. Thus, accurate identification of these two tumors is of great importance for formulating a correct treatment regimen and predicting prognosis.

However, in practice it has been found that from an imaging characterization point of view, both may show similar local density or signal changes in imaging examinations such as CT and MRI. Adrenal pheochromocytoma may appear as a dense uneven mass in CT, while adrenal cortical carcinoma may appear as a bulky, irregularly shaped solid mass. In MRI examination, signal unevenness may occur in both cases. These similar imaging features may lead to confusion among the physician in the initial judgment. Secondly, both may cause symptoms such as hypertension, metabolic disorder, etc. from the clinical manifestation, making it clinically indistinguishable. For example, typical symptoms of adrenal pheochromocytoma include paroxysmal hypertension, headache, palpitations, hyperhidrosis, etc., while adrenal cortical cancer may lead to symptoms of increased cortisol, increased aldosterone, etc. Overlapping of these symptoms increases the difficulty of diagnosis. In pathological features, although the cell morphology and tissue structure of adrenal pheochromocytoma and adrenal cortical carcinoma differ, in some cases, pathological examination may also be difficult to accurately distinguish between the two due to the complexity and heterogeneity of tumor tissue. From a biological behavior perspective, both may be invasive and metastatic, but the specific behavior may vary. Adrenal pheochromocytoma is mostly benign but still has the possibility of malignant transformation, while adrenal cortical carcinoma is a highly malignant tumor. These differences in biological behavior may also affect the accuracy of the diagnosis. Finally, in terms of diagnostic criteria, due to the similarity of the two in imaging, clinical presentation and pathological features, there is currently no single diagnostic criteria that can accurately distinguish between adrenal pheochromocytoma and adrenal cortical carcinoma, making qualitative diagnosis difficult.

Disclosure of Invention

The embodiment of the invention discloses a method and a system for identifying adrenal pheochromocytoma, which can improve the diagnosis efficiency of adrenal pheochromocytoma.

In a first aspect, an embodiment of the invention discloses a method for identifying adrenal pheochromocytoma, the method comprising:

Carrying out layer-by-layer sketching on a focus interested region of a CT scanning image by using focus image features, and carrying out three-dimensional reconstruction on the focus interested region by using a three-dimensional reconstruction algorithm to form a three-dimensional volume interested region; the CT scanning image comprises a plain scanning CT image and an enhanced scanning CT image;

After different texture image features and quantized focus texture parameters are obtained from the three-dimensional volume region of interest, determining a first judgment score of the current focus as adrenal pheochromocytoma according to the difference change of the focus texture parameters on the different texture image features; wherein different ones of the texture image features include fine texture image features, medium texture image features, and coarse texture image features;

After obtaining the 60 th percentile of the focus surface area to volume ratio, focus voxel volume and CT value from the three-dimensional volume interested region to determine the focus current form, determining a second judgment score of the current focus as adrenal chromaffin cell tumor according to the focus current form;

Substituting the regression dominance ratio of each image autotransformer data in the three-dimensional volume interested region into a regression prediction model to calculate a prediction probability, and determining a third judgment score of the current focus as adrenal pheochromocytoma according to the prediction probability; wherein the image self-variable data at least comprises a ratio of an enhanced peak value to a flat scanning CT value, a flat scanning CT value and a bag variable value;

Adding the first judgment score, the second judgment score and the third judgment score according to a certain proportion relation to calculate a total current focus judgment score;

and if the total judgment score of the current focus exceeds a first specified threshold, determining that the current focus is adrenal pheochromocytoma, and sending prompt information.

A second aspect of an embodiment of the present invention discloses an authentication system including:

The three-dimensional reconstruction unit is used for carrying out layer-by-layer delineation on the focus region of interest on the CT scanning image by utilizing focus image characteristics, and carrying out three-dimensional reconstruction on the focus region of interest by utilizing a three-dimensional reconstruction algorithm so as to form a three-dimensional volume region of interest; the CT scanning image comprises a plain scanning CT image and an enhanced scanning CT image;

A first determining unit, configured to determine a first judgment score that a current focus is adrenal pheochromocytoma according to a difference change of the focus texture parameter on different texture image features after obtaining different texture image features and quantized focus texture parameters from the three-dimensional volume region of interest; wherein different ones of the texture image features include fine texture image features, medium texture image features, and coarse texture image features;

The second determining unit is used for determining a second judgment score of the current focus as adrenal pheochromocytoma according to the current focus form after obtaining the focus surface area to volume ratio, focus voxel volume and 60 th percent of CT value from the three-dimensional volume interested region to determine the current focus form;

The third determining unit is used for determining a third judgment score of the current focus as adrenal pheochromocytoma according to the prediction probability after substituting the regression dominance ratio of the self-variable data of each image in the three-dimensional volume interest region into a regression prediction model to calculate the prediction probability; wherein the image self-variable data at least comprises a ratio of an enhanced peak value to a flat scanning CT value, a flat scanning CT value and a bag variable value;

the calculating unit is used for respectively adding the first judgment score, the second judgment score and the third judgment score according to a certain proportion relation so as to calculate the total current focus judgment score;

And the determining and prompting unit is used for determining that the current focus is adrenal pheochromocytoma and sending prompting information if the total judgment score of the current focus exceeds a first designated threshold.

A third aspect of an embodiment of the present invention discloses an authentication system including:

a memory storing executable program code;

a processor coupled to the memory;

The processor invokes the executable program code stored in the memory to perform a method for identifying adrenal pheochromocytoma as disclosed in the first aspect of the embodiment of the present invention.

A fourth aspect of the embodiments of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program causes a computer to execute a method for identifying adrenal pheochromocytoma disclosed in the first aspect of the embodiments of the present invention.

A fifth aspect of embodiments of the invention discloses a computer program product which, when run on a computer, causes the computer to perform part or all of the steps of any one of the methods of identifying adrenal pheochromocytoma of the first aspect.

A sixth aspect of the embodiments of the present invention discloses an application publishing platform for publishing a computer program product, wherein the computer program product, when run on a computer, causes the computer to perform part or all of the steps of any one of the methods of identifying adrenal pheochromocytoma of the first aspect.

Compared with the prior art, the embodiment of the invention has the following beneficial effects:

In the embodiment of the invention, focus image features are utilized to delineate focus interested areas layer by layer on CT scanning images, and three-dimensional reconstruction is carried out on the focus interested areas by utilizing a three-dimensional reconstruction algorithm so as to form three-dimensional volume interested areas; the CT scanning image comprises a plain scanning CT image and an enhanced scanning CT image; after different texture image features and quantized focus texture parameters are obtained from the three-dimensional volume region of interest, determining a first judgment score of the current focus as adrenal pheochromocytoma according to the difference change of the focus texture parameters on the different texture image features; wherein different ones of the texture image features include fine texture image features, medium texture image features, and coarse texture image features; after obtaining the 60 th percentile of the focus surface area to volume ratio, focus voxel volume and CT value from the three-dimensional volume interested region to determine the focus current form, determining a second judgment score of the current focus as adrenal chromaffin cell tumor according to the focus current form; substituting the regression dominance ratio of each image autotransformer data in the three-dimensional volume interested region into a regression prediction model to calculate a prediction probability, and determining a third judgment score of the current focus as adrenal pheochromocytoma according to the prediction probability; wherein the image self-variable data at least comprises a ratio of an enhanced peak value to a flat scanning CT value, a flat scanning CT value and a bag variable value; adding the first judgment score, the second judgment score and the third judgment score according to a certain proportion relation to calculate a total current focus judgment score; and if the total judgment score of the current focus exceeds a first specified threshold, determining that the current focus is adrenal pheochromocytoma, and sending prompt information. Therefore, the embodiment of the invention can improve the diagnosis efficiency of adrenal gland pheochromocytoma.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic flow chart of a method for identifying adrenal pheochromocytoma according to an embodiment of the present invention;

FIG. 2 is a schematic flow chart of another method for identifying adrenal pheochromocytoma according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of an authentication system according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of another authentication system according to an embodiment of the present invention;

fig. 5 is a schematic diagram of another authentication system according to an embodiment of the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

It should be noted that the terms "first," "second," "third," "fourth," and the like in the description and in the claims of the present invention are used for distinguishing between different objects and not necessarily for describing a particular sequential or chronological order. The terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.

The following detailed description refers to the accompanying drawings.

Example 1

Referring to fig. 1, fig. 1 is a flow chart of a method for identifying adrenal pheochromocytoma according to an embodiment of the present invention. As shown in fig. 1, the method of identifying adrenal pheochromocytoma may include the following steps.

101. The identification system utilizes focus image characteristics to delineate focus interested areas layer by layer on the CT scanning image, and utilizes a three-dimensional reconstruction algorithm to reconstruct the focus interested areas three-dimensionally so as to form three-dimensional volume interested areas; the CT scanning image comprises a plain scanning CT image and an enhanced scanning CT image.

As an optional implementation manner, in the embodiment of the present application, the feature of the focus image refers to the feature of a specific shape, density, size, etc. of the focus in the CT scan image. The identification system may carefully analyze each layer (i.e., each two-dimensional slice) of the CT scan image, and delineate the focal region of interest layer by layer based on the image characteristics of the focal. This process is similar to manually drawing the contours of lesions on an image, but is automated by an identification system, greatly improving efficiency and accuracy. After the focus interested area is sketched layer by layer, the identification system can integrate the two-dimensional focus outline information by utilizing a three-dimensional reconstruction algorithm to form a three-dimensional focus model. The model can more intuitively display the shape and the position of the focus in the three-dimensional space, and is helpful for doctors to more comprehensively know the condition of the focus.

As an alternative implementation, in the embodiment of the present invention, the identification system may finally generate a three-dimensional volumetric region of interest, that is, a three-dimensional spatial region containing the complete lesion, through a three-dimensional reconstruction algorithm. The area not only displays the shape and position of the focus, but also can provide more information about the volume, density distribution and the like of the focus, provides more comprehensive basis for diagnosis and treatment of doctors, is beneficial to the doctors to more accurately identify and analyze the focus, and provides powerful support for diagnosis and treatment.

As an alternative implementation, in the embodiment of the present invention, after layer-by-layer sketching is completed, the two-dimensional focus region of interest information is integrated by using a three-dimensional reconstruction algorithm to form a three-dimensional volumetric region of interest (3D VOI). The three-dimensional reconstruction technology can display the form, the position and the spatial relation between the focus and surrounding tissues in the three-dimensional space, and provides more comprehensive focus information for doctors. After the three-dimensional reconstruction is completed, the system can observe and analyze the three-dimensional volume region of interest at multiple angles and planes to more accurately evaluate the nature, extent and relationship of the lesion to surrounding tissue. This helps the physician to formulate more accurate treatment protocols such as surgical planning, radiotherapy positioning, etc. Meanwhile, the method can also be used for evaluating the treatment effect, observing the change condition of the focus and improving the accuracy and efficiency of diagnosis.

As an alternative implementation, in an embodiment of the present application, the CT scan image of the present application includes a flat scan CT image and an enhanced scan CT image, wherein the flat scan CT image is a CT scan performed without using any contrast agent. It provides basic morphological and structural information of the lesion, but may not be sufficient to clearly show the contrast of the lesion with surrounding tissue. Whereas the enhanced scan CT image is a CT scan performed after injection of contrast agent. The contrast agent may help to enhance the contrast of the lesion with surrounding tissue, making the lesion more prominent and sharp in the image. Enhanced scan CT images are commonly used to further evaluate the nature and extent of lesions.

102. The identification system obtains different texture image characteristics and quantized focus texture parameters from a three-dimensional volume region of interest, and then determines a first judgment score of the current focus as adrenal pheochromocytoma according to the difference change of the focus texture parameters on the different texture image characteristics; wherein the different texture image features include fine texture image features, medium texture image features and coarse texture image features.

As an alternative implementation manner, in the embodiment of the present application, the texture image features refer to quantitative descriptions of characteristics such as pixel gray scale, spatial arrangement, and relationship of focal areas in the image. In medical image processing, texture analysis is often used for assisted diagnosis of diseases.

As an optional implementation manner, in the embodiment of the present application, the present application may observe the texture variation of the lesion on different scales or layers according to the variation trend of the lesion texture parameter on different texture image features. This trend may reveal specific pathological features or growth patterns of the lesions. For example, adrenal pheochromocytomas may exhibit unique patterns of variation in specific texture features, such as high contrast on fine textures or specific periodicity on medium textures.

In an embodiment of the present application, a scoring system or classification rule may be formulated to determine whether the lesion is adrenal pheochromocytoma based on analysis of the trend of the texture parameter. The scoring system may combine a plurality of texture parameters and features to obtain a composite decision score by weighting, thresholding, or machine learning algorithms. This score reflects the similarity or matching of the current lesion to adrenal pheochromocytoma. Based on the first judgment score, the doctor can preliminarily evaluate the nature and likelihood of the lesion. If the score is high, it may suggest that the lesion is adrenal pheochromocytoma; otherwise, further examination or consideration of other diagnostics may be required. The judgment score can be used as one of the reference bases of diagnosis of doctors, and the final diagnosis and treatment scheme can be determined together with other clinical information and examination results.

103. The identification system obtains the ratio of the surface area to the volume of the focus, the voxel volume of the focus and the 60 th percent of CT value from the three-dimensional volume interested region to determine the current focus form, and then determines a second judgment score of the current focus as adrenal chromaffin cell tumor according to the current focus form.

As an alternative embodiment, in the present embodiment, the ratio of the surface area of the lesion to the volume of the lesion may reflect the size of the surface area of the lesion relative to the volume thereof, while a high ratio may mean that the surface of the lesion is rough or irregular, and a low ratio may mean that the lesion is smooth or regular. The lesion voxel volume refers to the total volume of voxels (three-dimensional counterparts of pixels) occupied by the lesion, which provides direct information about the size of the lesion. CT number 60 percent: this is a statistic of the CT value distribution in the lesion area, indicating that 60% of the voxels of the lesion have CT values less than or equal to this value, which helps to characterize the density distribution inside the lesion.

As an alternative embodiment, in the embodiment of the present application, the present application combines the morphological parameters described above, and may comprehensively evaluate the morphological characteristics of the lesion. For example, a lesion with a large surface area to volume ratio, a medium voxel volume, and a specific CT number distribution pattern may have a specific morphological feature. These features can be compared with the morphological features of known adrenal pheochromocytomas to preliminarily determine if the morphology of the lesion is consistent with adrenal pheochromocytomas.

As an alternative implementation manner, in the embodiment of the present application, a set of scoring system or classification rules may be formulated to determine the second judgment score based on the matching degree of the current morphology of the lesion and the morphology feature of the adrenal pheochromocytoma. The score may be a continuous number reflecting the similarity of the lesion to adrenal pheochromocytoma; a classification tag such as "highly suspected", "likely" or "unlikely" may be used. The second judgment score of the application can provide the doctor with additional information about the nature of the focus as a basis for assisting diagnosis. If the score is higher, the likelihood of the lesion being adrenal pheochromocytoma may be increased; conversely, further consideration of other diagnostics may be required. This score may be combined with the results of other diagnostic methods (laboratory tests, pathology tests, etc.), to comprehensively evaluate the nature of the lesion and the accuracy of the diagnosis.

104. The identification system substitutes the regression dominance ratio of each image self-variable data in the three-dimensional volume interested region into a regression prediction model to calculate the prediction probability, and then determines a third judgment score of the current focus as adrenal pheochromocytoma according to the prediction probability; the image self-variable data at least comprises a ratio of an enhanced peak value to a flat scanning CT value, a flat scanning CT value and a bag variable value.

As an optional implementation manner, in the embodiment of the present application, the image self-variable data is extracted from a CT scan image, and can reflect a quantization index of a lesion feature. In this example, the ratio of the enhanced peak value to the flat scan CT value, the flat scan CT value and the bag variation value are at least included. These indices can reflect the enhancement property, density distribution and presence or absence of cystic change region of the lesion, and are of great importance for diagnosis of adrenal pheochromocytoma.

As an alternative implementation, in embodiments of the present invention, logistic regression is a widely used statistical method for classifying problems that predicts the probability of a dependent variable occurring by fitting a relationship between the independent variable and the dependent variable. In this example, the independent variable is image data and the dependent variable is whether the lesion is adrenal pheochromocytoma.

As an alternative implementation, in the embodiment of the present invention, during the modeling process, the system may first select a set of known lesion samples as the training set, where each sample has a corresponding label (i.e., whether it is an adrenal pheochromocytoma). By carrying out regression analysis on the image auto-variable data of the samples, a Logistic regression model which can describe the relationship between the independent variable and the dependent variable can be constructed. After the model is built, the system can substitute the image self-variable data of the current focus into the model to calculate, and the prediction probability that the focus is adrenal pheochromocytoma is obtained. This probability value reflects the size of the probability that the lesion is adrenal pheochromocytoma under the current image characteristics.

As an alternative implementation manner, in the embodiment of the present application, a certain threshold value or a scoring standard may be set according to the prediction probability to determine the third judgment score. For example, it may be set that when the prediction probability is greater than a certain value, the score is "highly suspected"; when the prediction probability is within a certain range, the score is "possible"; when the prediction probability is less than a certain value, the score is "unlikely". The score can provide an objective and statistically-based auxiliary diagnosis basis for doctors, and helps the doctors to judge the focus properties more accurately.

105. The identification system respectively adds the first judgment score, the second judgment score and the third judgment score according to a certain proportion relation to calculate the total current focus judgment score.

As an alternative embodiment, in the embodiment of the present invention, it is assumed that the first judgment score is a, the second judgment score is b, and the third judgment score is c. Assume that the three scores are proportional to p1, p2, p3, where p1+p2+p3=1 (since they are proportional, the sum should be 1). Thus, the system may build the following mathematical model: the current lesion judgment total score = a x p1+ b x p2+ c x p3, this model is a simple weighted sum model in which each score is weighted according to its corresponding scale. The identification system adds the three judgment scores according to a certain proportion in a weighted summation mode to obtain the total judgment score of the current focus.

As an optional implementation manner, in the embodiment of the present application, the third judgment score is combined with the previous first judgment score and second judgment score, and multiple information is comprehensively considered, so that a more comprehensive diagnosis result can be obtained. If multiple judgment scores are all directed to adrenal pheochromocytoma, the doctor can make a diagnosis with more confidence; conversely, if there is a discrepancy or uncertainty between the scores, the physician may need to further consider other diagnostic methods or conduct more detailed examinations.

106. If the total score of the current focus judgment exceeds a first specified threshold, the identification system determines that the current focus is adrenal pheochromocytoma and sends prompt information.

As an alternative implementation, in the embodiment of the present invention, when the system determines that the current lesion is an adrenal pheochromocytoma, the sent prompt information may include detailed information such as the location, size, density, etc. of the lesion, and possible further processing advice. This helps the physician to quickly understand the condition of the lesion and to formulate a corresponding treatment regimen.

In the method of identifying adrenal pheochromocytoma of fig. 1, an identification system is described as an example of an execution subject. It should be noted that, the implementation subject of the method for identifying adrenal pheochromocytoma of fig. 1 may also be a stand-alone device associated with the identification system, and embodiments of the present invention are not limited thereto.

It can be seen that implementing a method for identifying adrenal pheochromocytoma as described in fig. 1 can improve the diagnosis efficiency of adrenal pheochromocytoma.

In addition, the implementation of the method for identifying adrenal pheochromocytoma described in fig. 1 can provide an objective, statistically-based auxiliary diagnosis basis, and help doctors to more accurately judge the nature of the lesions.

Example two

Referring to fig. 2, fig. 2 is a flow chart of another method for identifying adrenal pheochromocytoma according to an embodiment of the present invention. As shown in fig. 2, the method for identifying adrenal pheochromocytoma may include the steps of:

201. The identification system utilizes focus image characteristics to delineate focus interested areas layer by layer on the CT scanning image, and utilizes a three-dimensional reconstruction algorithm to reconstruct the focus interested areas three-dimensionally so as to form three-dimensional volume interested areas; the CT scanning image comprises a plain scanning CT image and an enhanced scanning CT image.

202. The identification system combines a Gaussian spatial band-pass filter and a Laplacian operator to filter the region of interest of the three-dimensional volume, so that different texture image features are obtained.

As an optional implementation manner, in the embodiment of the application, the method can carry out multi-level filtering treatment on the three-dimensional volume interested region by combining the Gaussian spatial band-pass filter and the Laplacian. Firstly, extracting texture information in a specific frequency range by using a Gaussian spatial band-pass filter; the laplacian is then applied to further highlight edges and detail features. This combination process can enhance the textural features of the lesion area, providing more valuable information for subsequent feature extraction and parameter quantification.

As an alternative implementation, in an embodiment of the present application, the application of the gaussian spatial bandpass filter may be as follows: the application can select proper Gaussian space band-pass filter parameters according to the characteristics of medical images and the texture characteristics of focuses. These parameters include spatial frequency range, directivity, etc. of the filter. The system may then apply the designed gaussian spatial bandpass filter to the three-dimensional volumetric region of interest. This process involves convolving each pixel or voxel in the image to highlight texture features within a particular frequency range. Finally, the filtered image is observed and analyzed to ensure that the specific texture features of the lesion area are enhanced while reducing background noise and interference.

As an alternative implementation, in an embodiment of the present invention, the application of the laplace operator may be as follows: first, the system may select an appropriate laplace operator based on the characteristics of the image and the type of edge or detail that needs to be extracted. The laplacian is then applied to the image processed through the gaussian spatial bandpass filter to further emphasize edge and detail features of the lesion area. Finally, the parameters of the Laplacian operator are adjusted to optimize the edge enhancement effect, and meanwhile excessive noise or artifacts are prevented from being introduced.

203. The identification system obtains quantized focus texture parameters from focus image features.

In this embodiment, the lesion image features at least include a gray-based histogram feature, a gray co-occurrence matrix feature, a gray run-length matrix feature, a gray-size region matrix feature, an adjacent gray tone difference matrix feature, a size and shape correlation feature, and the lesion texture parameters at least include an average gray value, a standard deviation, an entropy, an average value of positive pixels, a skewness and a kurtosis of the three-dimensional volume region of interest.

As an alternative implementation manner, in the embodiment of the present application, the present application may select a suitable lesion image feature for extraction according to the research purpose and medical knowledge. These features include gray level based histogram features, gray level co-occurrence matrix features, gray level run length matrix features, and the like. The system may then calculate a numerical representation of the selected feature using image processing algorithms and statistical methods. This may involve counting the gray values of the image, calculating the relationship of adjacent pixels, etc. Finally, the extracted features can be screened and optimized to retain the features most valuable for focus classification or diagnosis, thereby extracting focus image features.

As an alternative implementation, in the embodiment of the present application, the present application may determine the parameters of the lesion texture to be quantified, such as average gray value, standard deviation, entropy, etc., according to the research needs and the related knowledge in the medical field. The system may then calculate values for the selected texture parameters based on the extracted lesion image features. This may involve counting the eigenvalues, calculating the probability distribution, etc. Finally, the system may interpret and analyze the quantified texture parameters, understand their relationship to lesion properties, type, or severity, and thereby perform a quantification operation on the lesion texture parameters.

As an alternative implementation, in an embodiment of the present application, a gaussian spatial bandpass filter is used to highlight image features of different spatial scales. According to the application, the Spatial Scale Factor (SSF) of the filter is adjusted, so that the image characteristics without filtering (SSF 0) and with filtering values of fine (SSF 2), medium (SSF 3-5) and coarse (SSF 6) can be respectively extracted. These different scale features reflect the complexity and layering of the internal structure of the lesion. Next, the present application may utilize histogram analysis to quantify these image features. A histogram is a statistical chart representing the distribution of the number of pixels at different gray levels in an image.

As an optional implementation manner, in an embodiment of the present application, by analyzing the histogram, the present application may extract a plurality of quantized texture parameters, including: average gray value: an average value representing pixels within the region of interest reflecting the overall brightness level of the lesion; standard deviation: the variation or dispersion degree between the pixel value and the average value is represented, and the fluctuation condition of the gray value in the focus is measured; entropy: the irregularity of the pixel intensity distribution is represented, and the complexity and the information content of the internal structure of the focus are reflected; average value of positive pixels: mean values representing pixel values greater than zero, possibly used to exclude background noise or low intensity areas, focusing on the actual part of the lesion; degree of deviation: representing the asymmetry of the histogram to describe the degree of skew of the pixel value distribution, helping to identify texture features that are not normally distributed; kurtosis: representing the sharpness of the histogram, measures the sharpness of the distribution pattern of pixel values, and helps to distinguish between different types of texture structures.

As an alternative embodiment, in the present examples, these texture parameters provide rich information about the internal structure and tissue characteristics of the lesion, which can be used to aid in diagnosis of adrenal pheochromocytoma. By combining a plurality of texture parameters, the characteristics of the focus can be more comprehensively evaluated, and the accuracy and reliability of diagnosis are improved.

204. The discrimination system obtains a first judgment score when the average gray value and kurtosis have a significant difference on the fine texture image feature, the medium texture image feature, and the coarse texture image feature on the plain CT image, respectively, the standard deviation, the entropy, and the average value of the positive pixels have a significant difference on the fine texture image feature and the coarse texture image feature on the plain CT image, respectively, and the average gray value, the standard deviation, the entropy, and the average value of the positive pixels have a significant difference on the fine texture image feature, the medium texture image feature, and the coarse texture image feature on the enhanced scan CT image, respectively, and the kurtosis have a significant difference on the fine texture image feature and the coarse texture image feature on the enhanced scan CT image, respectively.

As an alternative implementation, in an embodiment of the present invention, the authentication system determines the first judgment score by further analyzing a plurality of texture image features of the pan-scan CT image and the enhanced-scan CT image. This score is derived based on the variation of the differences in the various texture parameters including average gray scale value, standard deviation, entropy, average value of positive pixels, kurtosis and skewness across the different texture features. On a plain CT image, the system first observes the variation of average gray values and kurtosis over fine, medium and coarse texture image features. It is a key basis for the decision if the average gray value and kurtosis show significant differences in these three texture features. In addition, the system analyzes differences in standard deviation, entropy, and mean of positive pixels over fine and coarse texture image features, and differences in skewness over medium texture image features. Significant differences in these parameters also have an important impact on the determination of the first judgment score. For enhanced scan CT images, the system then focuses on the variation of the average gray scale value, standard deviation, entropy, and average of positive pixels over fine, medium, and coarse texture image features. It would be another important criterion if these parameters exhibited significant differences in all three texture features. At the same time, the system is also particularly aware of the difference in kurtosis and skewness in the fine texture image features, as this is also a key factor affecting the first judgment score. After comprehensively considering all the factors, the identification system can calculate a specific first judgment score according to a preset algorithm and weight. The score reflects the similarity of the texture characteristics of the focus in the CT image and adrenal pheochromocytoma, and is one of important bases for the auxiliary diagnosis of the identification system.

205. The identification system extracts a focus three-dimensional surface model from the three-dimensional volume region of interest using a three-dimensional image processing algorithm to calculate a focus surface area, and divides the focus surface area by the focus total volume to calculate a focus surface area to volume ratio after calculating the focus total volume by a voxel counting or integration method based on voxel data of the three-dimensional volume region of interest.

As an alternative implementation, in the embodiment of the present invention, the system may use a three-dimensional image processing algorithm (such as threshold segmentation, region growing, morphological operations, etc.) to separate the boundaries of the lesion from the three-dimensional volumetric region of interest, and then construct a three-dimensional surface model of the lesion based on these boundary information, where the model is a three-dimensional grid structure, and can accurately represent the shape of the lesion.

As an alternative, in embodiments of the present invention, for an extracted three-dimensional surface model, the system may calculate its surface area, which typically involves calculating the area of each triangular or polygonal patch in the model and summing them to give the total surface area.

As an alternative implementation, in an embodiment of the present invention, the system may determine the total volume of the lesion by voxel counting or integration methods based on voxel data of the three-dimensional volumetric region of interest. The voxel counting method is to count the number of voxels in a lesion area and multiply the volume of each voxel (generally known) to obtain the total volume. The integration method is to estimate the total volume through an integration operation according to the density distribution of the focus area or other relevant parameters.

As an alternative embodiment, in the present example, after the system obtains the surface area and total volume of the lesion, the system can calculate the ratio of the surface area to the volume by a simple division operation. This ratio is a dimensionless quantity that reflects the geometric characteristics of the lesion and may have some reference value for distinguishing between different types of lesions (e.g., adrenal pheochromocytoma).

206. After determining the focus voxel set according to the focus segmentation result, the identification system multiplies the total number of focus voxels in the focus voxel set by a single voxel volume to calculate the focus voxel volume if the shape of each focus voxel in the focus voxel set is the same, and calculates the focus voxel volume by accumulating the volumes of each focus voxel if the shape of each focus voxel in the focus voxel set is different.

As an alternative implementation, in an embodiment of the present invention, if the shape and size of each focus voxel within the set of focus voxels is the same, the process of calculating the volume of focus voxels is relatively simple. The system only needs to multiply the total number of focus voxels in the focus voxel set by the volume of a single voxel to obtain the focus voxel volume. This is because each voxel represents the same three-dimensional space size, so their volumes are equal. By simple multiplication, the system can rapidly and accurately calculate the total volume of the focus.

As an alternative implementation, in the embodiment of the present invention, if the shape and size of each focus voxel in the focus voxel set are different, the process of calculating the focus voxel volume is slightly more complicated. In this case, the system may compute the volume of each lesion voxel individually and accumulate them to obtain the lesion voxel volume. This typically involves measuring or calculating the three-dimensional size of each voxel and then determining the volume of each voxel from these sizes. This process may require more computing resources and time, but it can more accurately reflect the actual size and shape of the lesion in three-dimensional space.

As an alternative implementation, in the embodiment of the present application, whichever method is adopted, the focal voxel volume calculated by the present application is an important basis for the identification system to perform subsequent analysis and diagnosis. From this volume information, the physician can assess the size, location and morphology of the lesion, and thereby formulate a more accurate treatment plan. Meanwhile, the calculation of the volume of the focus voxels also provides basic data for the calculation of other related indexes (such as the ratio of the focus surface area to the volume).

As an alternative implementation, in an embodiment of the present application, voxels within a focus voxel set may not be perfectly regular cubes or other simple shapes, but rather have a more complex morphology. Thus, while computing the volume of each voxel, the system may need to employ more complex algorithms or models to more accurately describe and compute their volumes, and the present application is not limited in any way. In addition, to ensure accuracy and reliability of the calculations, the system also needs to pre-process and correct the image data to eliminate possible noise and artifacts.

207. And after sequencing the CT values of each focus voxel in the three-dimensional volume interested region, the identification system selects the CT value of the focus voxel at the 60 th percentile to determine the 60 th percentile of the CT values.

As an alternative implementation, in an embodiment of the present invention, the identification system first ranks the CT values of each lesion voxel when performing CT value analysis on the lesion voxels in the region of interest of the three-dimensional volume. The CT value (Computed Tomography Number) is a value representing the density of tissue, and may be different for different tissues or lesions. By ordering the CT values, the system can learn the distribution of the density inside the focus. After the ordering is complete, the system will select a lesion voxel CT value at the 60 th percentile. This means that of the CT values of all lesion voxels, 60% have CT values lower than or equal to this selected value, while the remaining 40% are higher than this value. This particular percentile (60 th percentile) is chosen as a representative value because it reflects the CT value level of most voxels in the set of lesion voxels, while excluding some possible extremes or noise.

As an alternative implementation, in an embodiment of the present invention, the system may first arrange the CT values of all lesion voxels in order from small to large. The 60 th percentile position is then determined, typically by multiplying the total number by 0.6 (i.e., 60%). If the result is not an integer, the nearest integer is typically rounded up or selected as the case may be. And finding the CT value of the corresponding position in the ordered CT value list, wherein the CT value is the CT value of the 60 th bit percentage. The 60 th-position-percent CT value can be used as a characteristic index of the density distribution in the focus, which is helpful for the identification system to further analyze the properties and characteristics of the focus. The physician can make a more accurate diagnosis based on this index and other relevant information. It should be noted that the choice of this percentile is not arbitrary, but is determined based on experience of medical studies and practical applications, which can reflect some important characteristics of the lesion to some extent.

208. The identification system determines the shape regularity of the focus, the size of the focus volume and the distribution condition of the focus density according to the ratio of the focus surface area to the volume, the focus voxel volume and the 60 th percent of CT value, so as to obtain the current focus form.

In the embodiment of the application, the surface area and volume ratio of the focus are in inverse proportion to the shape regularity of the focus, the volume of the focus voxel is in direct proportion to the volume of the focus, and the 60 th percent of CT value is in direct proportion to the density distribution condition of the focus.

As an alternative embodiment, in an embodiment of the present invention, first, the ratio of the surface area of the lesion to the volume is used to determine the shape regularity of the lesion. This ratio is inversely proportional to the degree of regularity of the lesion shape, meaning that the smaller the ratio, the more regular the shape of the lesion; the larger the ratio, the more irregular the shape. Second, the lesion voxel volume directly reflects the volume size of the lesion. There is a direct proportional relationship between them, i.e. the larger the voxel volume, the larger the volume of the lesion. Finally, the 60 th percentile of CT values was used to describe the density distribution of lesions. This percentage is also proportional to the lesion density distribution, meaning that the higher the percentage, the more uniform or dense the lesion density distribution; the lower the percentage, the more uneven or sparse the density distribution may be. By combining the three parameters, the identification system can comprehensively evaluate the morphological characteristics of the focus, including the shape, the size and the density distribution of the focus, thereby obtaining the current form of the focus. The evaluation method combines information of multiple dimensions, and improves accuracy and reliability of morphological identification.

209. When the shape regularity of the focus is higher than a second designated threshold, the focus volume is within a designated range, and the focus density distribution is uneven, the discrimination system obtains a second judgment score.

As an alternative implementation, in an embodiment of the present invention, from the known description, the authentication system can pass through three key parameters: lesion shape regularity, lesion volume and lesion density distribution, to evaluate the morphology of the lesions. Each parameter has a specific decision criterion or threshold value based on which the system can derive a corresponding decision score.

In the embodiment of the present application, the identification system of the present application identifies benign adrenal pheochromocytoma, which is generally in the form of round-like regular bumps and uneven density, so when the focal shape regularity is higher than the second designated threshold, this generally means that the focal shape is relatively regular, without too many concave-convex or irregular edges, and meets the shape requirement of benign adrenal pheochromocytoma. Meanwhile, when the lesion volume is within the specified range, it is shown that the size of the lesion is within the size range of adrenal pheochromocytoma. But when the lesion density distribution is non-uniform, this may mean that there is a density difference inside the lesion, which may include different tissue types, necrotic areas, or other anomalies. That is, the density distribution requirement of benign adrenal gland pheochromocytoma is met. Thus, combining these three conditions, the authentication system obtains a second judgment score when they are all satisfied.

210. The identification system respectively sets the ratio of the enhanced enhancement peak value to the flat scanning CT value, the flat scanning CT value and the bag variable value as independent variables, and sets binary results corresponding to the enhanced enhancement peak value to the flat scanning CT value, the flat scanning CT value and the bag variable value as dependent variables, and then fits a logistic regression model.

As an alternative implementation, in embodiments of the present invention, the enhanced peak-to-flat CT ratio may reflect the change in lesion density between enhanced scanning and flat CT. Enhanced scanning is typically used to more clearly visualize the contours of the blood vessel and lesion, so this ratio may help the discrimination system judge the vascular richness, activity, or other relevant characteristics of the lesion. The plain CT value is the CT density value of a lesion when not scanned with enhancement, which provides information about the basal density of the lesion. Different tissue types will have different density profiles under CT scanning, so the flat scan CT value is an important reference for the identification system. The cystic change value may reflect the presence of a cystic component or a cystic change region within the lesion. Cystic changes are a feature of certain types of tumors or lesions, and thus this value is also significant to the identification system. By fitting the three independent variables and the corresponding binary results to the logistic regression model, the discrimination system can build a mathematical model describing the relationship between the independent and dependent variables. This model may help the system more accurately predict or classify a certain characteristic or state of a lesion.

211. And after extracting the regression coefficients of the reinforced peak value and the flat scanning CT value, the flat scanning CT value and the bag variable value from the logistic regression model, the identification system calculates the regression dominance ratio of the reinforced peak value and the flat scanning CT value, the regression dominance ratio of the flat scanning CT value and the regression dominance ratio of the bag variable value by using a regression dominance ratio calculation formula.

As an alternative implementation, in an embodiment of the present invention, the regression dominance ratio is a statistic used to account for the influence of an independent variable on a dependent variable in a logistic regression model. It represents the change in the ratio of the occurrence to non-occurrence probabilities of a dependent variable event when the independent variable is changed by one unit. In medicine and in biometrics, regression dominance is often used to describe the relationship between a certain exposure factor (e.g. a certain characteristic of a lesion) and the risk of developing a disease.

As an alternative implementation, in an embodiment of the present invention, regression dominance=exp (regression coefficient), where exp represents a natural exponential function. Therefore, for three independent variables of the ratio of the enhanced enhancement peak value to the flat scanning CT value, the flat scanning CT value and the bag variable value, the identification system substitutes the regression coefficients of the three independent variables into the formula to calculate the respective regression advantage ratios.

As an alternative embodiment, in embodiments of the present invention, enhancing the regression dominance ratio of the peak-to-average CT ratio may help the discrimination system evaluate the extent to which such changes affect lesion classification or condition judgment. The regression dominance ratio of the flat scan CT values reflects the importance of the basal density to lesion classification. Regression odds ratios of the capsular bag values may help identify the effect of the system on lesion classification by the decompensation. By calculating these regression odds ratios, the discrimination system can more deeply understand the role of the respective variables in the model, thereby more accurately interpreting the results of the model and providing a more reliable basis for subsequent medical diagnosis or treatment decisions.

212. The identification system multiplies the regression dominance ratio of the reinforced peak value and the flat scanning CT value ratio, the regression dominance ratio of the flat scanning CT value and the regression dominance ratio of the bag variation value by the corresponding weight ratio and then performs accumulated calculation to calculate the prediction probability.

In the embodiment of the present application, a predictive probability calculation formula in a regression prediction model=regression dominance ratio of the ratio of enhanced peak value to flat scan CT value ×first weight ratio+regression dominance ratio of flat scan CT value ×second weight ratio+regression dominance ratio of bag variable value ×third weight ratio. The formula multiplies the regression dominance ratio of each argument by the corresponding weight ratio and adds the results together to obtain a comprehensive prediction probability. This predictive probability can be used to assess the likelihood that a lesion belongs to an adrenal pheochromocytoma.

As an alternative implementation, in the embodiment of the invention, the identification system obtains the regression dominance ratios of the enhanced enhancement peak value to the flat scan CT value, the flat scan CT value and the bag variable value through a logistic regression model, and the regression dominance ratios reflect the influence degree of the respective variables on the dependent variable (usually a certain disease state or classification). To translate these impact levels into specific predictive probabilities, the discrimination system employs a weighted accumulation approach. In this approach, each regression dominance ratio is multiplied by a corresponding weight ratio. These weight ratios are typically derived by optimization algorithms in the model training process, which reflect the relative importance of each independent variable in the prediction process.

213. When the predictive probability is above a third specified threshold, the discrimination system obtains a third judgment score.

214. The identification system respectively adds the first judgment score, the second judgment score and the third judgment score according to a certain proportion relation to calculate the total current focus judgment score.

215. And if the total judgment score of the current focus exceeds a first specified threshold, the identification system determines that the current focus is adrenal pheochromocytoma and sends prompt information.

It can be seen that implementing another method for identifying adrenal pheochromocytoma described in fig. 2 can improve diagnosis efficiency of adrenal pheochromocytoma.

In addition, the implementation of the method for identifying adrenal pheochromocytoma described in fig. 2 can combine information in multiple dimensions, so that the accuracy and reliability of adrenal pheochromocytoma identification can be improved.

Example III

Referring to fig. 3, fig. 3 is a schematic diagram of an authentication system according to an embodiment of the present invention. As shown in fig. 3, the authentication system 300 may include a sketching and three-dimensional reconstruction unit 301, a first determination unit 302, a second determination unit 303, a third determination unit 304, a calculation unit 305, and a determination and prompting unit 306, wherein:

The delineating and three-dimensional reconstructing unit 301 performs layer-by-layer delineating of the focus region of interest on the CT scan image by using the focus image features, and performs three-dimensional reconstruction on the focus region of interest by using a three-dimensional reconstruction algorithm to form a three-dimensional volume region of interest; the CT scanning image comprises a plain scanning CT image and an enhanced scanning CT image.

A first determining unit 302, configured to determine, after obtaining different texture image features and quantized lesion texture parameters from the three-dimensional volume region of interest, a first judgment score that the current lesion is adrenal pheochromocytoma according to a difference change of the lesion texture parameters on the different texture image features; wherein the different texture image features include fine texture image features, medium texture image features and coarse texture image features.

The second determining unit 303 is configured to determine, according to the current focal form, a second judgment score that the current focal is adrenal pheochromocytoma after obtaining a focal surface area to volume ratio, a focal voxel volume, and a 60 th percentile of CT values from the three-dimensional volume region of interest to determine the current focal form.

A third determining unit 304, configured to determine, according to the prediction probability, a third judgment score that the current focus is adrenal pheochromocytoma after substituting the regression dominance ratio of each image autotransformer data in the three-dimensional volume region of interest into the regression prediction model to calculate the prediction probability; the image self-variable data at least comprises a ratio of an enhanced peak value to a flat scanning CT value, a flat scanning CT value and a bag variable value.

The calculating unit 305 is configured to add the first judgment score, the second judgment score and the third judgment score according to a certain proportional relationship, so as to calculate a total score of the current lesion judgment.

The determining and prompting unit 306 is configured to determine that the current focus is adrenal pheochromocytoma if the total score of the current focus is above a first specified threshold, and send a prompting message.

As an optional implementation manner, in the embodiment of the present application, the feature of the focus image refers to the feature of a specific shape, density, size, etc. of the focus in the CT scan image. The delineating and three-dimensional reconstruction unit 301 may carefully analyze each layer (i.e., each two-dimensional slice) of the CT scan image, and delineate the focal region of interest layer by layer based on the image characteristics of the focal. This process is similar to manually drawing the contours of the lesion on the image, but is done automatically by the delineation and three-dimensional reconstruction unit 301, greatly improving efficiency and accuracy. After the focal region of interest is delineated layer by layer, the delineating and three-dimensional reconstructing unit 301 may integrate the two-dimensional focal contour information by using a three-dimensional reconstruction algorithm to form a three-dimensional focal model. The model can more intuitively display the shape and the position of the focus in the three-dimensional space, and is helpful for doctors to more comprehensively know the condition of the focus.

As an alternative implementation, in the embodiment of the present invention, after the layer-by-layer sketching is completed, the sketching and three-dimensional reconstructing unit 301 integrates the two-dimensional focus region of interest information by using a three-dimensional reconstruction algorithm to form a three-dimensional volumetric region of interest (3D VOI). The three-dimensional reconstruction technology can display the form, the position and the spatial relation between the focus and surrounding tissues in the three-dimensional space, and provides more comprehensive focus information for doctors. After the three-dimensional reconstruction is completed, the system can observe and analyze the three-dimensional volume region of interest at multiple angles and planes to more accurately evaluate the nature, extent and relationship of the lesion to surrounding tissue. This helps the physician to formulate more accurate treatment protocols such as surgical planning, radiotherapy positioning, etc. Meanwhile, the method can also be used for evaluating the treatment effect, observing the change condition of the focus and improving the accuracy and efficiency of diagnosis.

As an alternative implementation manner, in an embodiment of the present invention, the first determining unit 302 may observe the texture variation of the lesion on different scales or layers according to the variation trend of the lesion texture parameter on different texture image features. This trend may reveal specific pathological features or growth patterns of the lesions. For example, adrenal pheochromocytomas may exhibit unique patterns of variation in specific texture features, such as high contrast on fine textures or specific periodicity on medium textures.

As an alternative implementation manner, in the embodiment of the present invention, the second determining unit 303 may comprehensively evaluate the morphological feature of the lesion in combination with the morphological parameters described above. For example, a lesion with a large surface area to volume ratio, a medium voxel volume, and a specific CT number distribution pattern may have a specific morphological feature. These features can be compared with the morphological features of known adrenal pheochromocytomas to preliminarily determine if the morphology of the lesion is consistent with adrenal pheochromocytomas.

As an alternative implementation, in the embodiment of the present invention, during the model building process, the third determining unit 304 may first select a set of known lesion samples as the training set, where each sample has a corresponding label (i.e. whether it is an adrenal pheochromocytoma). By carrying out regression analysis on the image auto-variable data of the samples, a Logistic regression model which can describe the relationship between the independent variable and the dependent variable can be constructed. After the model is built, the third determining unit 304 may substitute the image self-variable data of the current focus into the model to calculate, so as to obtain the prediction probability that the focus is adrenal pheochromocytoma. This probability value reflects the size of the probability that the lesion is adrenal pheochromocytoma under the current image characteristics.

As an alternative embodiment, in the embodiment of the present invention, it is assumed that the first judgment score is a, the second judgment score is b, and the third judgment score is c. Assume that the three scores are proportional to p1, p2, p3, where p1+p2+p3=1 (since they are proportional, the sum should be 1). Thus, the computing unit 305 may build the following mathematical model: the current lesion judgment total score = a x p1+ b x p2+ c x p3, this model is a simple weighted sum model in which each score is weighted according to its corresponding scale. The identification system adds the three judgment scores according to a certain proportion in a weighted summation mode to obtain the total judgment score of the current focus.

It can be seen that the diagnosis efficiency of adrenal pheochromocytoma can be improved by implementing the identification system described in fig. 3.

In addition, implementing the identification system described in fig. 3 can provide an objective, statistically-based, auxiliary diagnostic basis, helping doctors to more accurately determine the nature of lesions.

Example IV

Referring to fig. 4, fig. 4 is a schematic structural diagram of another authentication system according to an embodiment of the present invention. Wherein the authentication system of fig. 4 is optimized by the authentication system of fig. 3. In comparison with the authentication system of fig. 3, the first determining unit 302 of fig. 4 includes:

And the filtering subunit 3021 is configured to combine the gaussian spatial bandpass filter with the laplace operator, and perform filtering processing on the three-dimensional volumetric region of interest, so as to obtain different texture image features.

As an alternative implementation manner, in an embodiment of the present invention, the filtering subunit 3021 may perform a multi-level filtering process on the three-dimensional volumetric region of interest by combining a gaussian spatial bandpass filter and a laplace operator. Firstly, extracting texture information in a specific frequency range by using a Gaussian spatial band-pass filter; the laplacian is then applied to further highlight edges and detail features. This combination process can enhance the textural features of the lesion area, providing more valuable information for subsequent feature extraction and parameter quantification.

A first obtaining subunit 3022, configured to obtain a quantized lesion texture parameter from the lesion image feature; the focus image features at least comprise gray-based histogram features, gray co-occurrence matrix features, gray run length matrix features, gray size area matrix features, adjacent gray tone difference matrix features, size and shape related features, and the focus texture parameters at least comprise average gray values, standard deviations, entropy, average values of positive pixels, skewness and kurtosis of a three-dimensional volume region of interest.

In comparison with the authentication system of fig. 3, the second determining unit 303 of fig. 4 includes:

A first calculating subunit 3031 is configured to extract a three-dimensional surface model of the lesion from the three-dimensional volume of interest using a three-dimensional image processing algorithm to calculate a surface area of the lesion, and calculate a total volume of the lesion by dividing the surface area of the lesion by the total volume of the lesion based on voxel data of the three-dimensional volume of interest by a voxel counting or integration method to calculate a ratio of the surface area of the lesion to the volume of the lesion.

As an alternative implementation, in this embodiment of the present invention, the first computing subunit 3031 may use a three-dimensional image processing algorithm (such as threshold segmentation, region growing, morphological operation, etc.) to separate the boundaries of the lesion from the three-dimensional volumetric region of interest, and then construct a three-dimensional surface model of the lesion based on these boundary information, where the model is a three-dimensional grid structure, and can accurately represent the shape of the lesion.

As an alternative implementation, in an embodiment of the present invention, the first calculation subunit 3031 may determine the total volume of the lesion by a voxel counting or integration method based on voxel data of the three-dimensional volumetric region of interest. The voxel counting method is to count the number of voxels in a lesion area and multiply the volume of each voxel (generally known) to obtain the total volume. The integration method is to estimate the total volume through an integration operation according to the density distribution of the focus area or other relevant parameters.

As an alternative implementation, in the embodiment of the present invention, after the first calculating subunit 3031 obtains the surface area and the total volume of the lesion, the system may calculate the ratio of the surface area to the total volume by using a simple division operation. This ratio is a dimensionless quantity that reflects the geometric characteristics of the lesion and may have some reference value for distinguishing between different types of lesions (e.g., adrenal pheochromocytoma).

The second calculating subunit 3032 is configured to, after determining the focus voxel set according to the focus segmentation result, multiply the total number of focus voxels in the focus voxel set by a single voxel volume if the shape of each focus voxel in the focus voxel set is the same, calculate a focus voxel volume, and if the shape of each focus voxel in the focus voxel set is different, accumulate the volumes of each focus voxel and calculate a focus voxel volume.

As an alternative implementation, in an embodiment of the present invention, if the shape and size of each focus voxel within the set of focus voxels is the same, the process of calculating the volume of focus voxels is relatively simple. The second calculation subunit 3032 only needs to multiply the total number of lesion voxels within the set of lesion voxels by the volume of a single voxel to obtain the volume of a lesion voxel. This is because each voxel represents the same three-dimensional space size, so their volumes are equal. By simple multiplication, the system can rapidly and accurately calculate the total volume of the focus.

As an alternative implementation, in the embodiment of the present invention, if the shape and size of each focus voxel in the focus voxel set are different, the process of calculating the focus voxel volume is slightly more complicated. In this case, the second calculation subunit 3032 may calculate the volume of each lesion voxel individually and accumulate them to obtain a lesion voxel volume. This typically involves measuring or calculating the three-dimensional size of each voxel and then determining the volume of each voxel from these sizes. This process may require more computing resources and time, but it can more accurately reflect the actual size and shape of the lesion in three-dimensional space.

The selecting subunit 3033 is configured to sort the CT values of each focus voxel in the three-dimensional volume region of interest, and then select the CT value of the focus voxel located in the 60 th percentile to determine the 60 th percentile of the CT values.

As an alternative implementation, in the embodiment of the present invention, the selecting subunit 3033 may first arrange the CT values of all the focus voxels in order from small to large. The 60 th percentile position is then determined, typically by multiplying the total number by 0.6 (i.e., 60%). If the result is not an integer, the nearest integer is typically rounded up or selected as the case may be. And finding the CT value of the corresponding position in the ordered CT value list, wherein the CT value is the CT value of the 60 th bit percentage. The 60 th-position-percent CT value can be used as a characteristic index of the density distribution in the focus, which is helpful for the identification system to further analyze the properties and characteristics of the focus. The physician can make a more accurate diagnosis based on this index and other relevant information. It should be noted that the choice of this percentile is not arbitrary, but is determined based on experience of medical studies and practical applications, which can reflect some important characteristics of the lesion to some extent.

A determining subunit 3034, configured to determine a focal shape regularity, a focal volume size, and a focal density distribution condition according to the focal surface area to volume ratio, the focal voxel volume, and the 60 th percentile of the CT value, so as to obtain a current focal shape; wherein, the surface area and volume ratio of the focus are in inverse proportion to the regularity of the focus shape, the volume of focus voxels is in direct proportion to the volume of the focus, and the 60 th percent of CT value is in direct proportion to the distribution of focus density.

As an alternative embodiment, in an embodiment of the present invention, first, the ratio of the surface area of the lesion to the volume is used to determine the shape regularity of the lesion. This ratio is inversely proportional to the degree of regularity of the lesion shape, meaning that the smaller the ratio, the more regular the shape of the lesion; the larger the ratio, the more irregular the shape. Second, the lesion voxel volume directly reflects the volume size of the lesion. There is a direct proportional relationship between them, i.e. the larger the voxel volume, the larger the volume of the lesion. Finally, the 60 th percentile of CT values was used to describe the density distribution of lesions. This percentage is also proportional to the lesion density distribution, meaning that the higher the percentage, the more uniform or dense the lesion density distribution; the lower the percentage, the more uneven or sparse the density distribution may be. By combining these three parameters, the determination subunit 3034 is able to comprehensively evaluate the morphological features of the lesion, including its shape, size and density distribution, to thereby derive the current morphology of the lesion. The evaluation method combines information of multiple dimensions, and improves accuracy and reliability of morphological identification.

In comparison with the authentication system of fig. 3, the third determination unit 304 of fig. 4 includes:

The fitting subunit 3041 is configured to fit a logistic regression model after setting the ratio of the enhanced peak value to the flat scan CT value, and the bag variable value as independent variables, and setting binary results corresponding to the enhanced peak value to the flat scan CT value, and the bag variable value as dependent variables, respectively.

As an alternative implementation, in embodiments of the present invention, the fitting sub-unit 3041 may build a mathematical model describing the relationship between these independent and dependent variables by fitting the three independent variables and the corresponding binary results to a logistic regression model. This model may help the system more accurately predict or classify a certain characteristic or state of a lesion.

And the third calculation subunit 3042 is configured to calculate, by using a regression dominance ratio calculation formula, a regression dominance ratio of the enhanced peak value to the flat scan CT value, a regression dominance ratio of the flat scan CT value, and a regression dominance ratio of the bag variable value after extracting the regression coefficients of the enhanced peak value to the flat scan CT value, and the bag variable value from the logistic regression model.

As an alternative embodiment, in embodiments of the present invention, enhancing the regression dominance ratio of the peak-to-average CT ratio may help the discrimination system evaluate the extent to which such changes affect lesion classification or condition judgment. The regression dominance ratio of the flat scan CT values reflects the importance of the basal density to lesion classification. Regression odds ratios of the capsular bag values may help identify the effect of the system on lesion classification by the decompensation. By calculating these regression dominance ratios, the third calculation subunit 3042 may further understand the roles of the respective variables in the model, thereby more accurately interpreting the results of the model and providing a more reliable basis for subsequent medical diagnosis or treatment decisions.

The fourth calculating subunit 3043 is configured to substitute the regression dominance ratio of the enhanced peak value to the flat scan CT value, the regression dominance ratio of the flat scan CT value, and the regression dominance ratio of the bag variation value into the regression prediction model, respectively, to calculate the prediction probability.

In comparison with the authentication system of fig. 3, the fourth calculation subunit 3043 of fig. 4 includes:

the calculating module 30431 is configured to multiply the regression dominance ratio of the enhanced peak value to the flat scan CT value, the regression dominance ratio of the flat scan CT value, and the regression dominance ratio of the bag variable value by the corresponding weight ratios, and then perform accumulated calculation to calculate the prediction probability; the predictive probability calculation formula in the regression prediction model=the regression dominance ratio of the enhanced enhancement peak value to the flat scan CT value, the first weight ratio, the regression dominance ratio of the flat scan CT value, the second weight ratio, the regression dominance ratio of the bag variable value and the third weight ratio.

In comparison with the authentication system of fig. 3, the first determining unit 302 of fig. 4 further includes:

The second obtaining subunit 3023 is configured to obtain the first judgment score when the average gray value and the kurtosis have a significant difference on the fine texture image feature, the medium texture image feature, and the coarse texture image feature on the flat scan CT image, respectively, the standard deviation, the entropy, the average value of the positive pixels have a significant difference on the fine texture image feature and the coarse texture image feature on the flat scan CT image, respectively, and the average value of the average gray value, the standard deviation, the entropy, and the positive pixels have a significant difference on the fine texture image feature, the medium texture image feature, and the coarse texture image feature on the enhanced scan CT image, respectively, and the kurtosis have a significant difference on the fine texture image feature and the coarse texture image feature on the enhanced scan CT image, respectively.

As an alternative implementation, in an embodiment of the present invention, the second acquiring subunit 3023 determines the first judgment score by further analyzing multiple texture image features of the pan CT image and the enhancement scan CT image. This score is derived based on the variation of the differences in the various texture parameters including average gray scale value, standard deviation, entropy, average value of positive pixels, kurtosis and skewness across the different texture features. On a plain CT image, the second acquisition subunit 3023 first observes variations in average gray values and kurtosis over fine, medium, and coarse texture image features. It is a key basis for the decision if the average gray value and kurtosis show significant differences in these three texture features. In addition, the system analyzes differences in standard deviation, entropy, and mean of positive pixels over fine and coarse texture image features, and differences in skewness over medium texture image features. Significant differences in these parameters also have an important impact on the determination of the first judgment score. For enhanced scan CT images, the second acquisition subunit 3023 focuses on the variation of the average gray scale value, standard deviation, entropy, and average of positive pixels over fine, medium, and coarse texture image features. It would be another important criterion if these parameters exhibited significant differences in all three texture features. Meanwhile, the second acquisition subunit 3023 also pays particular attention to the difference in kurtosis and skewness in the fine texture image characteristics, since this is also a key factor affecting the first judgment score. After comprehensively considering all the above factors, the second obtaining subunit 3023 may calculate a specific first judgment score according to a preset algorithm and weight. The score reflects the similarity of the texture characteristics of the focus in the CT image and adrenal pheochromocytoma, and is one of important bases for the auxiliary diagnosis of the identification system.

Compared to the authentication system of fig. 3, the second determining unit 303 of fig. 4 further includes:

The third obtaining unit 3035 is configured to obtain a second judgment score when the rule degree of the focal shape is higher than the second specified threshold, the focal volume is within the specified range, and the focal density distribution is uneven.

As an alternative implementation manner, in the embodiment of the present invention, from the known description, the third obtaining unit 3035 may pass three key parameters: lesion shape regularity, lesion volume and lesion density distribution, to evaluate the morphology of the lesions. Each parameter has a specific decision criterion or threshold value, based on which the third acquisition unit 3035 can derive a corresponding decision score.

In the embodiment of the present application, the identification system of the present application identifies benign adrenal pheochromocytoma, which is generally in the form of round-like regular bumps and uneven density, so when the focal shape regularity is higher than the second designated threshold, this generally means that the focal shape is relatively regular, without too many concave-convex or irregular edges, and meets the shape requirement of benign adrenal pheochromocytoma. Meanwhile, when the lesion volume is within the specified range, it is shown that the size of the lesion is within the size range of adrenal pheochromocytoma. But when the lesion density distribution is non-uniform, this may mean that there is a density difference inside the lesion, which may include different tissue types, necrotic areas, or other anomalies. That is, the density distribution requirement of benign adrenal gland pheochromocytoma is met. Thus, the third acquisition unit 3035 obtains the second judgment score when all of the three conditions are satisfied.

In comparison with the authentication system of fig. 3, the third determining unit 304 of fig. 4 further includes:

A fourth acquiring unit 3044 for acquiring a third judgment score when the prediction probability is higher than the third specified threshold.

It can be seen that the diagnosis efficiency of adrenal pheochromocytoma can be improved by implementing the identification system described in fig. 4.

In addition, the implementation of the identification system described in fig. 4 can combine information in multiple dimensions to improve the accuracy and reliability of adrenal chromaffin cytoma identification.

Example five

Referring to fig. 5, fig. 5 is a schematic diagram illustrating an authentication system according to another embodiment of the present invention. As shown in fig. 5, the authentication system may include:

A memory 501 in which executable program codes are stored;

A processor 502 coupled to the memory 501;

The processor 502 invokes executable program codes stored in the memory 501 to execute any one of the methods for identifying adrenal gland pheochromocytoma of fig. 1-3.

The embodiment of the invention discloses a computer readable storage medium which stores a computer program, wherein the computer program enables a computer to execute any one of methods for identifying adrenal gland pheochromocytoma shown in fig. 1-2.

The embodiments of the present invention also disclose a computer program product, wherein the computer program product, when run on a computer, causes the computer to perform some or all of the steps of the method as in the method embodiments above.

Those of ordinary skill in the art will appreciate that all or part of the steps of the various methods of the above embodiments may be implemented by hardware associated with a program that may be stored in a computer-readable storage medium, including Read-Only Memory (ROM), random-access Memory (Random Access Memory, RAM), programmable Read-Only Memory (Programmable Read-Only Memory, PROM), erasable programmable Read-Only Memory (Erasable Programmable Read Only Memory, EPROM), one-time programmable Read-Only Memory (OTPROM), electrically erasable programmable Read-Only Memory (EEPROM), compact disc Read-Only Memory (Compact Disc Read-Only Memory, CD-ROM), or other optical disk Memory, magnetic disk Memory, tape Memory, or any other medium that can be used to carry or store data.

The above description of the method and the identification system for identifying adrenal pheochromocytoma disclosed in the embodiments of the present invention applies specific examples to illustrate the principles and embodiments of the present invention, and the above examples are only used to help understand the method and the core idea of the present invention; meanwhile, as those skilled in the art will vary in the specific embodiments and application scope according to the idea of the present invention, the present disclosure should not be construed as limiting the present invention in summary.

Claims

1. A method for identifying adrenal pheochromocytoma, characterized by comprising the following steps:

2. The method of claim 1, wherein said obtaining different texture image features and quantified lesion texture parameters from the three-dimensional volumetric region of interest comprises:

Filtering the three-dimensional volume interested region by combining a Gaussian spatial band-pass filter and a Laplacian operator, so as to obtain different texture image features;

Obtaining quantized focus texture parameters from the focus image characteristics; the focus image features at least comprise gray-based histogram features, gray-level co-occurrence matrix features, gray-level run-length matrix features, gray-level size area matrix features, adjacent gray tone difference matrix features, size and shape related features, and the focus texture parameters at least comprise average gray values, standard deviations, entropies, average values of positive pixels, skewness and kurtosis of the three-dimensional volume interested region.

3. The method of claim 1, wherein obtaining a lesion surface area to volume ratio, a lesion voxel volume, and a 60 th percentile CT value from the three-dimensional volumetric region of interest to determine a current morphology of the lesion comprises:

Extracting a focus three-dimensional surface model from the three-dimensional volume interested region by using a three-dimensional image processing algorithm to calculate focus surface area, and dividing the focus surface area by the focus total volume after calculating focus total volume by a voxel counting or integrating method based on voxel data of the three-dimensional volume interested region to calculate the focus surface area to volume ratio;

After determining a focus voxel set according to a focus segmentation result, if the shape and the size of each focus voxel in the focus voxel set are the same, multiplying the total number of focus voxels in the focus voxel set by a single voxel volume to calculate the focus voxel volume, and if the shape and the size of each focus voxel in the focus voxel set are different, accumulating the volumes of each focus voxel to calculate the focus voxel volume;

After sequencing the CT values of each focus voxel in the three-dimensional volume interested region, selecting a focus voxel CT value positioned at the 60 th percentile to determine the 60 th percentile of the CT values;

Determining the shape regularity, the size of the focus volume and the distribution condition of the focus density according to the surface area and the volume ratio of the focus, the voxel volume of the focus and the 60 th percent of the CT value so as to obtain the current shape of the focus; the surface area and volume ratio of the focus are in inverse proportion to the shape regularity of the focus, the volume of the focus voxels is in direct proportion to the volume of the focus, and the 60 th percent of the CT value is in direct proportion to the density distribution condition of the focus.

4. A method according to any one of claims 1 to 3, wherein substituting the regression dominance ratio of each image autotransformer data in the three-dimensional volumetric region of interest into a regression prediction model to calculate a prediction probability comprises:

After the ratio of the enhanced enhancement peak value to the flat scan CT value, the flat scan CT value and the bag variable value are respectively set as independent variables, the binary results corresponding to the enhanced enhancement peak value to the flat scan CT value, the flat scan CT value and the bag variable value are set as dependent variables, fitting a logistic regression model;

After extracting the ratio of the enhanced enhancement peak value to the flat scan CT value, the flat scan CT value and the regression coefficient of the bag variable value from the logistic regression model, calculating the regression dominance ratio of the enhanced enhancement peak value to the flat scan CT value, the regression dominance ratio of the flat scan CT value and the regression dominance ratio of the bag variable value by using a regression dominance ratio calculation formula;

And substituting the regression dominance ratio of the enhanced peak value to the flat scan CT value, the regression dominance ratio of the flat scan CT value and the regression dominance ratio of the bag variation value into a regression prediction model respectively, and calculating the prediction probability.

5. The method of claim 4, wherein substituting the regression dominance ratio of the enhanced peak-to-flat scan CT value ratio, the regression dominance ratio of the flat scan CT value, and the regression dominance ratio of the bag variation value into a regression prediction model, respectively, calculates the prediction probability, comprising:

multiplying the regression dominance ratio of the enhanced enhancement peak value to the flat scan CT value, the regression dominance ratio of the flat scan CT value and the regression dominance ratio of the bag variable value by corresponding weight ratios, and then performing accumulated calculation to calculate the prediction probability; wherein, the predictive probability calculation formula in the regression prediction model=the regression dominance ratio of the enhanced enhancement peak value to the flat scan CT value is the first weight ratio+the regression dominance ratio of the flat scan CT value is the second weight ratio+the regression dominance ratio of the bag variable value is the third weight ratio.

6. The method of claim 2, wherein determining a first judgment score for the current lesion as an adrenal pheochromocytoma based on the variance of the lesion texture parameters in different ones of the texture image features comprises:

The first judgment score is obtained when the average gray value and the kurtosis have significant differences on fine texture image features, medium texture image features and coarse texture image features on a flat scan CT image, respectively, the standard deviation, the entropy, the average value of the positive pixels have significant differences on fine texture image features and coarse texture image features on the flat scan CT image, respectively, and the skewness has significant differences on medium texture image features on the flat scan CT image, and the average gray value, the standard deviation, the entropy, and the average value of the positive pixels have significant differences on fine texture image features, medium texture image features and coarse texture image features on an enhanced scan CT image, respectively.

7. The method of claim 3, wherein determining a second judgment score for the current lesion as adrenal pheochromocytoma based on the current morphology of the lesion comprises:

And when the focus shape regularity is higher than a second designated threshold, the focus volume is within a designated range and the focus density distribution condition is uneven, obtaining the second judgment score.

8. The method of claim 4, wherein determining a third judgment score for the current lesion as adrenal pheochromocytoma based on the predictive probability comprises:

And when the prediction probability is higher than a third specified threshold, obtaining the third judgment score.

9. An authentication system, the authentication system comprising:

10. An authentication system, the authentication system comprising:

a memory storing executable program code;

a processor coupled to the memory;

the processor invokes the executable program code stored in the memory to perform the method of identifying adrenal pheochromocytoma of any of claims 1-8.