CN112801979A - Scoliosis detection method and system based on uncertainty reasoning - Google Patents

Abstract

The invention provides a scoliosis detection method based on uncertainty reasoning, which comprises the following steps: s1: acquiring a back image, and preprocessing the back image; s2: setting each key skeleton point, and extracting two-dimensional position information of each key skeleton point from the preprocessed back image; s3: obtaining uncertainty initial evidences corresponding to all key bone points through geometric calculation; s4: and carrying out uncertainty reasoning by using the uncertainty initial evidence to obtain a final conclusion as a scoliosis detection result. The invention also provides a scoliosis detection system based on uncertainty inference, which comprises a human-computer interface module, an image processing module, a key point extraction module, a geometric calculation module, an inference module, a knowledge base, a database and an interpretation module. The invention provides a scoliosis detection method and system based on uncertainty reasoning, and solves the problem that the existing scoliosis detection method is not convenient enough.

Description

Scoliosis detection method and system based on uncertainty reasoning

Technical Field

The invention relates to the technical field of spine detection, in particular to a scoliosis detection method and system based on uncertainty reasoning.

Background

Scoliosis refers to the lateral curvature of the human spine, which may be S-shaped or C-shaped. Lateral curvature of the spine may occur in the cervical, thoracic, lumbar, etc., with thoracic and lumbar being the most common. Eighty-five percent of scoliosis is of unknown cause, the remainder may be secondary to muscle neuropathy or abnormalities in the skeletal and connective tissues. Some people have stable and unchangeable scoliosis, and some people have gradually increased along with time. Early treatment was found to be effective in preventing scoliosis. For example, adolescent idiopathic scoliosis often causes patients to have no obvious clinical symptoms in early stage, no pain and no itch, and the patients have high concealment at the moment and are often easily ignored and miss the optimal treatment opportunity. With the increase of age, the scoliosis amplitude is larger and larger, and the treatment difficulty is increased.

The existing scoliosis detection methods are diversified, and mainly comprise a plumb line detection method, Adam's type detection, a vertebral lateral curvature degree instrument, muscle group palpation and muscle strength test, double-foot length detection, a Cobb's angulometry, vertebral elasticity and skeletal development detection and the like. The qualitative and quantitative analysis of the scoliosis is carried out by comprehensively checking the ulna of the wrist or the front and the side of the pelvis, the intestinal bone, the cervical vertebra, the thoracic vertebra, the lumbar vertebra, the sacrum, the left and the right pelvis, and the like. However, most detection methods still need to be combined with methods of X-ray examination and further examination by orthopedic doctors, and the procedures are complex, the combination of X-ray and manual work is needed, the detection is not convenient enough, the detection efficiency is low, and the cost is high.

In the prior art, as disclosed in the patent of china granted under 2021-01-08, a scoliosis detection method and apparatus, which is disclosed under CN109785297B, under non-X-ray and non-radiation conditions, the scoliosis condition of the person to be detected can be obtained by performing identification and analysis using images collected by a scanner, but detection is not performed by combining uncertainty reasoning, and the interpretability of the detection result is not strong enough.

Disclosure of Invention

The invention provides a scoliosis detection method and system based on uncertainty reasoning, aiming at overcoming the technical defect that the existing scoliosis detection method is not convenient enough.

In order to solve the technical problems, the technical scheme of the invention is as follows:

a scoliosis detection method based on uncertainty reasoning comprises the following steps:

s1: acquiring a back image, and preprocessing the back image;

s2: setting each key skeleton point, and extracting two-dimensional position information of each key skeleton point from the preprocessed back image;

s3: obtaining uncertainty initial evidence corresponding to each key skeleton point through geometric calculation according to the two-dimensional position information of each key skeleton point;

s4: and carrying out uncertainty reasoning by using the uncertainty initial evidence to obtain a final conclusion as a scoliosis detection result.

Preferably, in step S1, the preprocessing the back image includes: the back image was converted to RGB format and the size of the back image was corrected to 256 × 256, and then the gray values of the R, G, B three channel gray map were normalized for all positions of the back image.

Preferably, the set key skeletal points include: a left shoulder inflection point, a right shoulder inflection point, a left arm wrist, a right arm wrist, a left elbow, a right elbow, a vertex, an upper neck, a sternum stem, a sacral midpoint, a left hip joint, and a right hip joint.

Preferably, in step S2, the extracting two-dimensional position information of each key bone point from the preprocessed back image includes the following steps:

s2.1: extracting a channel heat map of each key bone point from the preprocessed back image by adopting a convolutional neural network based on heat map regression;

s2.2: post-processing the confidence of each channel heat map by using a non-maximum suppression mode;

s2.3: and selecting the maximum confidence coefficient in each channel heat map subjected to post-processing as the two-dimensional position information of the corresponding key bone point, and mapping the maximum confidence coefficient to the interval from-1 to convert the maximum confidence coefficient into the confidence coefficient of the corresponding key bone point.

Preferably, the uncertainty initial evidence comprises human body posture distance information, human body posture angle information and evidence reliability;

wherein,

the human body posture distance information is obtained by calculating two-dimensional Euclidean distance approximation of adjacent key skeleton points through an Euclidean distance formula;

the human body posture angle information is approximately obtained according to the human body posture distance information through a cosine formula;

evidence confidence is obtained by an uncertainty calculation method.

Preferably, before the uncertainty inference, the method further comprises: and calculating the included angle between the transverse section of the back of the tested person and the shooting direction according to the middle point of the sacrum, the left hip joint, the right hip joint, the left shoulder inflection point, the right key inflection point and the sternum stem, and correcting the human posture distance information and the human posture angle information.

Preferably, in step S4, the uncertainty inference includes combining evidence uncertainty calculation, uncertainty transfer, and uncertainty conclusion synthesis.

Preferably, in the uncertainty inference process, when the number of uncertainty generating formula rules in the knowledge base cannot obtain the inference result or the inference frequency reaches a preset upper limit, the inference is considered to be abnormal, and the abnormal condition is visually fed back through the display screen.

Preferably, after obtaining the scoliosis detection result, the method further comprises: and sequencing all intermediate conclusions and final conclusions obtained in the uncertainty reasoning process according to the time sequence, and performing visual feedback through a display screen in sequence.

A scoliosis detection system based on uncertainty reasoning is used for realizing the scoliosis detection method based on uncertainty reasoning, and comprises a human-computer interface module, an image processing module, a key point extraction module, a geometric calculation module, a reasoning module, a knowledge base, a database and an interpretation module; wherein,

the human-machine interface module comprises: the system comprises a camera for acquiring a back image, a display screen for visualization and a loudspeaker for audio feedback;

the image processing module is used for preprocessing the back image;

the key point extraction module is used for setting key skeleton points and extracting two-dimensional position information and corresponding credibility of each key skeleton point from the preprocessed back image;

the geometric calculation module is used for performing geometric calculation according to the two-dimensional position information of the key skeleton points to obtain an uncertainty initial evidence;

the reasoning module is used for carrying out uncertainty reasoning on the uncertainty initial evidence according to uncertainty knowledge in the knowledge base to obtain an intermediate conclusion and a final conclusion, and the final conclusion is used as a scoliosis detection result;

the knowledge base is used for storing uncertainty knowledge related to scoliosis detection; the uncertainty knowledge is expressed in an uncertainty generating rule;

the database is used for storing the uncertainty initial evidence, the intermediate reasoning process, the intermediate conclusion, the final conclusion and the corresponding credibility;

the interpretation module is used for recording each intermediate conclusion obtained in the uncertainty inference process and the corresponding credibility, converting each intermediate conclusion into a fuzzy conclusion according to different intervals according to the credibility, and sequencing the intermediate conclusion and the final conclusion according to a time sequence.

Compared with the prior art, the technical scheme of the invention has the beneficial effects that:

the invention provides a scoliosis detection method and system based on uncertainty reasoning, which are characterized in that the geometrical relation information of key skeleton points with credibility is obtained from a back image through geometrical calculation, and the uncertainty reasoning is carried out on the geometrical relation information of the key skeleton points with credibility and an uncertainty generating rule to obtain a scoliosis detection result, so that the scoliosis detection is realized without a specific environment or any wearable equipment, the convenience of scoliosis detection is improved, and the detection cost is reduced.

Drawings

FIG. 1 is a flow chart of the steps for implementing the technical solution of the present invention;

FIG. 2 is a schematic diagram of the module connection of the present invention;

wherein: 1. a human-machine interface module; 2. an image processing module; 3. a key point extraction module; 4. a geometric calculation module; 5. an inference module; 6. a knowledge base; 7. a database; 8. and an explanation module.

Detailed Description

The drawings are for illustrative purposes only and are not to be construed as limiting the patent;

for the purpose of better illustrating the embodiments, certain features of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product;

it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.

The technical solution of the present invention is further described below with reference to the accompanying drawings and examples.

Example 1

As shown in fig. 1, a scoliosis detection method based on uncertainty inference includes the following steps:

s1: acquiring a back image, and preprocessing the back image; in actual implementation, the camera is used for acquiring the back image of the upright upper half body of the human body;

s2: setting each key skeleton point, and extracting two-dimensional position information of each key skeleton point from the preprocessed back image;

s3: obtaining uncertainty initial evidence corresponding to each key skeleton point through geometric calculation according to the two-dimensional position information of each key skeleton point;

more specifically, the uncertainty initial evidence comprises human body posture distance information, human body posture angle information and evidence reliability; wherein,

calculating two-dimensional Euclidean distance approximation of adjacent key skeleton points through an Euclidean distance formula to obtain human posture distance information;

approximately obtaining human body posture angle information according to the human body posture distance information through a cosine formula;

obtaining evidence reliability through an uncertainty calculation method;

in the specific implementation process, the human posture distance information comprises: left humerus length, right humerus length, left forearm length, right forearm length, left shoulder inflection-manubrium distance, right shoulder inflection-manubrium distance, parietal-cervical distance, cervical-manubrium distance, manubrium-sacrum midpoint distance, left hip-sacrum midpoint distance, right hip-sacrum midpoint distance, left elbow-manubrium distance, left elbow-sacrum midpoint distance, left elbow-cervical distance, left elbow-parietal distance, left arm wrist-sternum midpoint distance, left arm wrist-sacrum midpoint distance, left arm wrist-napbrium distance, left arm wrist-parietal distance, right elbow-manubrium distance, right elbow-sacrum midpoint distance, right elbow-cervical distance, right elbow-parietal distance, right arm wrist-sternum distance, right arm wrist-manubrium distance, left arm, elbow-sacrum distance, left elbow-sacrum distance, right elbow-sacrum distance, a right arm wrist-sacrum midpoint distance, a right arm wrist-upper neck distance, and a right arm wrist-vertex distance;

the human body posture angle information comprises: a left shoulder corner, a right shoulder corner, a left shoulder-manubrium-upper neck corner, a right shoulder-manubrium-upper neck corner, a vertex-upper neck-manubrium corner, an upper neck-manubrium-sacrum midpoint corner, a left hip joint-body midline angle, a right hip joint-body midline angle, and a left shoulder corner-manubrium-right shoulder corner;

s4: carrying out uncertainty reasoning by using the uncertainty initial evidence to obtain a final conclusion as a scoliosis detection result;

in actual implementation, the uncertainty initial evidence and the uncertainty generating formula rule in the knowledge base 6 are tried to be subjected to uncertainty matching through the reasoning module 5, and an intermediate conclusion is obtained through uncertainty transmission and uncertainty conclusion synthesis; then, the intermediate conclusion and the initial evidence of the round are transmitted into a database 7 to be used as the initial evidence of the next round of uncertainty reasoning, and the uncertainty reasoning is repeatedly carried out by a reasoning module 5 until a preset reasoning finishing condition is reached to obtain a final conclusion as a scoliosis detection result;

wherein the uncertainty conclusion synthesis algorithm is as follows:

CF_1,2(H)＝CF₁(H)+CF₂(H)-CF₁(H)CF₂(H)，CF₁(H)≥0,CF₂(H)≥0

CF_1,2(H)＝CF₁(H)+CF₂(H)+CF₁(H)CF₂(H)，CF₁(H)<0,CF₂(H)<0

CF_1,2(H)＝(CF₁(H)+CF₂(H))/(1-min{CF₁(H)CF₂(H) }) other cases

Wherein, CF is the confidence of the corresponding conclusion, and the synthesized conclusion includes all uncertainty information of the sub-conclusion.

More specifically, in step S1, the preprocessing the back image includes: the back image is converted to RGB format and the size of the back image is corrected to 256 × 256, and then the gray values of all the positions of the R, G, B three channel gray map of the back image are normalized, so that the gray values are mapped from 0 to 255 to the interval of 0 to 1.

More specifically, the set key skeletal points include: a left shoulder inflection point, a right shoulder inflection point, a left arm wrist, a right arm wrist, a left elbow, a right elbow, a vertex, an upper neck, a sternum stem, a sacral midpoint, a left hip joint, and a right hip joint.

More specifically, in step S2, the extraction of the two-dimensional position information of each key bone point from the preprocessed back image includes the steps of:

s2.1: extracting a channel heat map of each key bone point from the preprocessed back image by adopting a convolutional neural network based on heat map regression; taking the confidence as the value of the heat map;

s2.2: post-processing the confidence of each channel heat map by using a non-maximum suppression mode;

s2.3: and selecting the maximum confidence coefficient in each channel heat map subjected to post-processing as the two-dimensional position information of the corresponding key bone point, and mapping the maximum confidence coefficient to the interval from-1 to convert the maximum confidence coefficient into the confidence coefficient of the corresponding key bone point.

More specifically, before uncertainty inference, the method further comprises: and calculating the included angle between the transverse section of the back of the tested person and the shooting direction according to the middle point of the sacrum, the left hip joint, the right hip joint, the left shoulder inflection point, the right key inflection point and the sternum stem, and correcting the human posture distance information and the human posture angle information.

More specifically, in step S4, uncertainty inference includes combining evidence uncertainty calculation, uncertainty transfer, and uncertainty conclusion synthesis.

More specifically, in the uncertainty inference process, when the number of uncertainty generating formula rules in the knowledge base 6 cannot obtain the inference result or the inference frequency reaches a preset upper limit, the inference is considered to be abnormal, and the abnormal condition is visually fed back through the display screen.

More specifically, after obtaining the scoliosis detection result, the method further comprises: and sequencing all intermediate conclusions and final conclusions obtained in the uncertainty reasoning process according to the time sequence, and performing visual feedback through a display screen in sequence.

Example 2

As shown in fig. 2, a scoliosis detection system based on uncertainty inference is used for implementing the scoliosis detection method based on uncertainty inference, and comprises a human-machine interface module 1, an image processing module 2, a key point extraction module 3, a geometric computation module 4, an inference module 5, a knowledge base 6, a database 7 and an interpretation module 8; wherein,

the human-machine interface module 1 comprises: the system comprises a camera for acquiring a back image, a display screen for visualization and a loudspeaker for audio feedback;

the image processing module 2 is used for preprocessing a back image;

the key point extraction module 3 is used for setting key skeleton points and extracting two-dimensional position information and corresponding credibility of each key skeleton point from the preprocessed back image;

the geometric computation module 4 is used for performing geometric computation according to the two-dimensional position information of the key skeleton points to obtain an uncertainty initial evidence;

the reasoning module 5 is used for carrying out uncertainty reasoning on the uncertainty initial evidence according to the uncertainty knowledge in the knowledge base 6 to obtain an intermediate conclusion and a final conclusion, and taking the final conclusion as a scoliosis detection result;

the knowledge base 6 is used for storing uncertainty knowledge related to scoliosis detection; the uncertainty knowledge is expressed in an uncertainty generating rule; the uncertainty generation formula rule comprises a precondition of uncertainty knowledge, a conclusion of the uncertainty knowledge and corresponding uncertainty knowledge credibility, and the uncertainty knowledge credibility is given by a professional orthopedist; the uncertainty knowledge includes: evaluating the symmetry of the corners of the shoulders of a human body, evaluating the lateral bending degree of the midline of the human body, evaluating the symmetry of inflection points of the shoulders and evaluating the symmetry of the distance between hands and body; wherein, the evaluation of the symmetry of the human body double-shoulder corners is based on the left shoulder corner, the right shoulder corner, the left shoulder-sternum handle-upper neck angle and the right shoulder-sternum handle-upper neck angle to obtain the credibility of the strict symmetry conclusion of the double-shoulder corners of the tested person; the evaluation of the lateral bending degree of the human midline is based on the credibility of the complete normal conclusion of the human midline of the tested person obtained by the vertex-upper neck-sternum handle angle, upper neck-sternum handle-sacrum middle point angle of the human body; the evaluation of the symmetry of the double shoulder inflection points obtains the credibility of a complete normal conclusion of the double shoulder inflection points of the testee based on the left shoulder inflection point, the sternum stem, the right shoulder inflection point angle, the right shoulder inflection point and the left shoulder inflection point of the human body; evaluating the Euclidean distance between the key points of the hand bones and the key points of the body vertical lines based on the length of the humerus and the length of the forearm of the human body to obtain the reliability of a conclusion that the distance between the hand and the body of the human body is completely symmetrical;

the database 7 is used for storing the uncertainty initial evidence, the intermediate reasoning process, the intermediate conclusion, the final conclusion and the corresponding credibility;

the interpretation module 8 is configured to record each intermediate conclusion obtained in the uncertainty inference process and the corresponding credibility, convert each intermediate conclusion into a fuzzy conclusion according to different intervals according to the credibility, and sort the intermediate conclusion and the final conclusion according to a time sequence.

It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.

Claims (10)

1. A scoliosis detection method based on uncertainty reasoning is characterized by comprising the following steps:

s1: acquiring a back image, and preprocessing the back image;

s2: setting each key skeleton point, and extracting two-dimensional position information of each key skeleton point from the preprocessed back image;

s3: obtaining uncertainty initial evidence corresponding to each key skeleton point through geometric calculation according to the two-dimensional position information of each key skeleton point;

s4: and carrying out uncertainty reasoning by using the uncertainty initial evidence to obtain a final conclusion as a scoliosis detection result.

2. The scoliosis detection method based on uncertainty inference as claimed in claim 1, wherein in step S1, the preprocessing of the back image comprises: the back image was converted to RGB format and the size of the back image was corrected to 256 × 256, and then the gray values of the R, G, B three channel gray map were normalized for all positions of the back image.

3. The method for scoliosis detection based on uncertainty inference as claimed in claim 1, wherein the set key skeletal points comprise: a left shoulder inflection point, a right shoulder inflection point, a left arm wrist, a right arm wrist, a left elbow, a right elbow, a vertex, an upper neck, a sternum stem, a sacral midpoint, a left hip joint, and a right hip joint.

4. The scoliosis detection method based on uncertainty inference as claimed in claim 1, wherein in step S2, extracting two-dimensional position information of each key bone point from the preprocessed back image comprises the following steps:

s2.1: extracting a channel heat map of each key bone point from the preprocessed back image by adopting a convolutional neural network based on heat map regression;

s2.2: post-processing the confidence of each channel heat map by using a non-maximum suppression mode;

s2.3: and selecting the maximum confidence coefficient in each channel heat map subjected to post-processing as the two-dimensional position information of the corresponding key bone point, and mapping the maximum confidence coefficient to the interval from-1 to convert the maximum confidence coefficient into the confidence coefficient of the corresponding key bone point.

5. The scoliosis detection method based on uncertainty inference as claimed in claim 1, wherein the initial evidence of uncertainty includes human posture distance information, human posture angle information and evidence confidence level;

wherein,

the human body posture distance information is obtained by calculating two-dimensional Euclidean distance approximation of adjacent key skeleton points through an Euclidean distance formula;

the human body posture angle information is approximately obtained according to the human body posture distance information through a cosine formula;

evidence confidence is obtained by an uncertainty calculation method.

6. The method for scoliosis detection based on uncertainty inference as claimed in claim 3, further comprising, before performing uncertainty inference: and calculating the included angle between the transverse section of the back of the tested person and the shooting direction according to the middle point of the sacrum, the left hip joint, the right hip joint, the left shoulder inflection point, the right key inflection point and the sternum stem, and correcting the human posture distance information and the human posture angle information.

7. The scoliosis detection method based on uncertainty inference as claimed in claim 1, wherein in step S4, the uncertainty inference includes combining evidence uncertainty calculation, uncertainty transmission and uncertainty conclusion synthesis.

8. The scoliosis detection method based on uncertainty inference as claimed in claim 1, wherein in the uncertainty inference process, when the number of uncertainty generating formula rules in the knowledge base cannot obtain the inference result or the inference times reaches a preset upper limit, the inference is considered to be abnormal, and the abnormal condition is visually fed back through the display screen.

9. The scoliosis detection method based on uncertainty inference as claimed in claim 1, further comprising after obtaining the scoliosis detection result: and sequencing all intermediate conclusions and final conclusions obtained in the uncertainty reasoning process according to the time sequence, and performing visual feedback through a display screen in sequence.

10. A scoliosis detection system based on uncertainty reasoning is characterized by comprising a human-computer interface module, an image processing module, a key point extraction module, a geometric calculation module, a reasoning module, a knowledge base, a database and an interpretation module; wherein,

the human-machine interface module comprises: the system comprises a camera for acquiring a back image, a display screen for visualization and a loudspeaker for audio feedback;

the image processing module is used for preprocessing the back image;

the key point extraction module is used for setting key skeleton points and extracting two-dimensional position information and corresponding credibility of each key skeleton point from the preprocessed back image;

the geometric calculation module is used for performing geometric calculation according to the two-dimensional position information of the key skeleton points to obtain an uncertainty initial evidence;

the reasoning module is used for carrying out uncertainty reasoning on the uncertainty initial evidence according to uncertainty knowledge in the knowledge base to obtain an intermediate conclusion and a final conclusion, and the final conclusion is used as a scoliosis detection result;

the knowledge base is used for storing uncertainty knowledge related to scoliosis detection; the uncertainty knowledge is expressed in an uncertainty generating rule;

the database is used for storing the uncertainty initial evidence, the intermediate reasoning process, the intermediate conclusion, the final conclusion and the corresponding credibility;

the interpretation module is used for recording each intermediate conclusion obtained in the uncertainty inference process and the corresponding credibility, converting each intermediate conclusion into a fuzzy conclusion according to different intervals according to the credibility, and sequencing the intermediate conclusion and the final conclusion according to a time sequence.

Images