Nothing Special   »   [go: up one dir, main page]
Next Article in Journal
Optimization of Second-Generation Biodiesel Blends to Enhance Diesel Engine Performance and Reduce Pollutant Emissions
Previous Article in Journal
The Impact of Grid-Forming vs. Grid-Following Converters on Frequency Regulation: Comparing Centralised or Distributed Photovoltaic Generation
You seem to have javascript disabled. Please note that many of the page functionalities won't work as expected without javascript enabled.
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Energies
Volume 17
Issue 23
10.3390/en17235828
energies-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Energy Efficiency in Measurement and Image Reconstruction Processes in Electrical Impedance Tomography

by
Barbara Stefaniak
1,
Tomasz Rymarczyk
1,2,*,
Dariusz Wójcik
1,2,
Marta Cholewa-Wiktor
3,
Tomasz Cieplak
3,
Zbigniew Orzeł
2,
Janusz Gudowski
2,
Ewa Golec
2,
Michał Oleszek
1,2 and
Marcin Kowalski
2
1
Research & Development Centre Netrix S.A., 20-704 Lublin, Poland
2
Institute of Computer Science and Innovative Technologies, WSEI University, 20-209 Lublin, Poland
3
Department of Management, Lublin University of Technology, 20-618 Lublin, Poland
*
Author to whom correspondence should be addressed.
Energies 2024, 17(23), 5828; https://doi.org/10.3390/en17235828
Submission received: 3 October 2024 / Revised: 5 November 2024 / Accepted: 7 November 2024 / Published: 21 November 2024
(This article belongs to the Special Issue Neural Networks and Deep Learning and Their Applications in Electrical Engineering)
Browse Figures
Figure 1
<p>Developed vest with 102 textile electrodes, A—central unit of the device (source own).</p> "> Figure 2
<p>Electrode schema [<a href="#B44-energies-17-05828" class="html-bibr">44</a>].</p> "> Figure 3
<p>The central unit of the device (signature letter A in <a href="#energies-17-05828-f001" class="html-fig">Figure 1</a>) (source own).</p> "> Figure 4
<p>Block diagram of the EIT data frame simulation process (source own).</p> "> Figure 5
<p>Healthy lungs model (source own).</p> "> Figure 6
<p>Models of considered medical lesions (source own).</p> "> Figure 7
<p>Boxplot for the first feature for all classes (source own).</p> "> Figure 8
<p>Confusion matrix for LightGBM model. 0—healthy, 1—COPD, 2—ARDS, 3—PTX, 4—PHTN, 5—PNA, and 6—Bronchospasm (source own).</p> "> Figure 9
<p>Beeswarm of SHAP values for COPD disease (source own).</p> ">
Versions Notes

Abstract

:
This paper presents an energy optimization approach to applying electrical impedance tomography (EIT) for medical diagnostics, particularly in detecting lung diseases. The designed Lung Electrical Tomography System (LETS) incorporates 102 electrodes and advanced image reconstruction algorithms. Energy efficiency is achieved through the use of modern electronic components and high-efficiency DC/DC converters that reduce the size and weight of the device without the need for additional cooling. Special attention is given to minimizing energy consumption during electromagnetic measurements and data processing, significantly improving the system’s overall performance. Research studies confirm the device’s high energy efficiency while maintaining the accuracy of the classification of lung disease using the LightGBM algorithm. This solution enables long-term patient monitoring and precise diagnosis with reduced energy consumption, marking a key step towards sustainable medical diagnostics based on EIT technology.

1. Introduction

Energy efficiency is a critical consideration in modern medical diagnostic systems, where the balance between accuracy and power consumption becomes increasingly important, especially in portable and long-term monitoring devices [1]. Electrical Impedance Tomography is an imaging technology widely applied in medical diagnostics, offering a non-invasive and radiation-free assessment of various health conditions, including respiratory diseases [2]; however, one of the primary challenges of implementing EIT systems lies in optimizing their energy usage during measurement and image reconstruction processes, without compromising diagnostic accuracy [3,4,5].
The article utilizes the following abbreviations: EIT (Electrical Impedance Tomography), LETS (Lung Electrical Tomography System), DC/DC (Direct Current to Direct Current converter), LightGBM (Light Gradient Boosting Machine), COPD (Chronic Obstructive Pulmonary Disease), ARDS (Acute Respiratory Distress Syndrome), PTX (Pneumothorax), PNA (Pneumonia), PHTN (Pulmonary Hypertension), EVLW (Extravascular Lung Water), WHO (World Health Organization), FEV1 (Forced Expiratory Volume in one second), CRP (C-reactive protein), LETSWEB (analytical module for LETS), BSPM (Body Surface Potential Mapping), PCB (Printed Circuit Board), and CT (Computed Tomography).
This paper focuses on energy optimization in EIT systems, particularly in the context of lung disease detection. The designed Lung Electrical Tomography System (LETS) utilizes 102 electrodes and advanced reconstruction algorithms to ensure precise measurements while minimizing energy consumption. The LETS system reduces power consumption by incorporating modern, high-efficiency DC/DC [5] converters and lightweight electronic components, enabling long-term monitoring without additional cooling mechanisms. This research investigates how these design choices improve the system’s energy performance, particularly during electromagnetic measurements and data processing, while maintaining high accuracy using machine learning algorithms such as LightGBM [6,7].
Respiratory diseases are a growing and increasingly serious problem in society. The most common of these include chronic obstructive pulmonary disease (COPD), acute respiratory distress syndrome (ARDS), pneumothorax (PTX), pneumonia (PNA), bronchospasm, and pulmonary hypertension (PHTN). COPD is characterized by an increasing, hardly reversible airflow restriction through the lower airways due to bronchospasm and loss of lung elasticity [8]. This is due to the inflammation in response to harmful dust and gases that hurt the bronchial mucosa [9,10]. Mucus secretion increases, mucus-producing glands overgrow, and the number of inflammatory cells in the mucosa increases. These cells secrete substances that cause damage to the lung tissue in the vicinity of tiny bronchioles, which leads to an irreversible reduction in the diameter of small bronchi and bronchioles and the destruction of the lung parenchyma in their surroundings [11]. A simplified model with a highlighted area containing damaged lungs for very small bronchioles will be presented. It is assumed that such damage occurs at the lung periphery. ARDS is a life-threatening condition in which the lungs cannot work correctly due to injury to the capillary wall. ARDS rarely occurs spontaneously; it usually results from another medical condition, a severe accident, or trauma [12]. Patients with ARDS show varying degrees of altered endothelial and pulmonary epithelial permeability, with concomitant pulmonary edema [13]. Compared with healthy subjects, patients with ARDS manifest increased extravascular lung water (EVLW) for pulmonary arterial pressure [14]. The simplified model primarily emphasizes small changes throughout the lungs, simulating the presence of EVLW. No effective pharmacological treatment for ARDS has been invented [12].
Pneumothorax is a medical condition in which air accumulates in the pleural cavity [15]. In a healthy person, only a negligible amount of fluid in the pleural cavity facilitates lung movement relative to the chest during respiratory movements. Air enters the pleural cavity if there is damage to the lung or chest wall. Its pressure causes the lung to be unable to fill appropriately with air. The accumulated air in the pleural space prevents the lungs from expanding correctly during inspiration, causing difficulty in breathing. If the emphysema is so extensive that there is a significant amount of air in the pleural cavity, the lung on that side stops functioning almost entirely [15]. The simplified model consists, among other things, of a reduced lung, air in the pleura, and a second healthy lung. Research on pneumothorax and its evaluation and management has increased over the past 10 years [16]. However, many questions and technologies have not yet been evaluated, likely forming the basis of future research.
PNA is a disease of the lower respiratory system, most often caused by bacteria and viruses [17]. It is an inflammatory condition of the lungs that affects one or both lungs and primarily attacks the alveoli [8]. The disease can be caused by various agents, such as bacteria, viruses, fungi, other microorganisms, and chemicals [6]. The simplified model encompasses two or one level of the lungs as a disease state.
Bronchospasm refers to the abrupt tightening of the smooth muscles in the bronchial walls, narrowing the airways. This process is triggered by the release of specific substances from immune cells, such as mast cells and basophils, which initiate the constriction. The autonomic nervous system regulates smooth muscle activity in the bronchi, where parasympathetic signals lead to muscle contraction, while sympathetic signals promote relaxation. When the parasympathetic system is unexpectedly activated, it leads to bronchoconstriction (bronchospasm), which can vary in intensity from mild to severe, spreading through the airways and causing breathing difficulties [18]. Bronchospasms can occur for many reasons, including asthma, chronic bronchitis, and anaphylaxis. The simplified model will primarily present a distinct area assigned to the bronchioles.
PHTN is a progressive condition marked by an ongoing rise in pulmonary arterial pressure and vascular resistance, which can eventually result in right ventricular failure and death [19]. This disease primarily affects the pulmonary blood vessels, causing elevated lung pressure. In patients with PHTN, the pulmonary circulation undergoes structural alterations in the vascular walls, pulmonary artery, and its branches—these vessels are responsible for carrying blood from the right ventricle to the lungs [20]. Although PHTN is considered a universally fatal disease, individual survival times can differ. For pulmonary arterial hypertension (WHO group I), the untreated median survival after diagnosis is typically 2 to 3 years, with right ventricular failure being the leading cause of death; however, survival duration is influenced by multiple factors [21]. The simplified model emphasizes an area approximating the location of blood vessels in the lungs.
In medicine, the following tests are used, among others, to help diagnose the presented diseases:
  • Spirometry measures the volume of air a patient can exhale in the first second (FEV1), helping assess the degree of airway obstruction. It also helps diagnose COPD [22], Bronchospasm [23], and PHTN [24].
  • Arterial blood gas analysis provides information on oxygen, carbon dioxide, and pH levels. It helps diagnose COPD [25], ARDS [26], PTX [27], and PHTN [28].
  • Chest computed tomography can assist in identifying features characteristic of COPD [29], ARDS [30], PTX [31], PNA [32], bronchospasm [33], and PHTN [34].
  • Pulmonary imaging can provide additional information about lung structural changes associated with COPD [35].
  • Biomarkers in alveolar fluid may help understand the pathogenesis of ARDS {Citation} and provide additional diagnostic information.
  • Lung ultrasonography, mainly using the BLUE protocol, can help assess lung conditions. It supports the diagnosis of ARDS [36] and PTX [37].
  • Chest X-ray is often the initial test to diagnose a pneumothorax [38]. Characteristic changes include Barkhausen lines, loss of lung vascular markings, and the presence of air in the pleural space. It allows for the observation of changes in lung structure, such as infiltrates or opacities, which may suggest the presence of infection characteristics for PNA [39].
  • Blood tests, especially complete blood count and CRP measurement, can provide information about the presence of infection and the degree of inflammation. They can also help diagnose PNA [32].
  • Sputum culture allows for the identification of the pathogen responsible for pneumonia [40], which is crucial for selecting appropriate antibiotic treatment.
  • Bronchial provocation tests, such as methacholine or histamine, can assess bronchial reactivity, diagnosing bronchial constriction [41].
  • Fractional exhaled nitric oxide (FeNO) measurements can be used to assess inflammatory status in the airways, which may be associated with bronchial constriction [42].
  • Echocardiography is often used to assess the structure and function of the heart and aids in diagnosing pulmonary hypertension by evaluating the pressure in the pulmonary artery [41].
To conduct the described examinations, the patient must adhere to specific guidelines and be instructed on how to behave correctly during the procedure. Sometimes, the results of the tests become available after a certain period, for example, when the appropriate doctor needs to analyze and describe them, which may require some time. The solution that allows for an approximate diagnosis in just a few minutes will be presented.
The research project is centered on developing and assessing the energy efficiency and diagnostic accuracy of the Lung Electrical Tomography System (LETS) using Electrical Impedance Tomography (EIT) technology. Key aspects of this study involve minimizing the system’s energy consumption by integrating modern electronic components and DC/DC converters, thereby enhancing energy efficiency and device durability for long-term patient monitoring.
The project also examines the diagnostic accuracy of the LETS device, employing the LightGBM algorithm to classify lung diseases based on images produced by the system. This approach evaluates the system’s precision and reliability in diagnosing various lung conditions, ensuring it meets clinical requirements for consistent, accurate results.
The methods chosen are carefully justified: Electrical Impedance Tomography (EIT) offers a safe and minimally invasive imaging technique that enables real-time monitoring of structural and functional changes in the lungs, making it highly suitable for continuous observation of patients with lung diseases. Additionally, LightGBM is used for its efficiency and accuracy in handling large datasets, as is often required in medical imaging and is optimized for computational efficiency to limit energy consumption.
By using advanced electronic components and DC/DC converters, the project achieves a reduction in energy loss and cooling requirements, leading to the design of a compact, lightweight device well-suited to clinical applications.

2. Materials and Methods

The future of medical diagnostics relies on devices that enable long-term patient monitoring, which allows for the detection of pathological conditions. The Lung Electrical Tomography System (LETS) is responding to the medical market’s demand. It is a mobile electrical impedance tomography system in three spatial dimensions for area monitoring. The system consists of the vest (Figure 1), the measuring device, and the analytics engine LETSWEB.

2.1. Hardware

The LETS system vest has 102 electrodes, 32 dedicated to electrical impedance tomography (EIT), arranged in two planes with 16 electrodes. Current is selectively injected between electrodes located on the same plane, and voltages are measured between adjacent electrodes, generating a data frame containing 896 voltage measurements. Additionally, the vest utilizes Body Surface Potential Mapping (BSPM) technology for continuously monitoring biopotentials [2,43], recording the electrical signals produced by the heart muscle. The EIT method applies low-current signals detected by the electrodes, while BSPM captures electrical signals from the body surface [43]. Data collected during these measurements is stored on a local memory device and transmitted to a database via Wi-Fi. The LETSWEB analytical module, a key component of the LETS system, is responsible for aggregating, processing, and analyzing medical data collected using specialized algorithms, aiding medical professionals in decision-making and enabling feedback.
In the LETS system vest, electrodes Figure 2 made of silicone conductive material with a specific resistivity = 50 Ωcm are employed. These electrodes offer better adaptation to body contours and enhanced adhesion, significantly reducing patient movement’s impact on the measured signal. Due to these properties, such electrodes are widely used in devices for electrical stimulation, such as TENS/EMS and electrical impedance tomography (EIT). The silicone material used for these electrodes also ensures high precision and reliability in biosignal measurements, contributing to improved quality of analyzed medical data.
The central unit of the device houses a single PCB (Figure 3, right side). The integration of modern advanced integrated circuits significantly reduces its dimensions. Using a high-frequency DC/DC converter to power the device ensures that the circuit remains compact while maintaining high efficiency. As a result, additional cooling is not required. These design choices contribute to the small size of the circuitry, which in turn reduces the mass of the PCB, components, and enclosure.
The system is powered by a lithium–polymer (Li-Po) cell (Figure 3 left side), which features a very high power density. The cell is encased in a thin pouch-type enclosure with extremely thin walls. These characteristics result in a relatively low mass compared with other solutions. Although this approach necessitates providing additional space in case the cell’s dimensions increase because of operational conditions, it does not affect the dimensions of the main device enclosure, thereby not increasing its mass; therefore, this solution is optimal considering parameters such as mass, power, and dimensions.
Electronic components were designed and selected to ensure minimal dimensions and weight while maintaining the device’s full functionality. The PCB measures 108.41 × 58.9 mm, and the selected cell has dimensions of 56 × 38 × 4 mm. An analysis of human anatomy determined that these dimensions are optimal for placing the device in the area around the sternum.

2.2. Models and Algorithms

A model of the male torso based on computer tomography (CT) images has been considered. Following the segmentation process to isolate the torso and lungs from the background, the next step involves generating a mesh composed of tetrahedral elements. A comprehensive description of the mesh generation procedure and the subsequent stages of data processing has been provided (Figure 4). The main steps are generating the finite element mesh, enumerating common and unique elements for disease models, defining material parameters, and solving the forward problem.
The initial stage of our algorithm involves importing a 3D image, which is expressed as a 3D matrix containing the values 0, 1, and 2. Each 0 denotes an element outside the field of view; 1 represents a human torso, and 2 signifies human lungs. The subsequent stage involves generating a mesh comprised of tetrahedrons.
The main goal of the article is to present models of respiratory diseases using a tetrahedral mesh model and classify them into the appropriate classes through electrical impedance tomography. We considered a model of a healthy person as well as models approximating diseases such as COPD, ARDS, PTX, PNA, Bronchospasm, and PHTN. Each model consists of up to seven regions, and their names and corresponding indices are presented in Table 1.
Figure 5 presents the central cross-sections of the model for a healthy case, showing projections onto the Oxy and Oyz planes. Image segmentation was employed to identify areas assigned to the lungs and bronchi. A morphological dilation operation was utilized to delineate the region approximating blood vessels. Figure 6 presents all diseases.
In Figure 6a, the lesion is the red region; it is damaged lungs for very small bronchioles. Figure 6b presents a model with ARDS lesions, which are small changes throughout the lungs, simulating extravascular water in the lungs. Figure 6c presents a reduced left lung, the air in the pleura (red color), and a healthy right lung. Figure 6d shows PNA lesions on two levels (index 6 and 7). Figure 6e simulates bronchospasm lesions with pathological change in distinct areas assigned to the bronchioles (index 4). Figure 6f is a model of PHTN disease; index 5 is responsible for enlarged blood vessels.

2.3. Dataset to Model Classification

The dataset represents seven different classes. First, they represent healthy persons, and the following classes represent ill persons. Cases representing lesions were considered, ranging from three to four:
  • Four for COPD (four levels of lung damage from 10% to 70% for very small bronchioles).
  • Three for ARDS (three different saturations of small changes throughout the lungs).
  • Four for PTX (two sizes of reduced lung separately for each lung).
  • Four for pneumonia (change at one level separately for each lung, change at one level for both lungs, and change at two levels for both lungs).
  • Four for bronchospasm (four levels of 10% to 70% reduction in airflow in the bronchi).
  • Four for PHTN (four levels of 10% to 70% narrowing of lung blood vessels).
In this way, a set of models with seven classes, generating twenty subclasses together, has been constructed. To obtain a classification database for each of these subclasses, a five-step process must be conducted.
  • Assignment of indices (1–7) to the individual components of the models.
  • Generation of respiratory phases using affine transformation for each model. The transform generates 90 frames with lungs of different sizes and shapes.
  • Assignment of material coefficients to indices Table 2.
  • For each respiratory phase, different material coefficients are assigned for healthy and reduced lungs in the case of pneumothorax.
  • Normalization of material coefficients concerning the reference (torso) Table 3.
  • Simulations of electrical impedance tomography measurements with the addition of noise.
The simulation voltages are obtained using the finite element method programmed in Python 3.10. The noise added to the simulation is normal random noise on a 2% level.
σbronchiα + (1 − α)σair
σ b r o n c h i α σ r e f + ( 1 − α ) σ air σ r e f
σblood vessel wallα + (1 − α)σblood
σ b l o o d   v e s s e l   w a l l α σ r e f + ( 1 − α ) σ b l o o d σ r e f
σlungα + (1 − α)σref
σ l u n g α σ r e f + ( 1 − α )
where σair = 10−10, σblood = 0.6625, σbronchi = 0.5576, σblood vessel wall = 0.2320, σref = 0.4610, σlung = 0.1111 [45]
Equations (1a), (2a) and (3a) have variable α, with a value taken from set {0.1, 0.3, 0.5, 0.7}. The variable α is accountable for the degree or advancement of the disease state.

3. Results

The electromagnetic compatibility tests were conducted in the electromagnetic compatibility laboratory. They measured the device’s electromagnetic emission.
The tests conducted as part of the tasks led to the following conclusions:
  • The simulator tested the LETS system on various available disease entities.
  • The LETS device results match the parameters measured using available medical devices.
  • The system could determine medical parameters from biosignal measurements and impedance tomography.
  • The expert system, part of LETS, could recognize all disease entities simulated as part of the tests in the operational environment.
Data processing in LETS involves analyzing biosignals and impedance measurements gathered from 102 electrodes, which collect detailed data on the patient’s lung health. Advanced algorithms were applied for accurate image reconstruction, transforming raw data into diagnostic images. Processing includes impedance-based reconstruction algorithms and data filtering to eliminate electromagnetic interference, which was verified during electromagnetic compatibility tests.
Model validation was conducted to ensure that the results generated by the LETS system align with those produced by available medical devices, demonstrating the model’s accurate reflection of medical parameters. The validation process involved testing the simulator with various disease entities, which allowed verification of the LETS expert system’s diagnostic accuracy. The LightGBM algorithm used in lung disease classification was tested for its ability to recognize all disease entities in operational conditions, thus ensuring high diagnostic effectiveness.
The methods for data processing and model validation were carefully selected to guarantee precise diagnostics while maintaining the device’s high energy efficiency.
In the classification of lung conditions, the Light Gradient Boosting Machine (LightGBM) was utilized for its fast training speed and high prediction accuracy. This model operates through continuous tree learning, leaf value calculation, and a top-leaf algorithm, which collectively help eliminate weak trees, enhancing overall performance [46].
The dataset used for training consisted of 10,500 cases, with 1500 cases per class, while the testing set comprised 3151 cases, approximately 450 cases per class. An analysis of class distributions indicated that most cases showed similar distributions, except for the PTX class. The ARDS group of patients displayed a higher median, highlighting significant inter-group differences (Figure 7). The diversity within distributions and the presence of outliers presented challenges in classification, reflecting the complex nature of these data.
The LightGBM model achieved a high classification accuracy of 91.8% (Figure 8). The highest rate of misclassification was observed between healthy cases and bronchospasm disease, attributable to the minimal differences between the models for these cases, which are distinguished primarily by a single value linked to the bronchi region in the lungs.
To enhance feature selection, mutual information estimation was applied, reducing the number of variables by assessing both linear and non-linear dependencies. This quantitative measure helped identify the features that most significantly impact the target variable, ensuring that the model focuses on the most informative features for classification tasks.
Shapley values were used to explain the influence of individual features on LightGBM’s predictions. Each feature value was conceptualized as a “player” in a cooperative game, where the overall prediction outcome represents the “reward”. Shapley values fairly distribute this reward among features, determining their contributions to the final classification. An analysis of the top three influential features demonstrated their impact on disease classification, revealing that only a specific subset of electrodes notably influences classification outcomes [47].
Analysis was performed for three of the most influential measures, showing their impact on disease classification. The results revealed that only a limited subset of electrodes significantly impacts disease classification. Figure 9 shows the distribution of SHAP values for individual features in COPD, where the top three features show a significantly larger impact on classification.
Furthermore, this selective influence underscores the potential for refining feature selection in disease classification models by prioritizing the most impactful measurements. By leveraging Shapley values, we can more accurately identify which features are critical for precise predictions, thereby enhancing the model’s interpretability and performance. This targeted approach is essential for optimizing diagnostic tools and advancing the field of machine learning in medical applications. The electrodes with the highest impact on classification are:
  • Healthy: 0, 1, 2, 3, 7, 8, 17, 18, 19, 21, 22, 31.
  • COPD: 2, 18, 19, 20, 21, 22, 25, 26.
  • ARDS: 0, 6, 7, 8, 9, 16, 18, 19.
  • PTX: 1, 2, 7, 8, 17, 18, 19, 20, 23, 24.
  • PHTN: 0, 2, 12, 13, 16, 18, 21, 22, 23, 24.
  • PNA: 2, 13, 14, 18, 21, 22, 25, 26.
  • Bronchospasm: 1, 2, 3, 4, 5, 6, 17, 18, 21, 22.
This indicates that electrodes with indices 10, 11, 15, 27, 28, 29, and 30 do not significantly impact disease classification. These results reveal that despite a wide range of potential features, only a limited subset of electrodes contributes substantially to the differentiation between different disease states. These results highlight the importance of focusing on these key electrodes to enhance the precision and effectiveness of the diagnostic model.
The results demonstrate that only a limited subset of electrodes has a substantial influence on differentiating between disease states. Concentrating on these key electrodes enables improved classification performance and more precise diagnostics, supporting a targeted approach in disease classification models. This methodology is essential for optimizing diagnostic tools and advancing machine-learning applications in medical diagnostics.

4. Discussion and Conclusions

The implementation of energy-efficient solutions in electrical impedance tomography systems represents a significant advancement in medical diagnostics, especially in the non-invasive detection of lung diseases. The Lung Electrical Tomography System (LETS), developed as part of this study, demonstrates a strong balance between diagnostic accuracy and power optimization, addressing one of the key challenges in deploying medical devices for long-term patient monitoring.
One of the key findings of the research is the ability to maintain classification precision using machine learning algorithms, such as LightGBM, while significantly reducing the energy required for electromagnetic measurements and data processing. This improvement not only increases the operational efficiency of the system but also responds to the growing need for sustainable solutions in medical technology. By reducing the device’s energy demand, long-term patient monitoring can be achieved with less frequent maintenance or battery replacements, enhancing patient comfort and lowering operational costs in clinical settings.
From a technical perspective, the selective use of electrodes during measurements plays a crucial role in reducing unnecessary energy consumption while maintaining the integrity of diagnostic results. The study’s findings highlight the importance of focusing on the most influential measurements supported by algorithms that identify key features contributing to accurate classification. This approach enables a streamlined diagnostic process, avoiding the burdens associated with more traditional, energy-intensive imaging methods.
Research indicates that the Lung Electrical Tomography System (LETS) can effectively classify conditions such as COPD, ARDS, pneumothorax, pneumonia, bronchospasm, and pulmonary hypertension, using less energy than conventional diagnostic methods. Specifically, the LightGBM model achieved a classification accuracy of 91.8%, demonstrating the high diagnostic efficacy of the system.
Additionally, the research identified specific electrodes that are most relevant to the classification process, contributing to the system’s energy efficiency. Focusing on key measurements can lead to further refinement of the classification process and optimization of diagnostic tools. Future plans include clinical trials and the integration of real-time patient data, which will allow for further improvement of algorithms and system performance.
This study contributes to the ongoing development of energy-efficient diagnostic tools and paves the way for the broader use of electrical impedance tomography in clinical environments. The LETS system sets a new standard for sustainable medical imaging technologies by reducing energy consumption without compromising diagnostic capabilities. The ultimate goal is to expand the application of EIT in a wider medical context, offering a sustainable, non-invasive, and cost-effective diagnostic solution that meets the growing demands of modern healthcare.

Author Contributions

Development of the concept, methodology, algorithms, and supervision B.S., T.R. and D.W.; development of the measurement, data acquisition M.O. and M.K.; development of research methodology and image reconstruction B.S., T.C. and Z.O.; literature review, formal analysis, writing, project administration, general review, visualization and editing of the manuscript, M.C.-W., J.G. and E.G. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Data Availability Statement

The raw data supporting the conclusions of this article will be made available by the corresponding author.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Przysucha, B.; Wójcik, D.; Rymarczyk, T.; Król, K.; Kozłowski, E.; Gąsior, M. Analysis of Reconstruction Energy Efficiency in EIT and ECT 3D Tomography Based on Elastic Net. Energies 2023, 16, 1490. [Google Scholar] [CrossRef]
  2. Baran, B.; Kozłowski, E.; Majerek, D.; Rymarczyk, T.; Soleimani, M.; Wójcik, D. Application of Machine Learning Algorithms to the Discretization Problem in Wearable Electrical Tomography Imaging for Bladder Tracking. Sensors 2023, 23, 1553. [Google Scholar] [CrossRef] [PubMed]
  3. Kulisz, M.; Kłosowski, G.; Rymarczyk, T.; Słoniec, J.; Gauda, K.; Cwynar, W. Optimizing the Neural Network Loss Function in Electrical Tomography to Increase Energy Efficiency in Industrial Reactors. Energies 2024, 7, 681. [Google Scholar] [CrossRef]
  4. Kulisz, M.; Kłosowski, G.; Rymarczyk, T.; Hoła, A.; Niderla, K.; Sikora, J. The Use of the Multi-Sequential LSTM in Electrical Tomography for Masonry Wall Moisture Detection. Measurement 2024, 234, 114860. [Google Scholar] [CrossRef]
  5. Wang, Z.; Su, X.; Zeng, N.; Jiang, J. Overview of Isolated Bidirectional DC–DC Converter Topology and Switching Strategies for Electric Vehicle Applications. Energies 2024, 17, 2434. [Google Scholar] [CrossRef]
  6. Panskyi, T.; Korzeniewska, E.; Firych-Nowacka, A. Educational Data Clustering in Secondary School Sensor-Based Engineering Courses Using Active Learning Approaches. Appl. Sci. 2024, 14, 5071. [Google Scholar] [CrossRef]
  7. Maciura, Ł.; Wójcik, D.; Rymarczyk, T.; Król, K. Novel hybrid algorithm using convolutional autoencoder with Svm for electrical impedance tomography and ultrasound computed tomography. Inform. Autom. Pomiary Gospod. Ochr. Śr. 2023, 13, 4–9. [Google Scholar] [CrossRef]
  8. Jassem, E.; Górecka, D. Ciężka i Terminalna Postać Przewlekłej Obturacyjnej Choroby Płuc, Severe and Terminal Chronic Obstructive Pulmonary Disease. Neumonol. Alergol. Pol. 2009, 77, 411–416. [Google Scholar]
  9. Górecka, D.; Jassem, E.; Pierzchała, W.; Śliwiński, P. Zalecenia Polskiego Towarzystwa Chorób Płuc Dotyczące Rozpoznawania i Leczenia Przewlekłej Obturacyjnej Choroby Płuc (POChP). Pneumonol. Alergol. Pol. 2012, 80, 220–254. [Google Scholar] [CrossRef]
  10. Antó, J.M.; Vermeire, P.; Vestbo, J.; Sunyer, J. Epidemiology of Chronic Obstructive Pulmonary Disease. Eur. Respir. J. 2001, 17, 982–994. [Google Scholar] [CrossRef]
  11. Przewlekła Obturacyjna Choroba Płuc (POChP)—Objawy i Profilaktyka 2021. Available online: http://www.mp.pl/social/article/54198 (accessed on 24 November 2023).
  12. European Lung Foundation Zespół Ostrej Niewydolności Oddechowej (ARDS). 2021. Available online: https://europeanlung.org/pl/information-hub/lung-conditions/zespol-ostrej-niewydolnosci-oddechowej-ards/ (accessed on 27 July 2021).
  13. Vignon, P.; Evrard, B.; Asfar, P.; Busana, M.; Calfee, C.S.; Coppola, S.; Demiselle, J.; Geri, G.; Jozwiak, M.; Martin, G.S.; et al. Fluid Administration and Monitoring in ARDS: Which Management? Intensive Care Med. 2020, 46, 2252–2264. [Google Scholar] [CrossRef] [PubMed]
  14. Neamu, R.F.; Martin, G.S. Fluid Management in Acute Respiratory Distress Syndrome. Curr. Opin. Crit. Care 2013, 19, 24–30. [Google Scholar] [CrossRef] [PubMed]
  15. Pulmonologia. Odma Opłucnowa. 2016. Available online: https://www.mp.pl/pacjent/pulmonologia/choroby/150463,odma-oplucnowa (accessed on 18 October 2016).
  16. Barton, E.C.; Maskell, N.A.; Walker, S.P. Expert Review on Spontaneous Pneumothorax: Advances, Controversies, and New Directions. Semin. Respir. Crit. Care Med. 2023, 44, 426–436. [Google Scholar] [CrossRef] [PubMed]
  17. Pulmonologia. Zapalenie Płuc—Objawy, Przyczyny, Leczenie, Rodzaje. 2023. Available online: https://www.mp.pl/pacjent/pulmonologia/choroby/65040,zapalenie-pluc (accessed on 1 October 2024).
  18. Cleveland Clinic. Bronchospasm: Symptoms, Treatment & What It Is. 2022. Available online: https://my.clevelandclinic.org/health/diseases/22620-bronchospasm (accessed on 18 March 2022).
  19. Sahni, S.; Ojrzanowski, M.; Majewski, S.; Talwar, A. Tętnicze Nadciśnienie Płucne: Aktualny Przegląd Leczenia. Pneumonol. Alergol. Pol. 2016, 1, A25–A40. [Google Scholar] [CrossRef]
  20. Wieteska-Miłek, M. Informator dla Pacjentów: Wszystko o Tętniczym Nadciśnieniu Płucnym, 4th ed.; Klinika Krążenia Płucnego, Chorób Zakrzepowo-Zatorowych i Kardiologii CMKP, Europejskie Centrum Zdrowia Otwock: Warszawa, Poland, 2018. [Google Scholar]
  21. Wikipedia. Pulmonary Hypertension. 2023. Available online: https://en.wikipedia.org/wiki/Pulmonary_hypertension (accessed on 20 November 2023).
  22. Global Initiative for Chronic Obstructive Lung Disease. Global Strategy for Prevention, Diagnosis and Management of Chronic Obstructive Pulmonary Disease: 2022 Report; GOLD: Deer Park, IL, USA, 2023. [Google Scholar]
  23. Pellegrino, R.; Viegi, G.; Brusasco, V.; Crapo, R.O.; Burgos, F.; Casaburi, R.E.A.; Coates, A.; Van Der Grinten, C.P.M.; Gustafsson, P.; Hankinson, J.; et al. Interpretative Strategies for Lung Function Tests. Eur. Respir. J. 2005, 26, 948–968. [Google Scholar] [CrossRef]
  24. Wanger, J.; Clausen, J.L.; Coates, A.; Pedersen, O.F.; Brusasco, V.; Burgos, F.; Casaburi, R.; Crapo, R.; Enright, P.; Van Der Grinten, C.P.M.; et al. Standardisation of the Measurement of Lung Volumes. Eur. Respir. J. 2005, 26, 511–522. [Google Scholar] [CrossRef]
  25. Celli, B.R.; MacNee, W.; Force, A.T. Standards for the Diagnosis and Treatment of Patients with COPD: A Summary of the ATS/ERS Position Paper. Eur. Respir. J. 2004, 23, 932–946. [Google Scholar] [CrossRef]
  26. Villar, J.; Pérez-Méndez, L.; Blanco, J.; Añón, J.M.; Blanch, L.; Belda, J.; Santos-Bouza, A.; Fernández, R.L.; Kacmarek, R.M. A Universal Definition of ARDS: The PaO2/FiO2 Ratio under a Standard Ventilatory Setting—A Prospective, Multicenter Validation Study. Intensive Care Med. 2013, 39, 583–592. [Google Scholar] [CrossRef]
  27. Gupta, D.; Hansell, A.; Nichols, T.; Duong, T.; Ayres, J.G.; Strachan, D. Epidemiology of Pneumothorax in England. Thorax 2000, 55, 666–671. [Google Scholar] [CrossRef]
  28. Hoeper, M.M.; Bogaard, H.J.; Condliffe, R.; Frantz, R.; Khanna, D.; Kurzyna, M.; Langleben, D.; Manes, A.; Satoh, T.; Torres, F. Definitions and Diagnosis of Pulmonary Hypertension. J. Am. Coll. Cardiol. 2013, 62, D42–D50. [Google Scholar] [CrossRef]
  29. Mets, O.M.; Schmidt, M.; Buckens, C.F.; Gondrie, M.J.; Isgum, I.; Oudkerk, M.; Vliegenthart, R.; de Koning, H.J.; van der Aalst, C.M.; Prokop, M. Diagnosis of Chronic Obstructive Pulmonary Disease in Lung Cancer Screening Computed Tomography Scans: Independent Contribution of Emphysema, Air Trapping and Bronchial Wall Thickening. Respir. Res. 2013, 14, 59. [Google Scholar] [CrossRef] [PubMed]
  30. Gattinoni, L.; Pelosi, P.; Suter, P.M.; Pedoto, A.; Vercesi, P.; Lissoni, A. Acute Respiratory Distress Syndrome Caused by Pulmonary and Extrapulmonary Disease. Different Syndromes? Am. J. Respir. Crit. Care Med. 1998, 158, 3–11. [Google Scholar] [CrossRef] [PubMed]
  31. MacDuff, A.; Arnold, A.; Harvey, J. Management of Spontaneous Pneumothorax: British Thoracic Society Pleural Disease Guideline 2010. Thorax 2010, 65 (Suppl. S2), ii18–ii31. [Google Scholar] [CrossRef] [PubMed]
  32. Torres Martí, A.; Lee, N.; Cillóniz, C.; Vila Estapé, J.; Van der Eerden, M. Laboratory Diagnosis of Pneumonia in the Molecular Age. Eur. Respir. J. 2016, 48, 1764–1778. [Google Scholar] [CrossRef] [PubMed]
  33. Global Initiative for Asthma. 2022 GINA Report, Global Strategy for Asthma Management and Prevention 2022. Available online: https://ginasthma.org/pocket-guide-for-asthma-management-and-prevention/ (accessed on 27 November 2023).
  34. Rajaram, S.; Swift, A.J.; Telfer, A.; Hurdman, J.; Marshall, H.; Lorenz, E.; Capener, D.; Davies, C.; Hill, C.; Elliot, C. 3D Contrast-Enhanced Lung Perfusion MRI Is an Effective Screening Tool for Chronic Thromboembolic Pulmonary Hypertension: Results from the ASPIRE Registry. Thorax 2013, 68, 677–678. [Google Scholar] [CrossRef]
  35. Lynch, D.A.; Austin, J.H.; Hogg, J.C.; Grenier, P.A.; Kauczor, H.U.; Bankier, A.A.; Barr, R.G.; Colby, T.V.; Galvin, J.R.; Gevenois, P.A. CT-Definable Subtypes of Chronic Obstructive Pulmonary Disease: A Statement of the Fleischner Society. Radiology 2015, 277, 192–205. [Google Scholar] [CrossRef] [PubMed]
  36. Lichtenstein, D.A.; Mezière, G.A. Relevance of Lung Ultrasound in the Diagnosis of Acute Respiratory Failure: The BLUE Protocol. Chest 2008, 134, 117–125. [Google Scholar] [CrossRef]
  37. Ding, W.; Shen, Y.; Yang, J.; He, X.; Zhang, M. Diagnosis of Pneumothorax by Radiography and Ultrasonography: A Meta-Analysis. Chest 2011, 140, 859–866. [Google Scholar] [CrossRef]
  38. McKnight, C.L.; Burns, B. Pneumothorax; Statpearls: Tampa, FL, USA, 2022. [Google Scholar]
  39. Metlay, J.P.; Waterer, G.W.; Long, A.C.; Anzueto, A.; Brozek, J.; Crothers, K.; Cooley, L.A.; Dean, N.C.; Fine, M.J.; Flanders, S.A. Diagnosis and Treatment of Adults with Community-Acquired Pneumonia. An Official Clinical Practice Guideline of the American Thoracic Society and Infectious Diseases Society of America. Am. J. Respir. Crit. Care Med. 2019, 200, e45–e67. [Google Scholar] [CrossRef]
  40. Murdoch, D.R.; Laing, R.T.; Mills, G.D.; Karalus, N.C.; Town, G.I.; Mirrett, S.; Reller, L.B. Evaluation of a Rapid Immunochromatographic Test for Detection of Streptococcus Pneumoniae Antigen in Urine Samples from Adults with Community-Acquired Pneumonia. J. Clin. Microbiol. 2001, 39, 3495–3498. [Google Scholar] [CrossRef]
  41. Cockcroft, D.W.; Davis, B.E. Mechanisms of Airway Hyperresponsiveness. J. Allergy Clin. Immunol. 2006, 118, 551–559. [Google Scholar] [CrossRef] [PubMed]
  42. Dweik, R.A.; Boggs, P.B.; Erzurum, S.C.; Irvin, C.G.; Leigh, M.W.; Lundberg, J.O.; Olin, A.C.; Plummer, A.L.; Taylor, D.R. An Official ATS Clinical Practice Guideline: Interpretation of Exhaled Nitric Oxide Levels (FENO) for Clinical Applications. Am. J. Respir. Crit. Care Med. 2011, 184, 602–615. [Google Scholar] [CrossRef] [PubMed]
  43. Kiczek, B.; Wójcik, D.; Oleszek, M.; Rymarczyk, T.; Król, K.; Kozłowski, E. Electrical Impedance Tomography Bladder Monitoring Wearable Device. In Proceedings of the 2022 ACM International Joint Conference on Pervasive and Ubiquitous Computing, Atlanta, GA, USA, 11–15 September 2022; ACM: Cambridge, UK, 2022; pp. 63–64. [Google Scholar]
  44. Rymarczyk, T. Wearable Sensor for Biopotential Measurements of Patients’ Health Monitoring. Przegląd Elektrotechniczny 2020, 1, 101–104. [Google Scholar] [CrossRef]
  45. Hasgall, P.A.; Gennaro, F.D.; Baumgartner, C.; Neufeld, E.; Lloyd, B.; Gosselin, M.C.; Payne, D.; Klingenböck, A.; Kuster, N. IT’IS Database for Thermal and Electromagnetic Parameters of Biological Tissues 2022. Available online: https://itis.swiss/virtual-population/tissue-properties/database/dielectric-properties (accessed on 22 February 2022).
  46. Ke, G.; Meng, Q.; Finley, T. LightGBM: A Highly Efficient Gradient Boosting Implementation. arXiv 2017, arXiv:170601952. [Google Scholar]
  47. Molnar, C. Interpretable Machine Learning; Leanpub: Victoria, BC, Canada, 2020. [Google Scholar]
Energies 17 05828 g001
Figure 1. Developed vest with 102 textile electrodes, A—central unit of the device (source own).
Figure 1. Developed vest with 102 textile electrodes, A—central unit of the device (source own).
Energies 17 05828 g001
Energies 17 05828 g002
Figure 2. Electrode schema [44].
Figure 2. Electrode schema [44].
Energies 17 05828 g002
Energies 17 05828 g003
Figure 3. The central unit of the device (signature letter A in Figure 1) (source own).
Figure 3. The central unit of the device (signature letter A in Figure 1) (source own).
Energies 17 05828 g003
Energies 17 05828 g004
Figure 4. Block diagram of the EIT data frame simulation process (source own).
Figure 4. Block diagram of the EIT data frame simulation process (source own).
Energies 17 05828 g004
Energies 17 05828 g005
Figure 5. Healthy lungs model (source own).
Figure 5. Healthy lungs model (source own).
Energies 17 05828 g005
Energies 17 05828 g006
Figure 6. Models of considered medical lesions (source own).
Figure 6. Models of considered medical lesions (source own).
Energies 17 05828 g006
Energies 17 05828 g007
Figure 7. Boxplot for the first feature for all classes (source own).
Figure 7. Boxplot for the first feature for all classes (source own).
Energies 17 05828 g007
Energies 17 05828 g008
Figure 8. Confusion matrix for LightGBM model. 0—healthy, 1—COPD, 2—ARDS, 3—PTX, 4—PHTN, 5—PNA, and 6—Bronchospasm (source own).
Figure 8. Confusion matrix for LightGBM model. 0—healthy, 1—COPD, 2—ARDS, 3—PTX, 4—PHTN, 5—PNA, and 6—Bronchospasm (source own).
Energies 17 05828 g008
Energies 17 05828 g009
Figure 9. Beeswarm of SHAP values for COPD disease (source own).
Figure 9. Beeswarm of SHAP values for COPD disease (source own).
Energies 17 05828 g009
Table 1. Regions comprising the models.
Table 1. Regions comprising the models.
Area NameArea Index
Out of model0
Torso1
Left lung2
Right lung3
Lungs bronchi4
Blood vessels around the lung’s bronchi5
First lesion6
Second lesion7
Table 2. Coefficient of materials parameters [45].
Table 2. Coefficient of materials parameters [45].
Area IndexHealthy PatientCOPDARDSPTXPNABronchial SpasmPHTN
10.46100.46100.46100.46100.46100.46100.4610
20.11110.11110.11110.12110.11110.11110.1111
30.11110.11110.11110.11110.11110.11110.1111
410−1010−1010−1010−1010−10(1a)10−10
50.66250.66250.66250.66250.66250.6625(2a)
6—(3a)2.679110−100.0016——
7————0.0032——
Table 3. Coefficients material parameters normalized.
Table 3. Coefficients material parameters normalized.
Area IndexHealthy PatientCOPDARDSPTXPNABronchial SpasmPHTN
11.01.01.01.01.01.01.0
20.24100.24100.24100.26270.24100.24100.2410
30.24100.24100.24100.24100.24100.24100.2410
42.2·10−102.2·10−102.2·10−102.2·10−102.2·10−10(1b)2.2·10−10
51.43701.43701.43701.43701.43701.4370(2b)
6—(3b)5.81152.2·10−100.0035——
7————0.0070——
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Stefaniak, B.; Rymarczyk, T.; Wójcik, D.; Cholewa-Wiktor, M.; Cieplak, T.; Orzeł, Z.; Gudowski, J.; Golec, E.; Oleszek, M.; Kowalski, M. Energy Efficiency in Measurement and Image Reconstruction Processes in Electrical Impedance Tomography. Energies 2024, 17, 5828. https://doi.org/10.3390/en17235828

AMA Style

Stefaniak B, Rymarczyk T, Wójcik D, Cholewa-Wiktor M, Cieplak T, Orzeł Z, Gudowski J, Golec E, Oleszek M, Kowalski M. Energy Efficiency in Measurement and Image Reconstruction Processes in Electrical Impedance Tomography. Energies. 2024; 17(23):5828. https://doi.org/10.3390/en17235828

Chicago/Turabian Style

Stefaniak, Barbara, Tomasz Rymarczyk, Dariusz Wójcik, Marta Cholewa-Wiktor, Tomasz Cieplak, Zbigniew Orzeł, Janusz Gudowski, Ewa Golec, Michał Oleszek, and Marcin Kowalski. 2024. "Energy Efficiency in Measurement and Image Reconstruction Processes in Electrical Impedance Tomography" Energies 17, no. 23: 5828. https://doi.org/10.3390/en17235828

APA Style

Stefaniak, B., Rymarczyk, T., Wójcik, D., Cholewa-Wiktor, M., Cieplak, T., Orzeł, Z., Gudowski, J., Golec, E., Oleszek, M., & Kowalski, M. (2024). Energy Efficiency in Measurement and Image Reconstruction Processes in Electrical Impedance Tomography. Energies, 17(23), 5828. https://doi.org/10.3390/en17235828

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop