Vitalii Babenko
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Vitalii Babenko has a total of 55 co-authors. Here are the top co-authors by number of shared publications:
- Ievgen Nastenko
Igor Sikorsky Kyiv Polytechnic Institute
6 shared publications - Olena Nosovets
National Technical University of Ukraine “Igor Sikorsky Kyiv Polytechnic Institute
5 shared publications - Volodymyr Pavlov
Національний технічний університет України «Київський політехнічний інститут імені Ігоря Сікорського
5 shared publications - Maryna Vasylivna Globa
SHUPYK NATIONAL HEALTHCARE UNIVERSITY OF UKRAINE, KYIV, UKRAINE
3 shared publications - Володимир Солодущенко
SI “Ukrainian Center of Maternity and Childhood of NAMS of Ukraine”, Kyiv, Ukraine; Shupyk National Healthcare University of Ukraine, Kyiv, Ukraine
3 shared publications - Iryna Mykolaivna Dykan
ДУ "Інститут ядерної медицини та променевої діагностики НАМН України
2 shared publications
Latest Publications
- Open Access
- 30 September 2026
- Published by Institute for Medical Radiology and Oncology of NAMS of Ukraine in Український радіологічний та онкологічний журнал
- Vol. 34 (3)
Multiparametric ultrasound phenotyping of paediatric liver cirrhosis for future artificial intelligence models
Background. Multiparametric ultrasonographic characterization of cirrhotic liver involvement may provide standardized information for development of interpretable clinical decision support systems. Purpose – to identify informative structural and haemodynamic characteristics and establish parameters for future artificial intelligence modelling. Materials and methods. This case-control investigation included 32 children aged 6–18 years with cirrhotic liver involvement and 30 controls. Aetiology was autoimmune hepatitis in 19 patients and Wilson disease in 13. B-mode ultrasonography, Doppler assessment and two-dimensional shear wave elastography were performed. Statistical estimates were calculated from aggregate measurements. Results. Left-lobe enlargement occurred in 65.6% of patients versus 10.0% of controls; segment I enlargement, in 43.8% versus 6.7%; surface irregularity, in 40.6% versus 0%; umbilicalvein recanalization, in 21.9% versus 0%; and splenomegaly, in 50.0% versus 3.3%. Triphasic hepatic venous waveforms were absent in all patients but present in 90.0% of controls. Absence of triphasicity showed 100.0% sensitivity (95% confidence interval 89.3–100.0%) and 90.0% specificity (74.4–96.5%). Portal venous velocity was lower and proper hepatic arterial resistive index was higher among patients in both age subgroups. All four comparisons: p < 0.001 (Welch test). Surface irregularity, monophasic venous waveforms and umbilical-vein recanalization each demonstrated 100.0% specificity. Conclusions. The identified characteristics describe structural and haemody-namic alterations in the investigated population and offer candidate parameters for interpretable modelling. Diagnostic estimates require independent confirmation; artificial intelligence-based fibrosis staging was not evaluated. The parameter set requires prospective multicentre evaluation against a reference standard.
- 30 September 2026
- Published by Institute for Medical Radiology and Oncology of NAMS of Ukraine in Український радіологічний та онкологічний журнал
- Vol. 34 (3)
Open AccessMultiparametric ultrasound phenotyping of paediatric liver cirrhosis for future artificial intelligence models
Background. Multiparametric ultrasonographic characterization of cirrhotic liver involvement may provide standardized information for development of interpretable clinical decision support systems. Purpose – to identify informative structural and haemodynamic characteristics and establish parameters for future artificial intelligence modelling. Materials and methods. This case-control investigation included 32 children aged 6–18 years with cirrhotic liver involvement and 30 controls. Aetiology was autoimmune hepatitis in 19 patients and Wilson disease in 13. B-mode ultrasonography, Doppler assessment and two-dimensional shear wave elastography were performed. Statistical estimates were calculated from aggregate measurements. Results. Left-lobe enlargement occurred in 65.6% of patients versus 10.0% of controls; segment I enlargement, in 43.8% versus 6.7%; surface irregularity, in 40.6% versus 0%; umbilicalvein recanalization, in 21.9% versus 0%; and splenomegaly, in 50.0% versus 3.3%. Triphasic hepatic venous waveforms were absent in all patients but present in 90.0% of controls. Absence of triphasicity showed 100.0% sensitivity (95% confidence interval 89.3–100.0%) and 90.0% specificity (74.4–96.5%). Portal venous velocity was lower and proper hepatic arterial resistive index was higher among patients in both age subgroups. All four comparisons: p < 0.001 (Welch test). Surface irregularity, monophasic venous waveforms and umbilical-vein recanalization each demonstrated 100.0% specificity. Conclusions. The identified characteristics describe structural and haemody-namic alterations in the investigated population and offer candidate parameters for interpretable modelling. Diagnostic estimates require independent confirmation; artificial intelligence-based fibrosis staging was not evaluated. The parameter set requires prospective multicentre evaluation against a reference standard.
- Open Access
- 11 September 2026
- Published by State Institution of Science "Center of Innovative Healthcare Technologies" State Administrative Department in Clinical and Preventive Medicine
- No. 6,p. 102-115
INTERPRETABLE AI-DRIVEN ULTRASOUND ANALYSIS FOR THE SELECTION OF PEDIATRIC PATIENTS FOR LIVER ELASTOGRAPHY: A PILOT STUDY
Introduction. Standard two-dimensional B-mode liver ultrasound is the most accessible modality for primary diagnosis; however, its conventional visual interpretation in pediatric chronic diffuse liver diseases is limited by subjectivity and low sensitivity to early-stage fibrosis. Shear wave elastography requires specialized equipment and is frequently unavailable in primary care settings. The development of interpretable artificial intelligence (AI) tools for objective computer-aided analysis of echograms enables automated patient triage and rational identification of priority cohorts for referral to liver elastography.Aim. To evaluate the diagnostic performance and clinical interpretability of a developed AI-based computer tool designed for the automated triage of children with chronic diffuse liver diseases prior to elastographic examination.Materials and methods: In a retrospective single-center study, the clinical echogram archive of 105 children aged 5–18 years was analyzed. The total dataset comprised 661 regions of interest (ROIs) extracted from 337 two-dimensional B-mode ultrasound images acquired via expert-class ultrasound systems. Two-dimensional shear wave elastography (2D-SWE) served as the reference standard (stratification into stages F0 vs. F1–F4 based on verified kPa thresholds). A logistic regression model was trained using local grayscale intensity profiles (32 bins) to construct spatial local-support maps and compute an integrated patient score. Validation was performed using a 5-fold cross-validation scheme configured to strictly prevent data leakage (patient-level splitting). Additionally, a selective classification analysis was conducted by introducing an indeterminate/uncertainty zone (rejection option) for equivocal cases.Results. In the binary task of differentiating normal tissue (F0) from fibrotic changes (F1–F4) at the patient level, the model achieved an area under the ROC curve (AUC) of 0.91 [95% confidence interval: 0.84–0.96], a balanced accuracy of 0.83 [0.76–0.90], a sensitivity of 0.80 [0.68–0.91], and a specificity of 0.86 [0.77–0.95] at a standard operating threshold of 0.50. Implementing a screening threshold of 0.32 increased the sensitivity for detecting fibrotic changes to 0.91. Exploratory selective classification analysis revealed that routing equivocal cases to the indeterminate category captured 15 out of 17 (88%) baseline model errors. After filtering out these indeterminate cases, the balanced accuracy on the retained subset of patients increased to 0.90, and specificity reached 1.00 with a sensitivity of 0.80.Conclusions. The developed intelligent method provides high triage accuracy for children requiring subsequent elastographic fibrosis stratification. Spatial local-support maps render the model's logic clinically interpretable for clinicians by providing a visual explanation of the calculated score. While the selective classification method effectively routes equivocal cases to an expert review zone, the low coverage (only 31.4% of patients remained outside the indeterminate zone) limits its utility as a standalone screening tool.
- 11 September 2026
- Published by State Institution of Science "Center of Innovative Healthcare Technologies" State Administrative Department in Clinical and Preventive Medicine
- No. 6,p. 102-115
Open AccessINTERPRETABLE AI-DRIVEN ULTRASOUND ANALYSIS FOR THE SELECTION OF PEDIATRIC PATIENTS FOR LIVER ELASTOGRAPHY: A PILOT STUDY
Introduction. Standard two-dimensional B-mode liver ultrasound is the most accessible modality for primary diagnosis; however, its conventional visual interpretation in pediatric chronic diffuse liver diseases is limited by subjectivity and low sensitivity to early-stage fibrosis. Shear wave elastography requires specialized equipment and is frequently unavailable in primary care settings. The development of interpretable artificial intelligence (AI) tools for objective computer-aided analysis of echograms enables automated patient triage and rational identification of priority cohorts for referral to liver elastography.Aim. To evaluate the diagnostic performance and clinical interpretability of a developed AI-based computer tool designed for the automated triage of children with chronic diffuse liver diseases prior to elastographic examination.Materials and methods: In a retrospective single-center study, the clinical echogram archive of 105 children aged 5–18 years was analyzed. The total dataset comprised 661 regions of interest (ROIs) extracted from 337 two-dimensional B-mode ultrasound images acquired via expert-class ultrasound systems. Two-dimensional shear wave elastography (2D-SWE) served as the reference standard (stratification into stages F0 vs. F1–F4 based on verified kPa thresholds). A logistic regression model was trained using local grayscale intensity profiles (32 bins) to construct spatial local-support maps and compute an integrated patient score. Validation was performed using a 5-fold cross-validation scheme configured to strictly prevent data leakage (patient-level splitting). Additionally, a selective classification analysis was conducted by introducing an indeterminate/uncertainty zone (rejection option) for equivocal cases.Results. In the binary task of differentiating normal tissue (F0) from fibrotic changes (F1–F4) at the patient level, the model achieved an area under the ROC curve (AUC) of 0.91 [95% confidence interval: 0.84–0.96], a balanced accuracy of 0.83 [0.76–0.90], a sensitivity of 0.80 [0.68–0.91], and a specificity of 0.86 [0.77–0.95] at a standard operating threshold of 0.50. Implementing a screening threshold of 0.32 increased the sensitivity for detecting fibrotic changes to 0.91. Exploratory selective classification analysis revealed that routing equivocal cases to the indeterminate category captured 15 out of 17 (88%) baseline model errors. After filtering out these indeterminate cases, the balanced accuracy on the retained subset of patients increased to 0.90, and specificity reached 1.00 with a sensitivity of 0.80.Conclusions. The developed intelligent method provides high triage accuracy for children requiring subsequent elastographic fibrosis stratification. Spatial local-support maps render the model's logic clinically interpretable for clinicians by providing a visual explanation of the calculated score. While the selective classification method effectively routes equivocal cases to an expert review zone, the low coverage (only 31.4% of patients remained outside the indeterminate zone) limits its utility as a standalone screening tool.
- Open Access
- 20 January 2026
- Published by State Institution of Science "Center of Innovative Healthcare Technologies" State Administrative Department in Clinical and Preventive Medicine
- No. 1,p. 104-112
DIAGNOSTIC PERFORMANCE OF MULTIPARAMETRIC ULTRASOUND WITH ARTIFICIAL INTELLIGENCE TOOLS IN PEDIATRIC DIFFUSE LIVER DISEASES
Introduction. Chronic diffuse liver diseases in children require safe noninvasive assessment and repeated monitoring of fibrosis. Liver biopsy remains the reference method for morphological verification; however, in pediatric practice its use is limited by invasiveness, risk of complications, and ethical considerations. Multiparametric ultrasound, including shear wave elastography and Doppler assessment of blood flow, is suitable for longitudinal follow-up, yet interpretation of ultrasound features remains operator-dependent. Automated image analysis using artificial intelligence algorithms may increase objectivity and reproducibility of fibrosis stratification and support clinical decision-making. Aim. To evaluate the diagnostic accuracy of multiparametric ultrasound, including two-dimensional shear wave elastography and machine learning algorithms, for staging liver fibrosis in children. Materials and methods: This retrospective single-center study included 185 children (140 with chronic diffuse liver diseases and 45 controls). All participants underwent two-dimensional shear wave elastography, hepatic vein Doppler assessment, and grayscale ultrasound image analysis using two machine learning approaches: Model A (random forest algorithm; binary classification) and Model B (hierarchical ensemble; multiclass classification). The reference approach was clinical verification and risk stratification based on elastography findings. Results. A strong positive association was observed between elastography measures and Model B predictions (Spearman rank correlation coefficient 0.77). Model A demonstrated 84% specificity in differentiating normal findings from pathological changes. Model B showed higher sensitivity (83%) but tended to overestimate fibrosis at the boundary between minimal and moderate stages. Doppler patterns, particularly a monophasic hepatic vein waveform, were significantly associated with higher fibrosis stages. Conclusions. Integrating machine learning algorithms into the ultrasound protocol increases diagnostic objectivity; however, hierarchical approaches require calibration to reduce false-positive results in the borderline zone of early fibrosis stages.
- 20 January 2026
- Published by State Institution of Science "Center of Innovative Healthcare Technologies" State Administrative Department in Clinical and Preventive Medicine
- No. 1,p. 104-112
Open AccessDIAGNOSTIC PERFORMANCE OF MULTIPARAMETRIC ULTRASOUND WITH ARTIFICIAL INTELLIGENCE TOOLS IN PEDIATRIC DIFFUSE LIVER DISEASES
Introduction. Chronic diffuse liver diseases in children require safe noninvasive assessment and repeated monitoring of fibrosis. Liver biopsy remains the reference method for morphological verification; however, in pediatric practice its use is limited by invasiveness, risk of complications, and ethical considerations. Multiparametric ultrasound, including shear wave elastography and Doppler assessment of blood flow, is suitable for longitudinal follow-up, yet interpretation of ultrasound features remains operator-dependent. Automated image analysis using artificial intelligence algorithms may increase objectivity and reproducibility of fibrosis stratification and support clinical decision-making. Aim. To evaluate the diagnostic accuracy of multiparametric ultrasound, including two-dimensional shear wave elastography and machine learning algorithms, for staging liver fibrosis in children. Materials and methods: This retrospective single-center study included 185 children (140 with chronic diffuse liver diseases and 45 controls). All participants underwent two-dimensional shear wave elastography, hepatic vein Doppler assessment, and grayscale ultrasound image analysis using two machine learning approaches: Model A (random forest algorithm; binary classification) and Model B (hierarchical ensemble; multiclass classification). The reference approach was clinical verification and risk stratification based on elastography findings. Results. A strong positive association was observed between elastography measures and Model B predictions (Spearman rank correlation coefficient 0.77). Model A demonstrated 84% specificity in differentiating normal findings from pathological changes. Model B showed higher sensitivity (83%) but tended to overestimate fibrosis at the boundary between minimal and moderate stages. Doppler patterns, particularly a monophasic hepatic vein waveform, were significantly associated with higher fibrosis stages. Conclusions. Integrating machine learning algorithms into the ultrasound protocol increases diagnostic objectivity; however, hierarchical approaches require calibration to reduce false-positive results in the borderline zone of early fibrosis stages.
- Open Access
- 24 November 2025
- Published by Igor Sikorsky Kyiv Polytechnic Institute in Innovative Biosystems and Bioengineering
- Vol. 9 (4) , 16-27
Diagnosis of COVID-19-associated Cardiopulmonary Pathology From CT Data Using Artificial Intelligence: a Review of Methods and Future Research Directions
Background. The COVID-19 pandemic, caused by SARS-CoV-2, has significantly impacted global health, emphasizing the importance of efficient diagnostic methods. Computed tomography (CT) imaging plays a crucial role in identifying COVID-19-associated lung pathologies, yet manual analysis of extensive imaging data remains burdensome. Machine learning (ML) methods offer promising automated solutions to expedite diagnostics and reduce workload on radiologists. Objective. To evaluate the effectiveness of machine learning algorithms, specifically convolutional neural networks (CNN) and texture analysis methods, in automated detection and classification of COVID-19-related cardiopulmonary pathology using chest CT imaging. Methods. This study analyzed chest CT datasets obtained from clinical resources, which included images from patients with COVID-19 exhibiting ground-glass opacities, crazy-paving patterns, and consolidations. Regions of interest (ROIs) were segmented and classified using various machine learning approaches: CNN combined with gray-level co-occurrence matrix (GLCM) texture analysis, logistic self-organizing forest (LSOF), group method of data handling (GMDH), and ensemble methods including random forest, XGBoost, LightGBM, and random forest of optimal complexity trees (RFOCT). Results. The CNN and texture-based hybrid classifiers achieved high accuracy, with overall classification accuracies ranging from 83% to 99%. Specifically, ground-glass opacity identification reached up to 100% accuracy, while crazy-paving patterns and consolidations showed slightly lower accuracies (71–95%). The ensemble method RFOCT achieved the highest accuracy (89%) in differentiating acute COVID-19 from Long COVID. Additionally, methods incorporating texture analysis significantly enhanced the accuracy and informativeness of CT-based diagnostics. Conclusions. Machine learning algorithms, particularly CNNs and advanced texture analysis, demonstrate significant potential in automating the diagnosis of COVID-19-associated lung pathology. These approaches not only increase diagnostic efficiency but also facilitate the detection of subtle pathological changes, crucial for clinical decision-making and patient management during epidemiological crises. Future research should address current limitations related to dataset size, computational complexity, and generalizability to further enhance clinical applicability.
- 24 November 2025
- Published by Igor Sikorsky Kyiv Polytechnic Institute in Innovative Biosystems and Bioengineering
- Vol. 9 (4) , 16-27
Open AccessDiagnosis of COVID-19-associated Cardiopulmonary Pathology From CT Data Using Artificial Intelligence: a Review of Methods and Future Research Directions
Background. The COVID-19 pandemic, caused by SARS-CoV-2, has significantly impacted global health, emphasizing the importance of efficient diagnostic methods. Computed tomography (CT) imaging plays a crucial role in identifying COVID-19-associated lung pathologies, yet manual analysis of extensive imaging data remains burdensome. Machine learning (ML) methods offer promising automated solutions to expedite diagnostics and reduce workload on radiologists. Objective. To evaluate the effectiveness of machine learning algorithms, specifically convolutional neural networks (CNN) and texture analysis methods, in automated detection and classification of COVID-19-related cardiopulmonary pathology using chest CT imaging. Methods. This study analyzed chest CT datasets obtained from clinical resources, which included images from patients with COVID-19 exhibiting ground-glass opacities, crazy-paving patterns, and consolidations. Regions of interest (ROIs) were segmented and classified using various machine learning approaches: CNN combined with gray-level co-occurrence matrix (GLCM) texture analysis, logistic self-organizing forest (LSOF), group method of data handling (GMDH), and ensemble methods including random forest, XGBoost, LightGBM, and random forest of optimal complexity trees (RFOCT). Results. The CNN and texture-based hybrid classifiers achieved high accuracy, with overall classification accuracies ranging from 83% to 99%. Specifically, ground-glass opacity identification reached up to 100% accuracy, while crazy-paving patterns and consolidations showed slightly lower accuracies (71–95%). The ensemble method RFOCT achieved the highest accuracy (89%) in differentiating acute COVID-19 from Long COVID. Additionally, methods incorporating texture analysis significantly enhanced the accuracy and informativeness of CT-based diagnostics. Conclusions. Machine learning algorithms, particularly CNNs and advanced texture analysis, demonstrate significant potential in automating the diagnosis of COVID-19-associated lung pathology. These approaches not only increase diagnostic efficiency but also facilitate the detection of subtle pathological changes, crucial for clinical decision-making and patient management during epidemiological crises. Future research should address current limitations related to dataset size, computational complexity, and generalizability to further enhance clinical applicability.
- Open Access
- 27 September 2024
- Published by National Institute of Cardiovascular Surgery named after M.M. Amosov of the NAMS of Ukraine in Ukrainian journal of cardiovascular surgery
- Vol. 32 (3) , 58-65
Development of Artificial Intelligence-Based Programs for the Diagnosis of Myocarditis in COVID-19 Using Chest Computed Tomography Data»
It has been established that 7.2% of patients hospitalized with coronavirus disease (COVID-19) exhibit signs of heart disease, with 23% of these patients experiencing heart failure. Currently, there is a lack of data on chest computed tomography (CT) for diagnosing myocarditis associated with COVID-19. The aim. To justify the feasibility and develop classification models for diagnosing myocarditis in COVID-19 patients based on chest CT data processing. Materials and methods. A retrospective analysis of data from 140 COVID-19 patients was conducted. Chest CT scans were analyzed using DRAGONFLY software, with permission from Object Research Systems. The COVID-CT-MD database, which includes CT data from 169 confirmed cases of SARS-CoV-2 infection, was used to build classification models. The regions of interest were fragments of heart CT images. Texture analysis methods were employed to create diagnostic models. Results. It was shown that the average density of the myocardium of a patient with a confirmed diagnosis of SARS-CoV-2 infection according to the Hounsfield scale does not essentially differ from the densitometric indicators of a healthy person. Therefore, the research was focused on finding structural changes in CT images for their use in constructing diagnostic models. The use of different classification algorithms had little effect on classification accuracy, probably due to the informational content of the input data. However, the obtained accuracy of the diagnostic models is acceptable and allows them to be used to support medical decision-making regarding diagnosis and treatment. Conclusions. Using classic methods, myocarditis was diagnosed in 7.1% of patients with severe pneumonia caused by the coronavirus. The global data closely aligns with the results of our clinical studies. The obtained results allowed for assessing structural changes in the myocardium characteristic of the acute form of SARS-CoV-2 infection. The constructed classification models indicate that specific changes in the myocardium during the acute form of SARS-CoV-2 infection can be identified using CT. The highest diagnostic accuracy on test samples reached 74%. The implementation of the developed diagnostic programs based on texture analysis of CT data and artificial intelligence technologies enables the diagnosis of myocarditis and the assessment of long-term treatment efficiency. Creation of these diagnostic programs using artificial intelligence technologies significantly simplifies the work of radiologists and improves the efficiency of myocarditis diagnosis in patients with SARS-CoV-2 infection.
- 27 September 2024
- Published by National Institute of Cardiovascular Surgery named after M.M. Amosov of the NAMS of Ukraine in Ukrainian journal of cardiovascular surgery
- Vol. 32 (3) , 58-65
Open AccessDevelopment of Artificial Intelligence-Based Programs for the Diagnosis of Myocarditis in COVID-19 Using Chest Computed Tomography Data»
It has been established that 7.2% of patients hospitalized with coronavirus disease (COVID-19) exhibit signs of heart disease, with 23% of these patients experiencing heart failure. Currently, there is a lack of data on chest computed tomography (CT) for diagnosing myocarditis associated with COVID-19. The aim. To justify the feasibility and develop classification models for diagnosing myocarditis in COVID-19 patients based on chest CT data processing. Materials and methods. A retrospective analysis of data from 140 COVID-19 patients was conducted. Chest CT scans were analyzed using DRAGONFLY software, with permission from Object Research Systems. The COVID-CT-MD database, which includes CT data from 169 confirmed cases of SARS-CoV-2 infection, was used to build classification models. The regions of interest were fragments of heart CT images. Texture analysis methods were employed to create diagnostic models. Results. It was shown that the average density of the myocardium of a patient with a confirmed diagnosis of SARS-CoV-2 infection according to the Hounsfield scale does not essentially differ from the densitometric indicators of a healthy person. Therefore, the research was focused on finding structural changes in CT images for their use in constructing diagnostic models. The use of different classification algorithms had little effect on classification accuracy, probably due to the informational content of the input data. However, the obtained accuracy of the diagnostic models is acceptable and allows them to be used to support medical decision-making regarding diagnosis and treatment. Conclusions. Using classic methods, myocarditis was diagnosed in 7.1% of patients with severe pneumonia caused by the coronavirus. The global data closely aligns with the results of our clinical studies. The obtained results allowed for assessing structural changes in the myocardium characteristic of the acute form of SARS-CoV-2 infection. The constructed classification models indicate that specific changes in the myocardium during the acute form of SARS-CoV-2 infection can be identified using CT. The highest diagnostic accuracy on test samples reached 74%. The implementation of the developed diagnostic programs based on texture analysis of CT data and artificial intelligence technologies enables the diagnosis of myocarditis and the assessment of long-term treatment efficiency. Creation of these diagnostic programs using artificial intelligence technologies significantly simplifies the work of radiologists and improves the efficiency of myocarditis diagnosis in patients with SARS-CoV-2 infection.
- Open Access
- 29 March 2024
- Published by Igor Sikorsky Kyiv Polytechnic Institute in System research and information technologies
- No. 1,p. 33-41
Cardiomyopathy prediction in patients with permanent ventricular pacing using machine learning methods
Pacing-induced cardiomyopathy is a notable issue in patients needing permanent ventricular pacing. Identifying risk groups early and swiftly preventing the ailment can reduce patient harm. However, current prognostic methods require clarity. We employed machine learning to develop predictive models using medical data. Three algorithms — decision tree, group method of data handling, and logistic regression — formed models that forecast pacing-induced cardiomyopathy. These models displayed high accuracy in predicting development, signifying soundness. Factors like age, paced QRS width, pacing mode, and ventricular index during implantation significantly influenced predictions. Machine learning can enhance pacing-induced cardiomyopathy prediction in ventricular pacing patients, aiding medical practice and preventive strategies. Кардіоміопатія, спричинена кардіостимуляцією, є важливою проблемою для пацієнтів, які потребують постійної шлуночкової кардіостимуляції. Раннє виявлення груп ризику та швидка профілактика недуги можуть зменшити шкоду для пацієнтів. Однак сучасні методи прогнозування потребують доопрацювання. Застосовано машинне навчання для розроблення прогностичних моделей на основі медичних даних. Три алгоритми — дерево рішень, група оброблення даних та логістична регресія — сформували моделі, які прогнозують кардіоміопатію, спричинену кардіостимуляцією. Ці моделі показали високу точність у прогнозуванні розвитку, що свідчить про їх надійність. Ключові фактори, такі як вік, ширина QRS, режим кардіостимуляції та шлуночковий індекс під час імплантації, суттєво впливали на прогнози. Машинне навчання може покращити прогнозування кардіоміопатії, спричиненої кардіостимуляцією, у пацієнтів, які перебувають на шлуночковій електрокардіостимуляції, допомагаючи медичній практиці та профілактичним стратегіям.
- 29 March 2024
- Published by Igor Sikorsky Kyiv Polytechnic Institute in System research and information technologies
- No. 1,p. 33-41
Open AccessCardiomyopathy prediction in patients with permanent ventricular pacing using machine learning methods
Pacing-induced cardiomyopathy is a notable issue in patients needing permanent ventricular pacing. Identifying risk groups early and swiftly preventing the ailment can reduce patient harm. However, current prognostic methods require clarity. We employed machine learning to develop predictive models using medical data. Three algorithms — decision tree, group method of data handling, and logistic regression — formed models that forecast pacing-induced cardiomyopathy. These models displayed high accuracy in predicting development, signifying soundness. Factors like age, paced QRS width, pacing mode, and ventricular index during implantation significantly influenced predictions. Machine learning can enhance pacing-induced cardiomyopathy prediction in ventricular pacing patients, aiding medical practice and preventive strategies. Кардіоміопатія, спричинена кардіостимуляцією, є важливою проблемою для пацієнтів, які потребують постійної шлуночкової кардіостимуляції. Раннє виявлення груп ризику та швидка профілактика недуги можуть зменшити шкоду для пацієнтів. Однак сучасні методи прогнозування потребують доопрацювання. Застосовано машинне навчання для розроблення прогностичних моделей на основі медичних даних. Три алгоритми — дерево рішень, група оброблення даних та логістична регресія — сформували моделі, які прогнозують кардіоміопатію, спричинену кардіостимуляцією. Ці моделі показали високу точність у прогнозуванні розвитку, що свідчить про їх надійність. Ключові фактори, такі як вік, ширина QRS, режим кардіостимуляції та шлуночковий індекс під час імплантації, суттєво впливали на прогнози. Машинне навчання може покращити прогнозування кардіоміопатії, спричиненої кардіостимуляцією, у пацієнтів, які перебувають на шлуночковій електрокардіостимуляції, допомагаючи медичній практиці та профілактичним стратегіям.
- 1 March 2023
- Published by Springer Nature in Cybernetics and Systems Analysis
- Vol. 59 (2) , 346-358
Classification of Pathologies on Medical Images Using the Algorithm of Random Forest of Optimal-Complexity Trees
- 1 March 2023
- Published by Springer Nature in Cybernetics and Systems Analysis
- Vol. 59 (2) , 346-358
Classification of Pathologies on Medical Images Using the Algorithm of Random Forest of Optimal-Complexity Trees
- 29 December 2022
- Published by Igor Sikorsky Kyiv Polytechnic Institute in Біомедична інженерія і технологія
- No. 8,p. 40-47
КОМБІНАЦІЯ ЛОКАЛЬНОЇ ПОРОГОВОЇ БІНАРИЗАЦІЇ ТА МАШИННОГО НАВЧАННЯ ДЛЯ КЛАСИФІКАЦІЇ ПУХЛИН МОЛОЧНОЇ ЗАЛОЗИ
Проблематика. Рання діагностика раку молочної залози має колосальне значення, оскільки дана патологія є одним із найбільш розповсюджених чинників летальності серед жінок по всьому світу. Чи не небезпечнішим підтипом раку молочної залози вважається інвазивна протокова карцинома. Зазвичай патологоанатоми фокусуються на областях з подібною карциномою, так як це дозволяє присвоїти оцінку агресивності усьому зразку монтування. Саме тому важливою задачею є автоматизоване виявлення карциноми при діагностиці ракових пухлин молочної залози. Мета. Встановлення основних етапів побудови діагностичних алгоритмів класифікації типу ракової пухлини молочної залози на основі аналізу гістологічних знімків. Методика реалізації. Запропоновано алгоритм на основі методу локальної порогової бінаризації для вилучення інформативних ознак з медичних зображень, та машинного навчання для побудови моделей розпізнавання типу ракової пухлини молочної залози за допомогою методів класифікації, таких як: метод групового урахування аргументів, логістична регресія, наївний байєсів класифікатор, метод k найближчих сусідів, та метод випадкового лісу. Перед виконанням алгоритму вибірку зображень було розбито на робочу (80%) для навчання моделей, та екзаменаційну (20%), яка не приймає жодної участі в експериментах аж до отримання результуючої моделі. Після застосування локальної порогової бінаризації отримуються дуети (комбінації із двох пікселей) і тріо (комбінації із трьох пікселей). При моделюванні робоча вибірка була розбита на тренувальну (75%) і тестову (25%). Результати дослідження. В результаті моделювання було отримано 10 моделей машинного навчання. Найкращою з них виявилась модель KNN, навчена на дуетах бінаризованих пікселей, що дала на тестовій вибірці 78.5% точності класифікації. Ключові слова – молочна залоза, ракові пухлини, гістологічні знімки, візуальна діагностика, бінаризація знімку, штучний інтелект, машинне навчання.
- 29 December 2022
- Published by Igor Sikorsky Kyiv Polytechnic Institute in Біомедична інженерія і технологія
- No. 8,p. 40-47
КОМБІНАЦІЯ ЛОКАЛЬНОЇ ПОРОГОВОЇ БІНАРИЗАЦІЇ ТА МАШИННОГО НАВЧАННЯ ДЛЯ КЛАСИФІКАЦІЇ ПУХЛИН МОЛОЧНОЇ ЗАЛОЗИ
Проблематика. Рання діагностика раку молочної залози має колосальне значення, оскільки дана патологія є одним із найбільш розповсюджених чинників летальності серед жінок по всьому світу. Чи не небезпечнішим підтипом раку молочної залози вважається інвазивна протокова карцинома. Зазвичай патологоанатоми фокусуються на областях з подібною карциномою, так як це дозволяє присвоїти оцінку агресивності усьому зразку монтування. Саме тому важливою задачею є автоматизоване виявлення карциноми при діагностиці ракових пухлин молочної залози. Мета. Встановлення основних етапів побудови діагностичних алгоритмів класифікації типу ракової пухлини молочної залози на основі аналізу гістологічних знімків. Методика реалізації. Запропоновано алгоритм на основі методу локальної порогової бінаризації для вилучення інформативних ознак з медичних зображень, та машинного навчання для побудови моделей розпізнавання типу ракової пухлини молочної залози за допомогою методів класифікації, таких як: метод групового урахування аргументів, логістична регресія, наївний байєсів класифікатор, метод k найближчих сусідів, та метод випадкового лісу. Перед виконанням алгоритму вибірку зображень було розбито на робочу (80%) для навчання моделей, та екзаменаційну (20%), яка не приймає жодної участі в експериментах аж до отримання результуючої моделі. Після застосування локальної порогової бінаризації отримуються дуети (комбінації із двох пікселей) і тріо (комбінації із трьох пікселей). При моделюванні робоча вибірка була розбита на тренувальну (75%) і тестову (25%). Результати дослідження. В результаті моделювання було отримано 10 моделей машинного навчання. Найкращою з них виявилась модель KNN, навчена на дуетах бінаризованих пікселей, що дала на тестовій вибірці 78.5% точності класифікації. Ключові слова – молочна залоза, ракові пухлини, гістологічні знімки, візуальна діагностика, бінаризація знімку, штучний інтелект, машинне навчання.
- 24 September 2021
- Published by Springer Nature
- p. 503-512
Forming the System with the Functionality of Clinical Pharmacist for Personalized Treatment Strategy Searching
- 24 September 2021
- Published by Springer Nature
- p. 503-512
Forming the System with the Functionality of Clinical Pharmacist for Personalized Treatment Strategy Searching
- Open Access
- 17 August 2021
- Published by Igor Sikorsky Kyiv Polytechnic Institute in Innovative Biosystems and Bioengineering
- Vol. 5 (3) , 153-166
Comparative Analysis of Classification Algorithms in the Analysis of Medical Images From Speckle Tracking Echocardiography Video Data
Background. Machine learning allows applying various intelligent algorithms to produce diagnostic and/or prognostic models. Such models can be used to determine the functional state of the heart, which is diagnosed by speckle-tracking echocardiography. To determine the patient's heart condition in detail, a classification approach is used in machine learning. Each of the classification algorithms has a different performance when applied to certain situations. Therefore, the actual task is to determine the most efficient algorithm for solving a specific task of classifying the patient's heart condition when applying the same speckle-tracking echocardiography data set. Objective. We are aimed to evaluate the effectiveness of the application of prognostic models of logistic regression, the group method of data handling (GMDH), random forest, and adaptive boosting (AdaBoost) in the construction of algorithms to support medical decision-making on the diagnosis of coronary heart disease. Methods. Video data from speckle-tracking echocardiography of 40 patients with coronary heart disease and 16 patients without cardiac pathology were used for the study. Echocardiography was recorded in B-mode in three positions: long axis, 4-chamber, and 2-chamber. Echocardiography frames that reflect the systole and diastole of the heart (308 samples in total) were taken as objects for classification. To obtain informative features of the selected objects, the genetic GMDH approach was applied to identify the best structure of harmonic textural features. We compared the efficiency of the following classification algorithms: logistic regression method, GMDH classifier, random forest method, and AdaBoost method. Results. Four classification models were constructed for each of the three B-mode echocardiography positions. For this purpose, the data samples were divided into 3: training sample (60%), validation sample (20%), and test sample (20%). Objective evaluation of the models on the test sample showed that the best classification method was random forest (90.3% accuracy on the 4-chamber echocardiography position, 74.2% on the 2-chamber, and 77.4% on the long axis). This was also confirmed by ROC analysis, wherein in all cases, the random forest was the most effective in classifying cardiac conditions. Conclusions. The best classification algorithm for cardiac diagnostics by speckle-tracking echocardiography was determined. It turned out to be a random forest, which can be explained by the ensemble approach of begging, which is inherent in this classification method. It will be the mainstay of further research, which is planned to be performed to develop a full-fledged decision support system for cardiac diagnostics.
- 17 August 2021
- Published by Igor Sikorsky Kyiv Polytechnic Institute in Innovative Biosystems and Bioengineering
- Vol. 5 (3) , 153-166
Open AccessComparative Analysis of Classification Algorithms in the Analysis of Medical Images From Speckle Tracking Echocardiography Video Data
Background. Machine learning allows applying various intelligent algorithms to produce diagnostic and/or prognostic models. Such models can be used to determine the functional state of the heart, which is diagnosed by speckle-tracking echocardiography. To determine the patient's heart condition in detail, a classification approach is used in machine learning. Each of the classification algorithms has a different performance when applied to certain situations. Therefore, the actual task is to determine the most efficient algorithm for solving a specific task of classifying the patient's heart condition when applying the same speckle-tracking echocardiography data set. Objective. We are aimed to evaluate the effectiveness of the application of prognostic models of logistic regression, the group method of data handling (GMDH), random forest, and adaptive boosting (AdaBoost) in the construction of algorithms to support medical decision-making on the diagnosis of coronary heart disease. Methods. Video data from speckle-tracking echocardiography of 40 patients with coronary heart disease and 16 patients without cardiac pathology were used for the study. Echocardiography was recorded in B-mode in three positions: long axis, 4-chamber, and 2-chamber. Echocardiography frames that reflect the systole and diastole of the heart (308 samples in total) were taken as objects for classification. To obtain informative features of the selected objects, the genetic GMDH approach was applied to identify the best structure of harmonic textural features. We compared the efficiency of the following classification algorithms: logistic regression method, GMDH classifier, random forest method, and AdaBoost method. Results. Four classification models were constructed for each of the three B-mode echocardiography positions. For this purpose, the data samples were divided into 3: training sample (60%), validation sample (20%), and test sample (20%). Objective evaluation of the models on the test sample showed that the best classification method was random forest (90.3% accuracy on the 4-chamber echocardiography position, 74.2% on the 2-chamber, and 77.4% on the long axis). This was also confirmed by ROC analysis, wherein in all cases, the random forest was the most effective in classifying cardiac conditions. Conclusions. The best classification algorithm for cardiac diagnostics by speckle-tracking echocardiography was determined. It turned out to be a random forest, which can be explained by the ensemble approach of begging, which is inherent in this classification method. It will be the mainstay of further research, which is planned to be performed to develop a full-fledged decision support system for cardiac diagnostics.