Vinera Baiturganova
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Vinera Baiturganova has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
- Zhanibek Issabekov
Satbayev University
4 shared publications - Bigaliyeva Zhanar
Department of Robotics and Technical Means of Automation, Satbayev University, Almaty 050013, Kazakhstan
3 shared publications - Perizat Rakhmetova
Satbayev University
3 shared publications - Ainur Bekbay
Satbayev University
2 shared publications - A. Baimakhanova1 shared publication
- Akezhan Sabibolda
Department of Electronics, Telecommunications and Space Technologies, Satbayev University, Almaty 050013, Kazakhstan
1 shared publication
Latest Publications
- Open Access
- 30 September 2026
- Published by Central Asian Academic Research Center LLP in NEWS OF THE NATIONAL ACADEMY OF SCIENCES OF THE REPUBLIC OF KAZAKHSTAN
- No. 3,p. 123–138
COMPARISON OF CNN ARCHITECTURES FOR AUTOMATED ACL INJURY CLASSIFICATION FROM KNEE MRI
Relevance. Anterior cruciate ligament (ACL) tears are among the most clinically significant knee injuries, and delayed or inaccurate diagnosis is associated with progressive joint degeneration. Magnetic resonance imaging (MRI) is the primary non-invasive modality for assessing ligament integrity, but its interpretation is time-consuming, requires considerable expertise and remains vulnerable to fatigue and cognitive bias, which motivates the development of automated decision-support tools. Aim. The study systematically evaluates and compares four convolutional neural network (CNN) architectures trained from scratch for three-class ACL injury classification (healthy, partial tear, complete tear). Methods. A publicly available dataset of 736 sagittal-plane knee MRI volumes (a subset of the KneeMRI dataset) was used. Each 3D volume was reduced to a 2D image by extracting the middle sagittal slice and resizing it to 128 × 128 pixels. ResNet-18, MobileNetV2, EfficientNet-B0 and MobileNetV4_small were trained end-to-end under a unified configuration with the AdamW optimizer, cosine annealing, cross-entropy loss and weighted random sampling to compensate for class imbalance. Each model was trained with five independent random seeds, and performance is reported as mean ± standard deviation of weighted precision, recall, F1-score and AUC. Results. ResNet-18 achieved the best performance (F1-score 0.6637 ± 0.0351; AUC 0.6622 ± 0.0136), followed by EfficientNet-B0, MobileNetV4_small, and MobileNetV2. Practical significance. The study provides a reproducible benchmark and practical guidance for architecture selection in small-scale medical imaging tasks, showing that, when training from scratch, the input representation and the availability of pre-trained features matter more than the choice of backbone, while clinical reliability will require 3D volumetric context, pre-training and external validation.
- 30 September 2026
- Published by Central Asian Academic Research Center LLP in NEWS OF THE NATIONAL ACADEMY OF SCIENCES OF THE REPUBLIC OF KAZAKHSTAN
- No. 3,p. 123–138
Open AccessCOMPARISON OF CNN ARCHITECTURES FOR AUTOMATED ACL INJURY CLASSIFICATION FROM KNEE MRI
Relevance. Anterior cruciate ligament (ACL) tears are among the most clinically significant knee injuries, and delayed or inaccurate diagnosis is associated with progressive joint degeneration. Magnetic resonance imaging (MRI) is the primary non-invasive modality for assessing ligament integrity, but its interpretation is time-consuming, requires considerable expertise and remains vulnerable to fatigue and cognitive bias, which motivates the development of automated decision-support tools. Aim. The study systematically evaluates and compares four convolutional neural network (CNN) architectures trained from scratch for three-class ACL injury classification (healthy, partial tear, complete tear). Methods. A publicly available dataset of 736 sagittal-plane knee MRI volumes (a subset of the KneeMRI dataset) was used. Each 3D volume was reduced to a 2D image by extracting the middle sagittal slice and resizing it to 128 × 128 pixels. ResNet-18, MobileNetV2, EfficientNet-B0 and MobileNetV4_small were trained end-to-end under a unified configuration with the AdamW optimizer, cosine annealing, cross-entropy loss and weighted random sampling to compensate for class imbalance. Each model was trained with five independent random seeds, and performance is reported as mean ± standard deviation of weighted precision, recall, F1-score and AUC. Results. ResNet-18 achieved the best performance (F1-score 0.6637 ± 0.0351; AUC 0.6622 ± 0.0136), followed by EfficientNet-B0, MobileNetV4_small, and MobileNetV2. Practical significance. The study provides a reproducible benchmark and practical guidance for architecture selection in small-scale medical imaging tasks, showing that, when training from scratch, the input representation and the availability of pre-trained features matter more than the choice of backbone, while clinical reliability will require 3D volumetric context, pre-training and external validation.
- 25 January 2026
- Published by Shakarim University in Bulletin of Shakarim University. Technical Sciences
- Vol. 1 (4(20)) , 65-72
USING ARTIFICIAL INTELLIGENCE TO ADAPT STUDENTS' LEARNING TRAJECTORIES
I In the modern era, where rapidly changing educational landscapes require adaptive learning mechanisms, integrating artificial intelligence into education is no longer a futuristic dream but a necessity. This paper presents a sophisticated intelligent system for real-time monitoring, detailed analyzing, and adaptive optimizing of competency acquisition throughout the learning process. Based on a neural network architecture augmented with ontological modeling and set-theoretic principles, this system provides a structured yet flexible framework for continuous learning improvement. Using the Six Sigma DMAIC (Define-Measure-Analyze-Improve-Control) methodology, the proposed model systematically improves educational trajectories through data-driven analysis and iterative improvements, ensuring precise alignment with industry and institutional requirements. In addition, the system incorporates predictive analytics and personalized feedback mechanisms that adapt instructional strategies to individual learner needs, thus bridging the gap between standardized curricula and personal learning paths. It further enhances decision-making for educators by providing actionable insights, real-time performance dashboards, and evidence-based recommendations. By combining advanced computational intelligence with proven educational methodologies, this research contributes to the creation of resilient, scalable, and future-ready learning environments that foster innovation, efficiency, and lifelong skill development.
- 25 January 2026
- Published by Shakarim University in Bulletin of Shakarim University. Technical Sciences
- Vol. 1 (4(20)) , 65-72
USING ARTIFICIAL INTELLIGENCE TO ADAPT STUDENTS' LEARNING TRAJECTORIES
I In the modern era, where rapidly changing educational landscapes require adaptive learning mechanisms, integrating artificial intelligence into education is no longer a futuristic dream but a necessity. This paper presents a sophisticated intelligent system for real-time monitoring, detailed analyzing, and adaptive optimizing of competency acquisition throughout the learning process. Based on a neural network architecture augmented with ontological modeling and set-theoretic principles, this system provides a structured yet flexible framework for continuous learning improvement. Using the Six Sigma DMAIC (Define-Measure-Analyze-Improve-Control) methodology, the proposed model systematically improves educational trajectories through data-driven analysis and iterative improvements, ensuring precise alignment with industry and institutional requirements. In addition, the system incorporates predictive analytics and personalized feedback mechanisms that adapt instructional strategies to individual learner needs, thus bridging the gap between standardized curricula and personal learning paths. It further enhances decision-making for educators by providing actionable insights, real-time performance dashboards, and evidence-based recommendations. By combining advanced computational intelligence with proven educational methodologies, this research contributes to the creation of resilient, scalable, and future-ready learning environments that foster innovation, efficiency, and lifelong skill development.
- 25 January 2026
- Published by Shakarim University in Bulletin of Shakarim University. Technical Sciences
- Vol. 1 (4(20)) , 6-13
AUTOMATED CLASSIFICATION OF HEMODYNAMICALLY SIGNIFICANT ARRHYTHMIAS BASED ON ECG FEATURES
This paper presents an automated method for the classification of hemodynamically significant arrhythmias (HSA) based solely on electrocardiographic (ECG) features, without the use of additional imaging diagnostic techniques. The proposed approach relies on key ECG parameters, including QRS complex duration, RR intervals, and heart rate (HR). The study is based on data from the open-access MIT-BIH Arrhythmia Database, which contains multiple types of cardiac rhythm disturbances. A comparative analysis of arrhythmia classes was conducted, leading to the identification of diagnostically significant predictors associated with hemodynamic instability. Logical decision rules and a decision tree model were developed to enable automatic recognition of HSA and clinical risk stratification. The proposed algorithm demonstrates high interpretability and practical applicability for real-time monitoring systems. The results confirm that ECG-based features alone can be effectively used for preliminary detection of dangerous arrhythmias. The developed approach is especially valuable for telemedicine systems and healthcare facilities with limited access to expensive diagnostic equipment.
- 25 January 2026
- Published by Shakarim University in Bulletin of Shakarim University. Technical Sciences
- Vol. 1 (4(20)) , 6-13
AUTOMATED CLASSIFICATION OF HEMODYNAMICALLY SIGNIFICANT ARRHYTHMIAS BASED ON ECG FEATURES
This paper presents an automated method for the classification of hemodynamically significant arrhythmias (HSA) based solely on electrocardiographic (ECG) features, without the use of additional imaging diagnostic techniques. The proposed approach relies on key ECG parameters, including QRS complex duration, RR intervals, and heart rate (HR). The study is based on data from the open-access MIT-BIH Arrhythmia Database, which contains multiple types of cardiac rhythm disturbances. A comparative analysis of arrhythmia classes was conducted, leading to the identification of diagnostically significant predictors associated with hemodynamic instability. Logical decision rules and a decision tree model were developed to enable automatic recognition of HSA and clinical risk stratification. The proposed algorithm demonstrates high interpretability and practical applicability for real-time monitoring systems. The results confirm that ECG-based features alone can be effectively used for preliminary detection of dangerous arrhythmias. The developed approach is especially valuable for telemedicine systems and healthcare facilities with limited access to expensive diagnostic equipment.
- Open Access
- 28 October 2025
- Published by Private Company Technology Center in Eastern-European Journal of Enterprise Technologies
- Vol. 5 (9 (137)) , 117-124
An interpretable ECG-based approach for detecting hemodynamically significant arrhythmias using lightweight machine learning models
The object of this study is the diagnostic process of patients with suspected hemodynamically significant arrhythmia in emergency and telemedicine settings, where rapid and interpretable decision support is required. The problem addressed is the limited access to echocardiographic assessment in emergency and resource-constrained environments, where interpretable and computationally efficient alternatives are urgently needed, particularly for mobile and field-deployed applications. The main results show that machine learning models, such as XGBoost, achieved strong diagnostic performance (F1-score = 0.84, AUC = 0.91), while rule-based classifiers provided clinically interpretable accuracy. These results enabled partial compensation for the absence of echocardiography and contributed to reliable triage in acute and time-sensitive settings. This effectiveness stems from key features of the method: reliance on interpretable ECG features (tQRS, tRR, HR, and EF derived from tQRS/tRR) and low computational complexity, setting it apart from more opaque deep learning methods. The results are explained by the strong correlation between these features and both electrical and mechanical heart function, enabling hemodynamic assessment without imaging. This supports clinical trust in the algorithm’s outputs. The proposed approach is applicable in primary screening, emergency triage, telemedicine, and remote monitoring, combining accuracy with explainability and autonomy from imaging tools. Therefore, research on interpretable ECG-based detection of hemodynamically significant arrhythmias remains highly relevant, especially in settings lacking access to specialized diagnostics
- 28 October 2025
- Published by Private Company Technology Center in Eastern-European Journal of Enterprise Technologies
- Vol. 5 (9 (137)) , 117-124
Open AccessAn interpretable ECG-based approach for detecting hemodynamically significant arrhythmias using lightweight machine learning models
The object of this study is the diagnostic process of patients with suspected hemodynamically significant arrhythmia in emergency and telemedicine settings, where rapid and interpretable decision support is required. The problem addressed is the limited access to echocardiographic assessment in emergency and resource-constrained environments, where interpretable and computationally efficient alternatives are urgently needed, particularly for mobile and field-deployed applications. The main results show that machine learning models, such as XGBoost, achieved strong diagnostic performance (F1-score = 0.84, AUC = 0.91), while rule-based classifiers provided clinically interpretable accuracy. These results enabled partial compensation for the absence of echocardiography and contributed to reliable triage in acute and time-sensitive settings. This effectiveness stems from key features of the method: reliance on interpretable ECG features (tQRS, tRR, HR, and EF derived from tQRS/tRR) and low computational complexity, setting it apart from more opaque deep learning methods. The results are explained by the strong correlation between these features and both electrical and mechanical heart function, enabling hemodynamic assessment without imaging. This supports clinical trust in the algorithm’s outputs. The proposed approach is applicable in primary screening, emergency triage, telemedicine, and remote monitoring, combining accuracy with explainability and autonomy from imaging tools. Therefore, research on interpretable ECG-based detection of hemodynamically significant arrhythmias remains highly relevant, especially in settings lacking access to specialized diagnostics
- Open Access
- 8 July 2025
- Published by Shakarim University in Bulletin of Shakarim University. Technical Sciences
- No. 2(18),p. 121-128
DEVELOPMENT OF A PREDICTIVE MAINTENANCE SYSTEM BASED ON MACHINE LEARNING
In traditional industrial settings, maintenance methods based on reactive repairs or scheduled time intervals often lead to downtime and reduced efficiency. Therefore, predictive maintenance uses data to predict failures before they actually occur. The objective of this study is to develop an intelligent predictive maintenance system based on machine learning and implement it in the industry of Kazakhstan, focusing on the principles of Industry 4.0. The system is based on collecting data from sensors (current, temperature, pressure, vibration) using Siemens PLCs, integrated via the OPC UA protocol into the industrial IoT infrastructure. Additionally, computer vision is used to monitor the equipment condition in real time. The obtained data is processed by machine learning algorithms, including neural networks, linear regression, and autoencoders. To adapt the model to changes, it is trained continuously using Bayesian updating. Visualization and interaction with users are implemented via SCADA for engineers and Power BI for managers. In addition, the paper discusses the challenges associated with the deployment of predictive maintenance solutions and suggests future directions for improving scalability, security, and real-time data processing capabilities. The obtained results contribute to the growing body of research in the field of predictive maintenance, demonstrating its potential to improve efficiency, reduce operating costs, and support the transition to datadriven, intelligent manufacturing systems. The work demonstrates the potential of predictive maintenance as a solution for aging industries with a shortage of engineering personnel and a step towards digitalization within the framework of Industry 4.0.
- 8 July 2025
- Published by Shakarim University in Bulletin of Shakarim University. Technical Sciences
- No. 2(18),p. 121-128
Open AccessDEVELOPMENT OF A PREDICTIVE MAINTENANCE SYSTEM BASED ON MACHINE LEARNING
In traditional industrial settings, maintenance methods based on reactive repairs or scheduled time intervals often lead to downtime and reduced efficiency. Therefore, predictive maintenance uses data to predict failures before they actually occur. The objective of this study is to develop an intelligent predictive maintenance system based on machine learning and implement it in the industry of Kazakhstan, focusing on the principles of Industry 4.0. The system is based on collecting data from sensors (current, temperature, pressure, vibration) using Siemens PLCs, integrated via the OPC UA protocol into the industrial IoT infrastructure. Additionally, computer vision is used to monitor the equipment condition in real time. The obtained data is processed by machine learning algorithms, including neural networks, linear regression, and autoencoders. To adapt the model to changes, it is trained continuously using Bayesian updating. Visualization and interaction with users are implemented via SCADA for engineers and Power BI for managers. In addition, the paper discusses the challenges associated with the deployment of predictive maintenance solutions and suggests future directions for improving scalability, security, and real-time data processing capabilities. The obtained results contribute to the growing body of research in the field of predictive maintenance, demonstrating its potential to improve efficiency, reduce operating costs, and support the transition to datadriven, intelligent manufacturing systems. The work demonstrates the potential of predictive maintenance as a solution for aging industries with a shortage of engineering personnel and a step towards digitalization within the framework of Industry 4.0.
- Open Access
- 30 October 2024
- Published by Private Company Technology Center in Eastern-European Journal of Enterprise Technologies
- Vol. 5 (2 (131)) , 103-111
Development of mathematical modeling of a mobile robot's motion control system
This study focuses on developing a mathematical model for precise control and stabilization of unmanned aerial vehicles (UAVs) in various spatial conditions. Addressing the problem of achieving precise control and stability, the proposed solution designs a control system based on a linear-quadratic controller (LQR) and simulates it using a proportional-derivative (PD) controller implemented in Matlab/Simulink. The results demonstrate high precision and stability in controlling the UAV motion parameters – roll, pitch, yaw, and altitude. This important level of performance is achieved due to the adaptivity of the LQR-based control system, which optimizes control actions according to the unsteady dynamics of the UAV. The integration of the PD controller improves responsiveness and stability, providing precise motion control over a range of spatial states. These features effectively solve the problem by handling the complex dynamics of the UAV and providing precise control. The results are explained by the ability of the LQR to provide optimal control laws that minimize deviations using a quadratic cost function, while the PD controller quickly corrects errors and responds to disturbances. The benefits of this approach include a significant reduction in control errors by about 25–30 %, increased response speed to external disturbances, and reduced computational latency due to efficient processing compared to more resource-intensive methods such as model predictive control. The developed mathematical model can be applied in practice in conditions requiring robust real-time control and adaptation to dynamic changes in the environment. It is especially suitable for industries such as logistics, surveillance, and environmental monitoring, providing an effective and optimal solution for stabilizing and controlling the motion of UAVs in various spatial states. This approach improves the performance of UAVs and expands their capabilities in various operating conditions.
- 30 October 2024
- Published by Private Company Technology Center in Eastern-European Journal of Enterprise Technologies
- Vol. 5 (2 (131)) , 103-111
Open AccessDevelopment of mathematical modeling of a mobile robot's motion control system
This study focuses on developing a mathematical model for precise control and stabilization of unmanned aerial vehicles (UAVs) in various spatial conditions. Addressing the problem of achieving precise control and stability, the proposed solution designs a control system based on a linear-quadratic controller (LQR) and simulates it using a proportional-derivative (PD) controller implemented in Matlab/Simulink. The results demonstrate high precision and stability in controlling the UAV motion parameters – roll, pitch, yaw, and altitude. This important level of performance is achieved due to the adaptivity of the LQR-based control system, which optimizes control actions according to the unsteady dynamics of the UAV. The integration of the PD controller improves responsiveness and stability, providing precise motion control over a range of spatial states. These features effectively solve the problem by handling the complex dynamics of the UAV and providing precise control. The results are explained by the ability of the LQR to provide optimal control laws that minimize deviations using a quadratic cost function, while the PD controller quickly corrects errors and responds to disturbances. The benefits of this approach include a significant reduction in control errors by about 25–30 %, increased response speed to external disturbances, and reduced computational latency due to efficient processing compared to more resource-intensive methods such as model predictive control. The developed mathematical model can be applied in practice in conditions requiring robust real-time control and adaptation to dynamic changes in the environment. It is especially suitable for industries such as logistics, surveillance, and environmental monitoring, providing an effective and optimal solution for stabilizing and controlling the motion of UAVs in various spatial states. This approach improves the performance of UAVs and expands their capabilities in various operating conditions.
- Open Access
- 5 September 2024
- Published by JSC Academy of Logistics and Transport in Вестник КазАТК
- Vol. 134 (5) , 393-402
SIMULATION OF COMPUTER VISION SYSTEMS WITH ARTIFICIAL INTELLIGENCE
. In recent years, the rapid development of artificial intelligence technologies has led to significant advances in the field of computer vision, dedicated to extracting valuable information from visual data. This article presents a comprehensive analysis of computer vision systems modeling, focusing on the integration of AI techniques to enhance their performance and capabilities. Through an extensive literature review, this research investigates the latest methodologies, algorithms, and architectures employed in computer vision systems to accomplish tasks such as object detection, recognition, tracking, and segmentation. The article first provides an overview of computer vision systems and their significance in various domains, including autonomous vehicles, surveillance, robotics, and medical imaging. It then delves into the advancements of AI, particularly deep learning techniques such as convolutional neural networks and recurrent neural networks, which have revolutionized the field of computer vision. The article highlights the effectiveness of AI models in improving the accuracy and efficiency of computer vision tasks, showcasing their superiority over traditional image processing techniques. Furthermore, this research explores the challenges and limitations faced in modeling computer vision systems with AI. Modeling of computer vision systems with elements of artificial intelligence is carried out in the Matlab/Simulink. Issues such as dataset bias, scalability, interpretability, and robustness are discussed, along with proposed solutions and ongoing research efforts. Finally, the article concludes with a comprehensive outlook on the future of computer vision systems modeling with AI.
- 5 September 2024
- Published by JSC Academy of Logistics and Transport in Вестник КазАТК
- Vol. 134 (5) , 393-402
Open AccessSIMULATION OF COMPUTER VISION SYSTEMS WITH ARTIFICIAL INTELLIGENCE
. In recent years, the rapid development of artificial intelligence technologies has led to significant advances in the field of computer vision, dedicated to extracting valuable information from visual data. This article presents a comprehensive analysis of computer vision systems modeling, focusing on the integration of AI techniques to enhance their performance and capabilities. Through an extensive literature review, this research investigates the latest methodologies, algorithms, and architectures employed in computer vision systems to accomplish tasks such as object detection, recognition, tracking, and segmentation. The article first provides an overview of computer vision systems and their significance in various domains, including autonomous vehicles, surveillance, robotics, and medical imaging. It then delves into the advancements of AI, particularly deep learning techniques such as convolutional neural networks and recurrent neural networks, which have revolutionized the field of computer vision. The article highlights the effectiveness of AI models in improving the accuracy and efficiency of computer vision tasks, showcasing their superiority over traditional image processing techniques. Furthermore, this research explores the challenges and limitations faced in modeling computer vision systems with AI. Modeling of computer vision systems with elements of artificial intelligence is carried out in the Matlab/Simulink. Issues such as dataset bias, scalability, interpretability, and robustness are discussed, along with proposed solutions and ongoing research efforts. Finally, the article concludes with a comprehensive outlook on the future of computer vision systems modeling with AI.