Abdulsamet Aktas
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Abdulsamet Aktas has a total of 50 co-authors. Here are the top co-authors by number of shared publications:
- Hamza Osman Ilhan
Department of Computer Engineering, Yildiz Technical University, İstanbul 34220, Türkiye
6 shared publications - Gorkem Serbes
Department of Biomedical Engineering, Yildiz Technical University, Istanbul 34220, Türkiye
5 shared publications - Hakkı Uzun
Recep Tayyip Erdoğan University
4 shared publications - Merve Huner Yigit
Department of Medical Biochemistry, Faculty of Medicine, Recep Tayyip Erdoğan University, Rize 53100, Türkiye
3 shared publications - Merve Pinar
Marmara Üniversitesi
3 shared publications - Ayşe Berna Altınel Girgin
Marmara Üniversitesi
2 shared publications
Latest Publications
- 26 June 2026
- Published by Istanbul Ticaret Universitesi in İstanbul Ticaret Üniversitesi Fen Bilimleri Dergisi
- Vol. 25 (49) , 49-76
COMPARATIVE ANALYSIS OF DEEP LEARNING MODELS FOR ISCHEMIC STROKE SEGMENTATION IN MRI
Stroke is one of the major global health concerns and a leading cause of death and long-term disability. Ischemic stroke occurs when a blockage in cerebral blood flow results in neurological impairment. Early diagnosis and treatment are critical in stroke management, and Magnetic Resonance Imaging (MRI) plays an important role in this phase. However, manual MRI interpretation is highly dependent on expert radiologists and time-consuming, highlighting the need for AI-based lesion detection methods in healthcare. This study evaluates the performance of various models for ischemic stroke lesion detection: YOLOv8, YOLO11, and RT-DETR for object detection, and DeepLabv3+ and SegFormer for segmentation. To enhance data quality and model performance, a comprehensive preprocessing pipeline was implemented. This included 3D to 2D image conversion, intensity normalization, and lesion mask refinement. The models are compared using F1-score and mean precision (mAP) metrics on two public datasets: Anatomical Tracings of Lesions After Stroke (ATLAS_2.0) and Ischemic Stroke Lesion Segmentation (ISLES’22). This study aims to bridge the gap between AI-driven lesion detection and real-world clinical applications by determining the advantages and disadvantages of each approach and contributing to more efficient and accurate stroke diagnosis.
- 26 June 2026
- Published by Istanbul Ticaret Universitesi in İstanbul Ticaret Üniversitesi Fen Bilimleri Dergisi
- Vol. 25 (49) , 49-76
COMPARATIVE ANALYSIS OF DEEP LEARNING MODELS FOR ISCHEMIC STROKE SEGMENTATION IN MRI
Stroke is one of the major global health concerns and a leading cause of death and long-term disability. Ischemic stroke occurs when a blockage in cerebral blood flow results in neurological impairment. Early diagnosis and treatment are critical in stroke management, and Magnetic Resonance Imaging (MRI) plays an important role in this phase. However, manual MRI interpretation is highly dependent on expert radiologists and time-consuming, highlighting the need for AI-based lesion detection methods in healthcare. This study evaluates the performance of various models for ischemic stroke lesion detection: YOLOv8, YOLO11, and RT-DETR for object detection, and DeepLabv3+ and SegFormer for segmentation. To enhance data quality and model performance, a comprehensive preprocessing pipeline was implemented. This included 3D to 2D image conversion, intensity normalization, and lesion mask refinement. The models are compared using F1-score and mean precision (mAP) metrics on two public datasets: Anatomical Tracings of Lesions After Stroke (ATLAS_2.0) and Ischemic Stroke Lesion Segmentation (ISLES’22). This study aims to bridge the gap between AI-driven lesion detection and real-world clinical applications by determining the advantages and disadvantages of each approach and contributing to more efficient and accurate stroke diagnosis.
- Open Access
- 22 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 215748-215770
HeaRT: An Innovative Health Representation Transformer With Clustered Feature Augmentation for Biomedical Text Classification
Medical text classification (MTC) poses significant challenges in health informatics due to contextual complexity, class imbalance, and limited labeled data. This study introduces HeaRT (Health Representation Transformer), a novel and explainable classification framework tailored for sparsely labeled and semantically complex medical texts. To the best of our knowledge, this is the first study to systematically combine TF-based lexical weighting, SBERT contextual embeddings, and cluster-enhanced structural features within a unified multi-vector fusion architecture for MTC. HeaRT combines TF-based lexical weighting with SBERT embeddings to capture both lexical and contextual nuances. A hybrid feature selection mechanism based on ANOVA and SHAP is employed to reduce redundancy and improve interpretability. To further enhance representational capacity, cluster-derived features from K-means, DBSCAN, and Agglomerative Clustering are added, introducing topological structure awareness into the learning process. The proposed framework is evaluated on two benchmark datasets: Medical Abstracts and the Biomedical Text Dataset, and is compared against state-of-the-art models such as BERT, and Doc2Sequence. Experimental results reveal that HeaRT achieves an F1-score of 60.74% with AdaBoost on Medical Abstracts and 94.02% with LightGBM on the Biomedical Text Dataset. Paired t-test results confirm the statistical significance of these gains (p < 0.05). These findings establish HeaRT as a robust, interpretable, and extensible solution for medical text classification, with strong potential for deployment in clinical decision support, biomedical literature mining, and healthcare informatics applications.
- 22 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 215748-215770
Open AccessHeaRT: An Innovative Health Representation Transformer With Clustered Feature Augmentation for Biomedical Text Classification
Medical text classification (MTC) poses significant challenges in health informatics due to contextual complexity, class imbalance, and limited labeled data. This study introduces HeaRT (Health Representation Transformer), a novel and explainable classification framework tailored for sparsely labeled and semantically complex medical texts. To the best of our knowledge, this is the first study to systematically combine TF-based lexical weighting, SBERT contextual embeddings, and cluster-enhanced structural features within a unified multi-vector fusion architecture for MTC. HeaRT combines TF-based lexical weighting with SBERT embeddings to capture both lexical and contextual nuances. A hybrid feature selection mechanism based on ANOVA and SHAP is employed to reduce redundancy and improve interpretability. To further enhance representational capacity, cluster-derived features from K-means, DBSCAN, and Agglomerative Clustering are added, introducing topological structure awareness into the learning process. The proposed framework is evaluated on two benchmark datasets: Medical Abstracts and the Biomedical Text Dataset, and is compared against state-of-the-art models such as BERT, and Doc2Sequence. Experimental results reveal that HeaRT achieves an F1-score of 60.74% with AdaBoost on Medical Abstracts and 94.02% with LightGBM on the Biomedical Text Dataset. Paired t-test results confirm the statistical significance of these gains (p < 0.05). These findings establish HeaRT as a robust, interpretable, and extensible solution for medical text classification, with strong potential for deployment in clinical decision support, biomedical literature mining, and healthcare informatics applications.
- Open Access
- 10 September 2025
- Published by PeerJ in PeerJ Computer Science
- Vol. 11, e3173
Unveiling the capabilities of vision transformers in sperm morphology analysis: a comparative evaluation
Traditional sperm morphology assessment relies on manual visual inspection or semi-automated computer-aided sperm analysis (CASA) systems, which often require labor-intensive pre-processing steps. While recent machine learning approaches, particularly convolutional neural networks (CNNs), have improved feature extraction from sperm images, achieving a fully automated and highly accurate system remains challenging due to the complexity of sperm morphology and the need for specialized image adjustments. This study presents a novel, end-to-end automated sperm morphology analysis framework based on vision transformers (ViTs), which processes raw sperm images from two benchmark datasets-Human Sperm Head Morphology (HuSHeM) and Sperm Morphology Image Data Set (SMIDS)-without manual pre-processing. We conducted an extensive hyperparameter optimization study across eight ViT variants, evaluating learning rates, optimization algorithms, and data augmentation scales. Our experiments demonstrated that data augmentation significantly enhances ViT performance by improving generalization, particularly in limited-data scenarios. A comparative analysis of CNNs, hybrid models, and pure ViTs revealed that transformer-based architectures consistently outperform traditional methods. The BEiT_Base model achieved state-of-the-art accuracies of 92.5% (SMIDS) and 93.52% (HuSHeM), surpassing prior CNN-based approaches by 1.63% and 1.42%, respectively. Statistical significance (p < 0.05, t-test) confirmed these improvements. Visualization techniques (Attention Maps, Grad-CAM) further validated ViTs’ superior ability to capture long-range spatial dependencies and discriminative morphological features, such as head shape and tail integrity. Our work bridges a critical gap in reproductive medicine by delivering a scalable, fully automated solution that eliminates manual intervention while improving diagnostic accuracy. These findings underscore the potential of transformer-based models in clinical andrology, with implications for broader applications in biomedical image analysis.
- 10 September 2025
- Published by PeerJ in PeerJ Computer Science
- Vol. 11, e3173
Open AccessUnveiling the capabilities of vision transformers in sperm morphology analysis: a comparative evaluation
Traditional sperm morphology assessment relies on manual visual inspection or semi-automated computer-aided sperm analysis (CASA) systems, which often require labor-intensive pre-processing steps. While recent machine learning approaches, particularly convolutional neural networks (CNNs), have improved feature extraction from sperm images, achieving a fully automated and highly accurate system remains challenging due to the complexity of sperm morphology and the need for specialized image adjustments. This study presents a novel, end-to-end automated sperm morphology analysis framework based on vision transformers (ViTs), which processes raw sperm images from two benchmark datasets-Human Sperm Head Morphology (HuSHeM) and Sperm Morphology Image Data Set (SMIDS)-without manual pre-processing. We conducted an extensive hyperparameter optimization study across eight ViT variants, evaluating learning rates, optimization algorithms, and data augmentation scales. Our experiments demonstrated that data augmentation significantly enhances ViT performance by improving generalization, particularly in limited-data scenarios. A comparative analysis of CNNs, hybrid models, and pure ViTs revealed that transformer-based architectures consistently outperform traditional methods. The BEiT_Base model achieved state-of-the-art accuracies of 92.5% (SMIDS) and 93.52% (HuSHeM), surpassing prior CNN-based approaches by 1.63% and 1.42%, respectively. Statistical significance (p < 0.05, t-test) confirmed these improvements. Visualization techniques (Attention Maps, Grad-CAM) further validated ViTs’ superior ability to capture long-range spatial dependencies and discriminative morphological features, such as head shape and tail integrity. Our work bridges a critical gap in reproductive medicine by delivering a scalable, fully automated solution that eliminates manual intervention while improving diagnostic accuracy. These findings underscore the potential of transformer-based models in clinical andrology, with implications for broader applications in biomedical image analysis.
- Open Access
- 3 September 2025
- Published by MDPI AG in Diagnostics
- Vol. 15 (17) , 2234
Category-Aware Two-Stage Divide-and-Ensemble Framework for Sperm Morphology Classification
Introduction: Sperm morphology is a fundamental parameter in the evaluation of male infertility, offering critical insights into reproductive health. However, traditional manual assessments under microscopy are limited by operator dependency and subjective interpretation caused by biological variation. To overcome these limitations, there is a need for accurate and fully automated classification systems. Objectives: This study aims to develop a two-stage, fully automated sperm morphology classification framework that can accurately identify a wide spectrum of abnormalities. The framework is designed to reduce subjectivity, minimize misclassification between visually similar categories, and provide more reliable diagnostic support in reproductive healthcare. Methods: A novel two-stage deep learning-based framework is proposed utilizing images from three staining-specific versions of a comprehensive 18-class dataset. In the first stage, sperm images are categorized into two principal groups: (1) head and neck region abnormalities, and (2) normal morphology together with tail-related abnormalities. In the second stage, a customized ensemble model—integrating four distinct deep learning architectures, including DeepMind’s NFNet-F4 and vision transformer (ViT) variants—is employed for detailed abnormality classification. Unlike conventional majority voting, a structured multi-stage voting strategy is introduced to enhance decision reliability. Results: The proposed framework consistently outperforms single-model baselines, achieving accuracies of 69.43%, 71.34%, and 68.41% across the three staining protocols. These results correspond to a statistically significant 4.38% improvement over prior approaches in the literature. Moreover, the two-stage system substantially reduces misclassification among visually similar categories, demonstrating enhanced ability to detect subtle morphological variations. Conclusions: The proposed two-stage, ensemble-based framework provides a robust and accurate solution for automated sperm morphology classification. By combining hierarchical classification with structured decision fusion, the method advances beyond traditional and single-model approaches, offering a reliable and scalable tool for clinical decision-making in male fertility assessment.
- 3 September 2025
- Published by MDPI AG in Diagnostics
- Vol. 15 (17) , 2234
Open AccessCategory-Aware Two-Stage Divide-and-Ensemble Framework for Sperm Morphology Classification
Introduction: Sperm morphology is a fundamental parameter in the evaluation of male infertility, offering critical insights into reproductive health. However, traditional manual assessments under microscopy are limited by operator dependency and subjective interpretation caused by biological variation. To overcome these limitations, there is a need for accurate and fully automated classification systems. Objectives: This study aims to develop a two-stage, fully automated sperm morphology classification framework that can accurately identify a wide spectrum of abnormalities. The framework is designed to reduce subjectivity, minimize misclassification between visually similar categories, and provide more reliable diagnostic support in reproductive healthcare. Methods: A novel two-stage deep learning-based framework is proposed utilizing images from three staining-specific versions of a comprehensive 18-class dataset. In the first stage, sperm images are categorized into two principal groups: (1) head and neck region abnormalities, and (2) normal morphology together with tail-related abnormalities. In the second stage, a customized ensemble model—integrating four distinct deep learning architectures, including DeepMind’s NFNet-F4 and vision transformer (ViT) variants—is employed for detailed abnormality classification. Unlike conventional majority voting, a structured multi-stage voting strategy is introduced to enhance decision reliability. Results: The proposed framework consistently outperforms single-model baselines, achieving accuracies of 69.43%, 71.34%, and 68.41% across the three staining protocols. These results correspond to a statistically significant 4.38% improvement over prior approaches in the literature. Moreover, the two-stage system substantially reduces misclassification among visually similar categories, demonstrating enhanced ability to detect subtle morphological variations. Conclusions: The proposed two-stage, ensemble-based framework provides a robust and accurate solution for automated sperm morphology classification. By combining hierarchical classification with structured decision fusion, the method advances beyond traditional and single-model approaches, offering a reliable and scalable tool for clinical decision-making in male fertility assessment.
- 28 August 2025
- Published by Springer Nature in Journal of Imaging Informatics in Medicine
- Vol. 39 (3) , 2537-2552
Detection of Jaw Lesions on Panoramic Radiographs Using Deep Learning Method
This study aimed to evaluate and compare the performance of state-of-the-art deep learning models for detecting and segmenting both radiolucent and radiopaque jaw lesions on panoramic radiographs. A total of 2371 anonymized panoramic radiographs containing jaw lesions were retrospectively collected and categorized into radiolucent and radiopaque datasets. Expert annotation was performed to delineate lesion boundaries and assign anatomical localization (anterior/posterior maxilla and mandible). Four deep learning architectures—YOLOv8, YOLOv11, Mask R-CNN, and RT-DETR—were trained and evaluated under three experimental scenarios: (I) training without spatial labels, (II) data augmentation with unlabeled background images, and (III) inclusion of spatial localization annotations. Performance metrics included precision, recall, F1-score, and mean average precision (mAP@0.5 and mAP@0.5–0.95), with paired t-tests used for statistical comparison. In Scenario I, YOLOv11x-seg and YOLOv8x-seg achieved the highest segmentation performance for radiolucent and radiopaque lesions, respectively. For detection, YOLOv8x performed best on radiolucent lesions, while RT-DETR-L outperformed others on radiopaque lesions. In Scenario II, while YOLOv8x-seg achieved the best segmentation results across both lesion types, RT-DETR-L demonstrated superior detection performance, particularly for radiolucent lesions. In Scenario III, RT-DETR-L consistently outperformed all models across both lesion types. This study demonstrates the potential of state-of-the-art deep learning models for effective detection of lesions in panoramic radiographs. The developed models may offer valuable support to clinicians in lesion evaluation; however, it is recommended that they be employed primarily as decision support tools within clinical workflows, rather than as standalone diagnostic systems.
- 28 August 2025
- Published by Springer Nature in Journal of Imaging Informatics in Medicine
- Vol. 39 (3) , 2537-2552
Detection of Jaw Lesions on Panoramic Radiographs Using Deep Learning Method
This study aimed to evaluate and compare the performance of state-of-the-art deep learning models for detecting and segmenting both radiolucent and radiopaque jaw lesions on panoramic radiographs. A total of 2371 anonymized panoramic radiographs containing jaw lesions were retrospectively collected and categorized into radiolucent and radiopaque datasets. Expert annotation was performed to delineate lesion boundaries and assign anatomical localization (anterior/posterior maxilla and mandible). Four deep learning architectures—YOLOv8, YOLOv11, Mask R-CNN, and RT-DETR—were trained and evaluated under three experimental scenarios: (I) training without spatial labels, (II) data augmentation with unlabeled background images, and (III) inclusion of spatial localization annotations. Performance metrics included precision, recall, F1-score, and mean average precision (mAP@0.5 and mAP@0.5–0.95), with paired t-tests used for statistical comparison. In Scenario I, YOLOv11x-seg and YOLOv8x-seg achieved the highest segmentation performance for radiolucent and radiopaque lesions, respectively. For detection, YOLOv8x performed best on radiolucent lesions, while RT-DETR-L outperformed others on radiopaque lesions. In Scenario II, while YOLOv8x-seg achieved the best segmentation results across both lesion types, RT-DETR-L demonstrated superior detection performance, particularly for radiolucent lesions. In Scenario III, RT-DETR-L consistently outperformed all models across both lesion types. This study demonstrates the potential of state-of-the-art deep learning models for effective detection of lesions in panoramic radiographs. The developed models may offer valuable support to clinicians in lesion evaluation; however, it is recommended that they be employed primarily as decision support tools within clinical workflows, rather than as standalone diagnostic systems.
- Open Access
- 10 July 2025
- Published by Wiley in Advanced Intelligent Systems
- Vol. 7 (12)
Hi‐LabSpermTracking: A Novel and High‐Quality Sperm Tracking Dataset with an Advanced Ensemble Detection and Tracking Approach for Real‐World Clinical Scenarios
Sperm motility, a critical factor in diagnosing male infertility, requires computer‐based solutions due to the limitations of manual evaluation methods. This study introduces the Hi‐LabSpermTracking dataset, comprising 66 videos (60 s each, 10 fps) collected from 14 patients and meticulously annotated by experts. Unlike similar datasets, these uninterrupted, long‐duration videos enable continuous tracking of individual sperm cells, each assigned a unique ID throughout the video, supporting both sperm detection and tracking tasks. Experimental evaluations employ you only look once v8 (YOLOv8), real‐time detection transformer, and simple online and realtime tracking with a deep association metric across three scenarios. In Scenario I (sperm detection), the YOLOv8n model achieves 98.9% mAP50 and 97.9% F1‐score. In Scenario II (sperm tracking), performance metrics include 83.88% mAP50, 87.63% F1‐score, 72.27% higher order tracking accuracy (HOTA), and 77.88% multiple object tracking accuracy (MOTA). Scenario III simulates real‐world challenges by separating training and testing videos. Ensemble methods are applied, with the proposed mean ensemble achieving superior results: 86.55% mAP50, 87.87% F1‐score, 66.66% HOTA, and 76.42% MOTA. The Hi‐LabSpermTracking dataset enables robust sperm tracking research, while the mean ensemble method amplifies accuracy by uniting model strengths.
- 10 July 2025
- Published by Wiley in Advanced Intelligent Systems
- Vol. 7 (12)
Open AccessHi‐LabSpermTracking: A Novel and High‐Quality Sperm Tracking Dataset with an Advanced Ensemble Detection and Tracking Approach for Real‐World Clinical Scenarios
Sperm motility, a critical factor in diagnosing male infertility, requires computer‐based solutions due to the limitations of manual evaluation methods. This study introduces the Hi‐LabSpermTracking dataset, comprising 66 videos (60 s each, 10 fps) collected from 14 patients and meticulously annotated by experts. Unlike similar datasets, these uninterrupted, long‐duration videos enable continuous tracking of individual sperm cells, each assigned a unique ID throughout the video, supporting both sperm detection and tracking tasks. Experimental evaluations employ you only look once v8 (YOLOv8), real‐time detection transformer, and simple online and realtime tracking with a deep association metric across three scenarios. In Scenario I (sperm detection), the YOLOv8n model achieves 98.9% mAP50 and 97.9% F1‐score. In Scenario II (sperm tracking), performance metrics include 83.88% mAP50, 87.63% F1‐score, 72.27% higher order tracking accuracy (HOTA), and 77.88% multiple object tracking accuracy (MOTA). Scenario III simulates real‐world challenges by separating training and testing videos. Ensemble methods are applied, with the proposed mean ensemble achieving superior results: 86.55% mAP50, 87.87% F1‐score, 66.66% HOTA, and 76.42% MOTA. The Hi‐LabSpermTracking dataset enables robust sperm tracking research, while the mean ensemble method amplifies accuracy by uniting model strengths.
- Open Access
- 19 June 2025
- Published by MDPI AG in Diagnostics
- Vol. 15 (12) , 1564
Advanced Multi-Level Ensemble Learning Approaches for Comprehensive Sperm Morphology Assessment
Introduction: Fertility is fundamental to human well-being, significantly impacting both individual lives and societal development. In particular, sperm morphology—referring to the shape, size, and structural integrity of sperm cells—is a key indicator in diagnosing male infertility and selecting viable sperm in assisted reproductive technologies such as in vitro fertilisation (IVF) and intracytoplasmic sperm injection (ICSI). However, traditional manual evaluation methods are highly subjective and inconsistent, creating a need for standardized, automated systems. Objectives: This study aims to develop a robust and fully automated sperm morphology classification framework capable of accurately identifying a wide range of morphological abnormalities, thereby minimizing observer variability and improving diagnostic support in reproductive healthcare. Methods: We propose a novel ensemble-based classification approach that combines convolutional neural network (CNN)-derived features using both feature-level and decision-level fusion techniques. Features extracted from multiple EfficientNetV2 variants are fused and classified using Support Vector Machines (SVM), Random Forest (RF), and Multi-Layer Perceptron with Attention (MLP-Attention). Decision-level fusion is achieved via soft voting to enhance robustness and accuracy. Results: The proposed ensemble framework was evaluated using the Hi-LabSpermMorpho dataset, which contains 18 distinct sperm morphology classes. The fusion-based model achieved an accuracy of 67.70%, significantly outperforming individual classifiers. The integration of multiple CNN architectures and ensemble techniques effectively mitigated class imbalance and enhanced the generalizability of the model. Conclusions: The presented methodology demonstrates a substantial improvement over traditional and single-model approaches in automated sperm morphology classification. By leveraging ensemble learning and multi-level fusion, the model provides a reliable and scalable solution for clinical decision-making in male fertility assessment.
- 19 June 2025
- Published by MDPI AG in Diagnostics
- Vol. 15 (12) , 1564
Open AccessAdvanced Multi-Level Ensemble Learning Approaches for Comprehensive Sperm Morphology Assessment
Introduction: Fertility is fundamental to human well-being, significantly impacting both individual lives and societal development. In particular, sperm morphology—referring to the shape, size, and structural integrity of sperm cells—is a key indicator in diagnosing male infertility and selecting viable sperm in assisted reproductive technologies such as in vitro fertilisation (IVF) and intracytoplasmic sperm injection (ICSI). However, traditional manual evaluation methods are highly subjective and inconsistent, creating a need for standardized, automated systems. Objectives: This study aims to develop a robust and fully automated sperm morphology classification framework capable of accurately identifying a wide range of morphological abnormalities, thereby minimizing observer variability and improving diagnostic support in reproductive healthcare. Methods: We propose a novel ensemble-based classification approach that combines convolutional neural network (CNN)-derived features using both feature-level and decision-level fusion techniques. Features extracted from multiple EfficientNetV2 variants are fused and classified using Support Vector Machines (SVM), Random Forest (RF), and Multi-Layer Perceptron with Attention (MLP-Attention). Decision-level fusion is achieved via soft voting to enhance robustness and accuracy. Results: The proposed ensemble framework was evaluated using the Hi-LabSpermMorpho dataset, which contains 18 distinct sperm morphology classes. The fusion-based model achieved an accuracy of 67.70%, significantly outperforming individual classifiers. The integration of multiple CNN architectures and ensemble techniques effectively mitigated class imbalance and enhanced the generalizability of the model. Conclusions: The presented methodology demonstrates a substantial improvement over traditional and single-model approaches in automated sperm morphology classification. By leveraging ensemble learning and multi-level fusion, the model provides a reliable and scalable solution for clinical decision-making in male fertility assessment.
- 1 January 2025
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 186, 109603
Feature efficiency in IoMT security: A comprehensive framework for threat detection with DNN and ML
- 1 January 2025
- Published by Elsevier in Computers in Biology and Medicine
- Vol. 186, 109603
Feature efficiency in IoMT security: A comprehensive framework for threat detection with DNN and ML
- Open Access
- 23 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 196070-196091
Hi-LabSpermMorpho: A Novel Expert-Labeled Dataset With Extensive Abnormality Classes for Deep Learning-Based Sperm Morphology Analysis
Sperm morphology is crucial in semen analysis for diagnosing male infertility. To reduce limitations in visual assessment, such as variability in biological conditions and the biologist’s experience, developing computer-based sperm analysis techniques is imperative. In this study, a total of 49345 RGB sperm morphology patches were obtained using the proposed image acquisition technique and three different Diff-Quick staining methods: BesLab, Histoplus, and GBL. The images were labeled by experts under 18 classes, including sperm head, neck, and tail abnormality types, along with a normal class. The head category includes amorphous, tapered, double, pyriform, pin, vacuolated, narrow acrosome, and round. The neck category encompasses thin, thick, twisted, and asymmetrical. The tail category includes double, curly, long, short, and twisted. The Efficient-V2-Medium achieved accuracy rates of 65.05% and 67.42% on the BesLab and Histoplus datasets, respectively, while the GBL dataset yielded an accuracy of 63.58% using the Efficient-V2-Small. This study experimentally demonstrates that the Histoplus staining method is more suitable for deep learning-based automated analysis systems. As a reference for future studies, 35 different deep learning architectures were trained on the proposed dataset, establishing a classification baseline. The results show that the dataset can be successfully applied to complex deep learning models. Additionally, it addresses the absence of a large-scale sperm morpholgy analysis public datasets and can serve as a standard benchmark for future studies.
- 23 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 196070-196091
Open AccessHi-LabSpermMorpho: A Novel Expert-Labeled Dataset With Extensive Abnormality Classes for Deep Learning-Based Sperm Morphology Analysis
Sperm morphology is crucial in semen analysis for diagnosing male infertility. To reduce limitations in visual assessment, such as variability in biological conditions and the biologist’s experience, developing computer-based sperm analysis techniques is imperative. In this study, a total of 49345 RGB sperm morphology patches were obtained using the proposed image acquisition technique and three different Diff-Quick staining methods: BesLab, Histoplus, and GBL. The images were labeled by experts under 18 classes, including sperm head, neck, and tail abnormality types, along with a normal class. The head category includes amorphous, tapered, double, pyriform, pin, vacuolated, narrow acrosome, and round. The neck category encompasses thin, thick, twisted, and asymmetrical. The tail category includes double, curly, long, short, and twisted. The Efficient-V2-Medium achieved accuracy rates of 65.05% and 67.42% on the BesLab and Histoplus datasets, respectively, while the GBL dataset yielded an accuracy of 63.58% using the Efficient-V2-Small. This study experimentally demonstrates that the Histoplus staining method is more suitable for deep learning-based automated analysis systems. As a reference for future studies, 35 different deep learning architectures were trained on the proposed dataset, establishing a classification baseline. The results show that the dataset can be successfully applied to complex deep learning models. Additionally, it addresses the absence of a large-scale sperm morpholgy analysis public datasets and can serve as a standard benchmark for future studies.
- Open Access
- 29 October 2024
- Published by Turkish Journal of Internal Medicine in Turkish Journal of Internal Medicine
- Vol. 6 (4) , 144-154
Machine Learning to Predict Disease Severity and Progression in Hospitalized COVID-19 Patients Using Laboratory Data on Admission
Background: Herein, we aimed to develop and test machine learning (ML) models to predict disease severity and/or progression in hospitalized COVID-19 patients through baseline laboratory features. Methods: In this retrospective study of hospitalized COVID-19 patients admitted to a tertiary care center, we evaluated routine admission data to determine the accuracy rates of different ML algorithms: k-nearest neighbor classifier, bagging classifier, random forest (RF), and decision tree. These models were compared over three outcomes: those who needed oxygen supplementation vs. who did not on admission (Analysis 1, n: 180), those who later developed oxygen requirement vs. those who did not (Analysis 2, n: 112), and those who needed invasive mechanical ventilation vs. those who did not during hospitalization (Analysis 3, n: 164). Results: The median age of the patients was 55 (44-68) years, with males constituting 47.2% of the subjects. At admission, 37.8% of the patients required oxygen supplementation. During hospitalization, 17.5% needed mechanical ventilation, and 8.3% died. For all analyses, RF had the highest accuracy in classifying the need for oxygen supplementation on admission (89.4%) or during hospitalization (91.1%) and for invasive mechanical ventilation (92.2%). These were followed by a bagging classifier for Analysis 1 (88.3%) and Analysis 3 (91.0%) and by a decision tree for Analysis 2 (88.4%). C-reactive protein, monocyte distribution width, and high-sensitive troponin-T were the most crucial laboratory contributors to Analysis 1, Analysis 2, and Analysis 3, respectively. Conclusion: Our study showed that ML algorithms could predict the need for oxygen supplementation and mechanical ventilation during hospitalization using baseline laboratory data, suggesting a slight superiority of RF, among others.
- 29 October 2024
- Published by Turkish Journal of Internal Medicine in Turkish Journal of Internal Medicine
- Vol. 6 (4) , 144-154
Open AccessMachine Learning to Predict Disease Severity and Progression in Hospitalized COVID-19 Patients Using Laboratory Data on Admission
Background: Herein, we aimed to develop and test machine learning (ML) models to predict disease severity and/or progression in hospitalized COVID-19 patients through baseline laboratory features. Methods: In this retrospective study of hospitalized COVID-19 patients admitted to a tertiary care center, we evaluated routine admission data to determine the accuracy rates of different ML algorithms: k-nearest neighbor classifier, bagging classifier, random forest (RF), and decision tree. These models were compared over three outcomes: those who needed oxygen supplementation vs. who did not on admission (Analysis 1, n: 180), those who later developed oxygen requirement vs. those who did not (Analysis 2, n: 112), and those who needed invasive mechanical ventilation vs. those who did not during hospitalization (Analysis 3, n: 164). Results: The median age of the patients was 55 (44-68) years, with males constituting 47.2% of the subjects. At admission, 37.8% of the patients required oxygen supplementation. During hospitalization, 17.5% needed mechanical ventilation, and 8.3% died. For all analyses, RF had the highest accuracy in classifying the need for oxygen supplementation on admission (89.4%) or during hospitalization (91.1%) and for invasive mechanical ventilation (92.2%). These were followed by a bagging classifier for Analysis 1 (88.3%) and Analysis 3 (91.0%) and by a decision tree for Analysis 2 (88.4%). C-reactive protein, monocyte distribution width, and high-sensitive troponin-T were the most crucial laboratory contributors to Analysis 1, Analysis 2, and Analysis 3, respectively. Conclusion: Our study showed that ML algorithms could predict the need for oxygen supplementation and mechanical ventilation during hospitalization using baseline laboratory data, suggesting a slight superiority of RF, among others.