Soner Kotan
Soner Kotan has a total of 58 co-authors. Here are the top co-authors by number of shared publications:
- Ahmet Katilmiş
Marmara Üniversitesi
1 shared publication - Aydin Akan
Department of Electrical and Electronics Engineering, Izmir University of Economics, 35330 Izmir, Türkiye
1 shared publication - Ayşe Kübra Aktaş
Bartin Universitesi
1 shared publication - Başak Göfner
Milli Eğitim Bakanlığı
1 shared publication - Bilal Yorulmaz
Marmara Üniversitesi
1 shared publication - Birsen Berfu Akaydın
Kocaeli Üniversitesi, Eğitim Fakültesi
1 shared publication
Latest Publications
- Open Access
- 12 November 2026
- Published by Elsevier in European Journal of Radiology Artificial Intelligence
- Vol. 8
MammoBIRD: A comprehensive bilingual (TR/EN) mammography report corpus for BI-RADS classification and Turkish-language Named Entity Recognition
- 12 November 2026
- Published by Elsevier in European Journal of Radiology Artificial Intelligence
- Vol. 8
Open AccessMammoBIRD: A comprehensive bilingual (TR/EN) mammography report corpus for BI-RADS classification and Turkish-language Named Entity Recognition
- 3 July 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Neural Systems
- Vol. 37 (02) , 2750005
Deep Model Families for EEG-Based Multiclass Dementia Classification
Deep learning (DL) has shown considerable promise for EEG-based dementia assessment; however, rigorous cross-family comparisons under leakage-free and clinically meaningful evaluation protocols remain limited. To address this gap, we benchmarked twelve DL architectures from four major families—Recurrent Neural Networks (RNNs), Temporal Convolutional Networks (TCNs), Transformers, and Self-Supervised Learning (SSL)—for the multi-class differentiation of Alzheimer’s disease (AD), Frontotemporal Dementia (FTD), and Healthy Controls (HC). The analysis was conducted on a clinically acquired EEG dataset comprising 36 AD patients, 23 FTD patients, and 29 HC participants. To prevent subject-level data leakage, we employed strict subject-independent 5-fold cross-validation, and to improve clinical relevance, final predictions were aggregated at the subject level using majority voting across EEG segments. Model performance was evaluated using classification metrics, learning stability, statistical ranking, computational efficiency, and interpretability analysis. Among all benchmarked architectures, the self-supervised SimCLR-Time framework achieved the strongest subject-level performance, reaching a mean accuracy of [ ] and a weighted F1-score of [ ]. The Time-Series Transformer (TST) was the best-performing fully supervised model, with an accuracy of [ ], followed closely by the hybrid LSTM-FCN model at [ ]. In contrast, the EEG-Seq2Seq model showed the weakest performance [ ]. Computational profiling further showed that TCN-Attention offered a favorable trade-off between accuracy and edge-oriented efficiency, while SimCLR-Time combined strong diagnostic performance with the fastest inference latency. In addition, Integrated Gradients-based saliency analysis suggested that the top-performing model relied on structured spatiotemporal EEG patterns rather than arbitrary temporal artifacts. Overall, these findings provide a controlled within-dataset benchmark for multi-class dementia classification from raw EEG and suggest that self-supervised and attention-based architectures are promising directions for clinically oriented EEG decision-support systems. However, external validation on larger and more heterogeneous cohorts remains necessary before broader generalization can be claimed.
- 3 July 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Neural Systems
- Vol. 37 (02) , 2750005
Deep Model Families for EEG-Based Multiclass Dementia Classification
Deep learning (DL) has shown considerable promise for EEG-based dementia assessment; however, rigorous cross-family comparisons under leakage-free and clinically meaningful evaluation protocols remain limited. To address this gap, we benchmarked twelve DL architectures from four major families—Recurrent Neural Networks (RNNs), Temporal Convolutional Networks (TCNs), Transformers, and Self-Supervised Learning (SSL)—for the multi-class differentiation of Alzheimer’s disease (AD), Frontotemporal Dementia (FTD), and Healthy Controls (HC). The analysis was conducted on a clinically acquired EEG dataset comprising 36 AD patients, 23 FTD patients, and 29 HC participants. To prevent subject-level data leakage, we employed strict subject-independent 5-fold cross-validation, and to improve clinical relevance, final predictions were aggregated at the subject level using majority voting across EEG segments. Model performance was evaluated using classification metrics, learning stability, statistical ranking, computational efficiency, and interpretability analysis. Among all benchmarked architectures, the self-supervised SimCLR-Time framework achieved the strongest subject-level performance, reaching a mean accuracy of [ ] and a weighted F1-score of [ ]. The Time-Series Transformer (TST) was the best-performing fully supervised model, with an accuracy of [ ], followed closely by the hybrid LSTM-FCN model at [ ]. In contrast, the EEG-Seq2Seq model showed the weakest performance [ ]. Computational profiling further showed that TCN-Attention offered a favorable trade-off between accuracy and edge-oriented efficiency, while SimCLR-Time combined strong diagnostic performance with the fastest inference latency. In addition, Integrated Gradients-based saliency analysis suggested that the top-performing model relied on structured spatiotemporal EEG patterns rather than arbitrary temporal artifacts. Overall, these findings provide a controlled within-dataset benchmark for multi-class dementia classification from raw EEG and suggest that self-supervised and attention-based architectures are promising directions for clinically oriented EEG decision-support systems. However, external validation on larger and more heterogeneous cohorts remains necessary before broader generalization can be claimed.
- Open Access
- 1 January 2026
- Published by FapUNIFESP (SciELO) in Revista da Associação Médica Brasileira
- Vol. 72 (9)
Rational Drug Use Support System: design, artificial intelligence integration, and pilot feasibility trial
SUMMARY OBJECTIVE: The Rational Drug Use Support System was developed to assist patients in using the right drug, at the right time and at the right dose according to their treatment protocols. The aim of this study was to evaluate the performance of an artificial intelligence-enhanced mobile application designed to support rational drug use and to assess its feasibility and usability through a pilot trial. METHODS: The system includes an artificial intelligence -powered mobile application and a web-based interface integrating convolutional neural networks and object detection algorithms for human activity recognition. To build an image library, real patient videos were recorded in a hospital setting under physician supervision. A total of 357 original medication-intake videos were collected from 31 patients, which were then segmented into 1,626 short clips to train the artificial intelligence model. The system was first tested under controlled laboratory conditions and then evaluated for feasibility in a hospital pilot trial involving eight participants over 15 days. RESULTS: The proposed artificial intelligence model achieved 94% accuracy in recognizing medication intake activities in the test set, suggesting promising performance that warrants further evaluation. Exploratory analyses of the 36-Item Short Form Health Survey Questionnaire administered to patients who utilized the mobile application for 15 days revealed improvements in the domains of social functioning, vitality, and mental health. CONCLUSION: The findings suggest that the "Rational Drug Use Support System" appears feasible for implementation in telemedicine and clinical settings to support rational drug use and patient engagement in chronic disease management. However, larger controlled studies are warranted to confirm these preliminary findings.
- 1 January 2026
- Published by FapUNIFESP (SciELO) in Revista da Associação Médica Brasileira
- Vol. 72 (9)
Open AccessRational Drug Use Support System: design, artificial intelligence integration, and pilot feasibility trial
SUMMARY OBJECTIVE: The Rational Drug Use Support System was developed to assist patients in using the right drug, at the right time and at the right dose according to their treatment protocols. The aim of this study was to evaluate the performance of an artificial intelligence-enhanced mobile application designed to support rational drug use and to assess its feasibility and usability through a pilot trial. METHODS: The system includes an artificial intelligence -powered mobile application and a web-based interface integrating convolutional neural networks and object detection algorithms for human activity recognition. To build an image library, real patient videos were recorded in a hospital setting under physician supervision. A total of 357 original medication-intake videos were collected from 31 patients, which were then segmented into 1,626 short clips to train the artificial intelligence model. The system was first tested under controlled laboratory conditions and then evaluated for feasibility in a hospital pilot trial involving eight participants over 15 days. RESULTS: The proposed artificial intelligence model achieved 94% accuracy in recognizing medication intake activities in the test set, suggesting promising performance that warrants further evaluation. Exploratory analyses of the 36-Item Short Form Health Survey Questionnaire administered to patients who utilized the mobile application for 15 days revealed improvements in the domains of social functioning, vitality, and mental health. CONCLUSION: The findings suggest that the "Rational Drug Use Support System" appears feasible for implementation in telemedicine and clinical settings to support rational drug use and patient engagement in chronic disease management. However, larger controlled studies are warranted to confirm these preliminary findings.
- 1 January 2021
- Published by Pegem Akademi Yayıncılık Eğitim Danışmanlık Hizmetleri Ticaret A.Ş
Kuramdan Uygulamaya Örnek Olay Yaşamı Sınıfa Taşımak
Kuramdan Uygulamaya Örnek Olay: Yaşamı Sınıfa Taşımak Eğitim Programları ve Öğretim “Yaşamak nedir” çok felsefi bir soru olup bünyesinde olası binlerce yanıtı barındırmaktadır. Bu olası yanıtlardan biri de “bitip tükenmek bilmeyen problem/sorun çözme süreci”dir. Problem/sorun olarak görülecek durum, kadın/erkek,
- 1 January 2021
- Published by Pegem Akademi Yayıncılık Eğitim Danışmanlık Hizmetleri Ticaret A.Ş
Kuramdan Uygulamaya Örnek Olay Yaşamı Sınıfa Taşımak
Kuramdan Uygulamaya Örnek Olay: Yaşamı Sınıfa Taşımak Eğitim Programları ve Öğretim “Yaşamak nedir” çok felsefi bir soru olup bünyesinde olası binlerce yanıtı barındırmaktadır. Bu olası yanıtlardan biri de “bitip tükenmek bilmeyen problem/sorun çözme süreci”dir. Problem/sorun olarak görülecek durum, kadın/erkek,