Solak, Ahmet
Loading...
Profile URL
Name Variants
Job Title
Dr. Öğr. Üyesi
Email Address
asolak@ktun.edu.tr
Main Affiliation
02.04. Department of Electrical and Electronics Engineering
Status
Current Staff
ORCID ID
Scopus Author ID
Turkish CoHE Profile ID
Google Scholar ID
WoS Researcher ID
Research Topics
Domains
Health SciencesPhysical Sciences
Fields
MedicineComputer Science
Subfields
Radiology, Nuclear Medicine and ImagingArtificial IntelligenceComputer Vision and Pattern RecognitionAnesthesiology and Pain Medicine
Specific Research Areas
COVID-19 diagnosis using AI
Radiomics and Machine Learning in Medical Imaging
AI in cancer detection
Advanced Neural Network Applications
Anesthesia and Sedative Agents
Sustainable Development Goals
1NO POVERTY
0
Research Products
2ZERO HUNGER
0
Research Products
3GOOD HEALTH AND WELL-BEING
8
Research Products
4QUALITY EDUCATION
0
Research Products
5GENDER EQUALITY
0
Research Products
6CLEAN WATER AND SANITATION
0
Research Products
7AFFORDABLE AND CLEAN ENERGY
0
Research Products
8DECENT WORK AND ECONOMIC GROWTH
0
Research Products
9INDUSTRY, INNOVATION AND INFRASTRUCTURE
0
Research Products
10REDUCED INEQUALITIES
0
Research Products
11SUSTAINABLE CITIES AND COMMUNITIES
0
Research Products
12RESPONSIBLE CONSUMPTION AND PRODUCTION
0
Research Products
13CLIMATE ACTION
0
Research Products
14LIFE BELOW WATER
0
Research Products
15LIFE ON LAND
0
Research Products
16PEACE, JUSTICE AND STRONG INSTITUTIONS
0
Research Products
17PARTNERSHIPS FOR THE GOALS
1
Research Products
Documents
7
Citations
16
h-index
2
Documents
8
Citations
22
Publication Collaboration
| Affiliation Name | Count |
|---|---|
| Konya Technical University | 7 |
| Selçuk University | 3 |
| Chemnitz University of Technology | 1 |
| ETH Zurich | 1 |
| Inspire | 1 |
1 / 2
Data obtained from OpenAlex
Scholarly Output
14
Articles
10
Views / Downloads
35/30
Supervised MSc Theses
1
Supervised PhD Theses
1
WoS Citation Count
15
Scopus Citation Count
14
Patents
0
Projects
0
WoS Citations per Publication
1.07
Scopus Citations per Publication
1.00
Open Access Source
8
Supervised Theses
2
| Journal | Count |
|---|---|
| 2020 28TH SIGNAL PROCESSING AND COMMUNICATIONS APPLICATIONS CONFERENCE (SIU) | 1 |
| 28th Signal Processing and Communications Applications Conference (SIU) -- OCT 05-07, 2020 -- ELECTR NETWORK | 1 |
| Biomimetics | 1 |
| Düzce Üniversitesi Bilim ve Teknoloji Dergisi | 1 |
| Environmental Earth Sciences | 1 |
Current Page: 1 / 3
Scopus Quartile Distribution
Competency Cloud
14 results
Scholarly Output Search Results
Now showing 1 - 10 of 14
Article Citation - WoS: 1Citation - Scopus: 1Ensemble-Based Hybrid Deep Learning for Monkeypox Detection: Merging Instance-Normalized Transformers With CNNs for Enhanced Diagnostic Precision(Springer London Ltd, 2025-06-05) Solak, AhmetMonkeypox has re-emerged as a global public health threat, particularly in regions lacking extensive laboratory infrastructure. To address the urgent need for rapid, non-invasive diagnosis, we introduce a hybrid deep-learning framework that fuses an instance-normalized vision transformer (IN-ViT) with ResNet-50. Our approach first applies instance normalization within each transformer encoder to stabilize per-patch feature statistics, then concatenates these global contextual embeddings with ResNet-50's locally extracted features via a lightweight multilayer perceptron. We evaluate performance on the publicly available Monkeypox Skin Lesion Dataset, comprising 3192 augmented images of monkeypox, chickenpox, and measles lesions, partitioned into 70% train, 10% validation, and 20% test sets. Against standalone baselines-VGG-16, VGG-19, ResNet-50, and a standard ViT-our IN-ViT + ResNet-50 ensemble achieves 96.26% accuracy, 96.35% precision, 96.26% recall, and 96.24% F1-score, representing a >= 1% improvement over prior state-of-the-art. Crucially, the model sustains real-time inference (similar to 30 ms per image on Tesla T4 GPU) and can be readily deployed in telemedicine or point-of-care screening. These results demonstrate that combining fine-grained instance normalization with feature-level fusion yields a robust and interpretable diagnostic tool. Future work will explore federated learning for cross-site generalization, advanced data-augmentation regimes to mitigate class imbalance, and clinical validation across diverse patient populations.Article Citation - WoS: 2Citation - Scopus: 2Adrenal Lesion Classification With Abdomen Caps and the Effect of Roi Size(Springer, 2023-04-25) Solak, Ahmet; Ceylan, Rahime; Bozkurt, Mustafa Alper; Cebeci, Hakan; Koplay, MustafaAccurate classification of adrenal lesions on magnetic resonance (MR) images are very important for diagnosis and treatment planning. The detection and classification of lesions in medical imaging heavily rely on several key factors, including the specialist's level of experience, work intensity, and fatigue of the clinician. These factors are critical determinants of the accuracy and effectiveness of the diagnostic process, which in turn has a direct impact on patient health outcomes. With the spread of artificial intelligence, the use of computer-aided diagnosis (CAD) systems in disease diagnosis has also increased. In this study, adrenal lesion classification was performed using deep learning on MR images. The data set used was obtained from the Department of Radiology, Faculty of Medicine, Selcuk University, and all adrenal lesions were identified and reviewed in consensus by two radiologists experienced with abdominal MR. Studies were carried out on two different data sets created by T1- and T2-weighted MR images. The data set consisted of 112 benign and 10 malignant lesions for each mode. Experiments were performed with regions of interest (ROIs) of different sizes to increase the working performance. Thus, the effect of the selected ROI size on the classification performance was assessed. In addition, instead of the convolutional neural network (CNN) models used in deep learning, a unique classification model structure called Abdomen Caps was proposed. When the data sets used in classification studies are manually separated for training, validation, and testing, different results are obtained with different data sets for each stage. To eliminate this imbalance, tenfold cross-validation was used in this study. The best results obtained were 0.982, 0.999, 0.969, 0.983, 0.998, and 0.964 for accuracy, precision, recall, F1-score, area under the curve (AUC) score, and kappa score, respectively.Conference Object Citation - WoS: 1Citation - Scopus: 1Classification of Mammography Images by Transfer Learning(IEEE, 2020) Solak, Ahmet; Ceylan, RahimeBreast cancer is the most common cancer type in women worldwide. Diagnosis and early detection of cancer by mammography images are of great importance in cancer treatment. The use of deep learning in Computer Assisted Diagnostic systems has gained a great momentum especially since 2012. In this study, benign and malignant mass images were reproduced with data augmentation and the data sets obtained were classified with deep learning networks. In this study, a scratch Convolutional Neural Network (CNN) architecture was created and transfer learning was realized with different network models which trained on IMAGENET images. In the transfer learning section, separate training results were obtained by performing feature extraction and fine tuning of network parameters. As a result of the study, the best results were obtained with MobileNet, NASNetLarge and InceptionResNetV2 models which are used in transfer learning models.Doctoral Thesis Tıbbi Görüntü İşlemede Kapsül Ağlar(Konya Teknik Üniversitesi, 2023) Solak, Ahmet; Ceylan, Rahi̇meBu doktora tezi, farklı görüntüleme yöntemlerinden elde edilen tıbbi görüntülerin otomatik sınıflandırılması ve segmentasyonu için Bilgisayar Destekli Tanı (BDT) sistemlerinin geliştirilmesine odaklanmaktadır. Tıbbi görüntüleme, vücudun iç yapılarını ve işlevlerini görselleştirmek için invazif olmayan araçlar sağlayarak sağlık hizmetlerinde kritik bir rol oynamaktadır. Bununla birlikte, tıbbi görüntülerin radyologlar tarafından analizi ve yorumlanması genellikle öznel ve zaman alıcıdır. Bu nedenle bu tez, tıbbi görüntülerdeki ilgi alanlarının tespitini ve sınıflandırılmasını otomatikleştirmek için yapay zekanın, özellikle de derin öğrenme modellerinin kullanımını araştırmaktadır. Tez, her biri farklı bir tıbbi görüntüleme yöntemine odaklanan dört çalışma içermektedir. İlk çalışma, mamogram görüntülerindeki iyi huylu/kötü huylu kitlelerin klasik Konvolüsyonel Sinir Ağları (KSA) ve transfer öğrenme modelleri kullanılarak sınıflandırılmasını ve ardından bu tezde geliştirilen kapsül ağ modeli kullanılarak sınıflandırılmasını içermektedir. İkinci çalışma, modifiye edilmiş bir U-Net modeli kullanarak kolonoskopi görüntülerindeki poliplerin segmentasyonuna ve segmentasyon performansını optimize etmek için farklı parametrelerin analizine sonrasında kapsül ağ tabanlı bir segmentasyon modelinin performansına odaklanmaktadır. Üçüncü çalışmada, abdominal MR görüntülerindeki iyi huylu/kötü huylu adrenal lezyonların hem sınıflandırılması hem de segmentasyonu incelenmiştir. Sınıflandırma için farklı ilgi bölgeleri çıkarılmış, kapsül ağı tabanlı ve KSA tabanlı modeller kullanılarak ayrı çalışmalar yapılmıştır. Segmentasyon için klasik U-Net modeli modifiye edilerek yeni bir model önerilmiş, farklı parametrelerin ve kapsül ağ tabanlı segmentasyon modelinin performans üzerindeki etkisi değerlendirilmiştir. Son olarak, X-ışını görüntüleri kullanılarak çocuklarda pnömoni sınıflandırması için özel bir kapsül ağ yapısı önerilmiştir. Tüm çalışmalar literatürdeki benzer çalışmalarla veya en son modellerle karşılaştırılmış ve üstünlükleri gösterilmiştir. Genel olarak bu tez, tıbbi görüntülerin otomatik sınıflandırılması ve segmentasyonu için tıbbi teşhislerin doğruluğunu, verimliliğini ve tutarlılığını potansiyel olarak artırabilecek yeni derin öğrenme modelleri sunmaktadır.Article HCHS-Net: a Multimodal Handcrafted Feature and Metadata Framework for Interpretable Skin Lesion Classification(Multidisciplinary Digital Publishing Institute (MDPI), 2026-02-19) Solak A.; Solak, AhmetAccurate and timely classification of skin lesions is critical for early cancer detection, yet current deep learning approaches suffer from high computational costs, limited interpretability, and poor transparency for clinical deployment. This study presents HCHS-Net, a lightweight and interpretable multimodal framework for six-class skin lesion classification on the PAD-UFES-20 dataset. The proposed framework extracts a 116-dimensional visual feature vector through three complementary handcrafted modules: a Color Module employing multi-channel histogram analysis to capture chromatic diagnostic patterns, a Haralick Module deriving texture descriptors from the gray-level co-occurrence matrix (GLCM) that quantify surface characteristics correlated with malignancy, and a Shape Module encoding morphological properties via Hu moment invariants aligned with the clinical ABCD rule. The architectural design of HCHS-Net adopts a biomimetic approach by emulating the hierarchical information processing of the human visual system and the cognitive diagnostic workflows of expert dermatologists. Unlike conventional black-box deep learning models, this framework employs parallel processing branches that simulate the selective attention mechanisms of the human eye by focusing on biologically significant visual cues such as chromatic variance, textural entropy, and morphological asymmetry. These visual features are concatenated with a 12-dimensional clinical metadata vector encompassing patient demographics and lesion characteristics, yielding a compact 128-dimensional multimodal representation. Classification is performed through an ensemble of three gradient boosting algorithms (XGBoost, LightGBM, CatBoost) with majority voting. HCHS-Net achieves 97.76% classification accuracy with only 0.25 M parameters, outperforming deep learning baselines, including VGG-16 (94.60%), ResNet-50 (94.80%), and EfficientNet-B2 (95.16%), which require 60–97× more parameters. The framework delivers an inference time of 0.11 ms per image, enabling real-time classification on standard CPUs without GPU acceleration. Ablation analysis confirms the complementary contribution of each feature module, with metadata integration providing a 2.53% accuracy gain. The model achieves perfect melanoma and nevus recall (100%) with 99.55% specificity, maintaining reliable discrimination at safety-critical diagnostic boundaries. Comprehensive benchmarking against 13 published methods demonstrates that domain-informed handcrafted features combined with clinical metadata can match or exceed deep learning fusion approaches while offering superior interpretability and computational efficiency for point-of-care deployment. © 2026 by the author.Article Citation - WoS: 7Citation - Scopus: 8A Sensitivity Analysis for Polyp Segmentation With U-Net(Springer, 2023-07-29) Solak, Ahmet; Ceylan, RahimeColorectal Cancer (CRC) is one of the most common cancer diseases in the world. Early diagnosis of the disease is of great importance for the recovery of the patient. Colonoscopy is the gold standard procedure used in the diagnosis of CRC. In this context, this study focused on the detection of polyps with high accuracy in order to contribute to the early diagnosis of CRC. Within the scope of the study, polyp segmentation was performed on the public CVC-Clinic DB polyp dataset. In the study, the basic U-Net model and its derivatives (modified U-Net, modified U-Net with transfer learning (VGG-16, VGG-19) in the encoding part) were used for the segmentation process. For sensitivity analysis, models were trained on three separate datasets prepared with different preprocessing methods in addition to the raw dataset with k-fold cross validations (k = 2,3,4) and different batch numbers (1,2,3,4,5) in each cross validation. As a result of the analysis, the best performance was obtained as 0.868, 0.799, 0.873 and 0.994 for Dice, Jaccard, Sensitivity, Specificity when the batch size was taken as 1 with fourfold cross validation in the modified U-Net trained with the Discrete Wavelet Transform (DWT) dataset. This model and its parameters were then tested with public datasets Kvasir-Seg and Etis-Larib Polyp DB. Moreover, different models were trained with the parameters of the most successful model. The results of all analyzes were interpreted and compared with the literature.Article Adrenal Lesion Classification on T1-Weighted Abdomen Images With Convolutional Neural Networks(2022-12-31) Solak, Ahmet; Ceylan, Rahime; Bozkurt, Mustafa Alper; Cebeci, Hakan; Koplay, MustafaAdrenal lesions are usually discovered incidentally during other health screenings and are usually benign. However, it is vital to take precautions when a malignant adrenal lesion is detected. Especially deep learning models developed in the last ten years give successful results on medical images. In this paper, adrenal lesion characterization on T1-weighted magnetic resonance abdomen images was aimed using convolutional neural network (CNN) which is one of the deep learning methods. Firstly, effects of important model parameters are assessed on performance of CNN, so optimum CNN model is obtained for classification of adrenal lesions. For a fixed number of convolution filters determined in the first stage of the study, CNN model implemented by different kernel sizes were trained. According to the best result obtained, this time the kernel size was kept constant, and experiments were made for different filter numbers. Finally, studies were carried out with CNN structures of different depths and the results were compared. As a result of the studies, when filter is selected as [5 20], the best results in the trainings conducted with a single-block CNN structure are obtained 0.97, 0.90, 0.98, 0.90, 0.90, and 0.94, for accuracy, sensitivity, specificity, precision, F1-score, and AUC score, respectively. The study was compared with the studies in the literature, and it was seen that it was superior to them.Article Citation - WoS: 2Citation - Scopus: 1Pneumonia Detection With Chest-Caps(Int Information & Engineering Technology Assoc, 2022-12-31) Solak, Ahmet; Ceylan, RahimePneumonia is one of the diseases with the highest mortality in children. Early diagnosis is vital for the recovery of children and saving their lives. With the developments in artificial intelligence, the use of computer aided systems has become widespread. This has increased reliable, accurate and fast on studies about classification, segmentation and detection. In this study, pneumonia and healthy chest X-ray images were classified using capsule network. This model is specialized and adapted to the study in a specific way. K-fold cross validation and preprocessing of images were also applied to improve the study performance. As a result of the study, accuracy, precision, recall, F1-score and AUC scores were obtained as 0.984, 0.996, 0.971, 0.983, 0.974, respectively. The proposed model has been compared with state-of-the-art models and studies in the literature, and it is seen that our study has achieved excellent results.Article Solunum Hastalıklarının Yapay Zeka Destekli Teşhisi: COVID-19 ve Akciğer Patolojileri için Görü Dönüştürücüleri ve Resnet Mimarilerinin Değerlendirilmesi(2025-10-15) Solak, AhmetBu çalışma, gelişmiş derin öğrenme mimarilerinin – özellikle Vision Transformers (ViT) ve çeşitli ResNet modellerinin (ResNet50, ResNet101, ResNet152) – göğüs röntgenlerini Normal, Akciğer Opasitesi, Viral Pnömoni ve COVID-19 olmak üzere dört klinik açıdan önemli tanısal kategoriye sınıflandırmadaki etkinliğini sistematik olarak değerlendirmektedir. Modellerin performansını, hassasiyet, geri çağırma, F1-skora ve doğruluk gibi temel değerlendirme metrikleri üzerinden ölçmek amacıyla özenle hazırlanmış 21.165 göğüs X-ışını görüntüsünden oluşan bir veri seti kullanılmıştır. Deneysel değerlendirmeler, ViT modelinin %90.25 doğruluk, %91.56 hassasiyet, %89.22 geri çağırma ve %90.25 F1-skora elde ettiğini ortaya koymaktadır. Bu bulgular, yapay zeka temelli yaklaşımların tıbbi tanı süreçlerini güçlendirme, tanı doğruluğunu artırma ve özellikle kaynak kısıtlı ortamlarda sağlık hizmetlerinin sunumunu iyileştirme potansiyeline işaret etmektedir. Çalışma, karmaşık tıbbi görüntüleme görevlerinde Vision Transformers'ın uygulanabilirliğini vurgulamakta ve solunum yolu hastalıkları ile diğer sağlık sorunlarına yönelik yapay zeka temelli çözümleri destekleyen artan araştırma literatürüne katkıda bulunmaktadır.Article Beyin Tümörü Sınıflandırması için Görü Dönüştürücü ve Transfer Öğrenmenin Karşılaştırmalı Analizi(2025-01-30) Solak, AhmetBeyin tümörlerinin doğru sınıflandırılması, nöro-onkolojide tedavi planlarını yönlendirmek ve hasta sonuçlarını iyileştirmek için kritik öneme sahiptir. Bu çalışmada, Manyetik Rezonans (MR) görüntüleri kullanılarak Vision Transformers (ViTs) yönteminin beyin tümörlerinin ikili sınıflandırmasındaki etkinliği araştırılmış ve VGG16, VGG19 ve ResNet50 gibi CNN tabanlı modellerle karşılaştırılmıştır. Doğruluk, kesinlik, duyarlılık ve F1-skoru gibi kapsamlı değerlendirme metrikleri, ViTs’in üstün performansını ortaya koymuştur; ViTs, %92,59 doğrulukla VGG16 (%85,19), VGG19 (%74,04) ve ResNet50'yi (%88,89) geride bırakmıştır. Bu bulgular, ViTs’in nöro-onkolojide tanısal doğruluğu artıran ve hasta bakımını iyileştiren dönüştürücü bir araç olarak klinik uygulamalara entegrasyonu için umut vadeden bir yöntem olduğunu göstermektedir.