Browsing by Author "Aksakalli, Isil Karabey"

Now showing 1 - 10 of 10

An Attention-Based Transfer Learning Model for Diagnosing Subluxation in Temporomandibular Joint Panoramic Radiographs
(Nature Portfolio, 2025) Sancar, Yasin; Aksakalli, Isil Karabey; Sokmen, Ozlem; Barin, Hande; Torenek-Agirman, Kubra
Artificial intelligence (AI) and deep learning (DL) techniques have great potential to accelerate diagnostic processes, increase accuracy, and support clinical decision-making in healthcare. In this study, we propose a transfer learning-based approach-one of the DL techniques-to improve subluxation (SL) detection in temporomandibular joint panoramic radiography (TMJ-PR) images. For this purpose, we prepared and publicly released a dataset comprising 3,425 annotated TMJ-PR images to encourage reproducibility and further research in this domain. Several transfer learning models including MobileNet, ResNet50V2, InceptionV3, Xception, EfficientNetV2B0, InceptionResNetV2, and DenseNet201 were trained and evaluated using a 5-fold cross-validation method. By integrating a self-attention mechanism into the DenseNet201 model which achieved the highest baseline performance across all metrics, the proposed attention-based version yielded further improvements, achieving an accuracy of 90.7%, precision of 90.7%, recall of 90.7%, specificity of 89.4%, and F1-score of 90.7%. The results indicate that the proposed model achieves superior F1-score performance compared to all baseline models with relative improvements ranging from +2.40% (vs. DenseNet201) to +14.26% (vs. EfficientNetV2B0). The findings demonstrate that the proposed model not only improves subluxation detection performance but also offers a promising foundation for integration into a clinical decision support system (CDSS), enhancing early diagnosis and treatment planning using low-cost TMJ-PR images. The publicly shared dataset further supports transparency and reproducibility in future medical AI research.
Comprehensive Insights Into Artificial Intelligence for Dental Lesion Detection: A Systematic Review
(MDPI, 2024) Demir, Kubra; Sokmen, Ozlem; Aksakalli, Isil Karabey; Torenek-Agirman, Kubra
Background/Objectives: The growing demand for artificial intelligence (AI) in healthcare is driven by the need for more robust and automated diagnostic systems. These methods not only provide accurate diagnoses but also promise to enhance operational efficiency and optimize resource utilization in clinical workflows. In the field of dental lesion detection, the application of deep learning models to various imaging techniques has gained significant prominence. This study presents a comprehensive systematic review of the utilization of deep learning methods for detecting dental lesions across different imaging modalities, including panoramic imaging, periapical radiographs, and cone-beam computed tomography (CBCT). A systematic search was conducted following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to ensure a structured and transparent review process. Methods: This study addresses four key research questions related to the types of objects used for AI in dental images, state-of-the-art approaches for detecting lesions in dental images, data augmentation methods, and challenges and possible solutions to the existing AI-based dental lesion detection. Furthermore, this systematic review was performed on 29 primary studies identified from multiple electronic databases. This review focused on studies published between 2019 and 2024, sourced from IEEE, Web of Knowledge, Springer, ScienceDirect, PubMed, and Google Scholar. Results: We identified five types of lesions in dental images as periapical lesions, cyst lesions, jawbone lesions, dental caries, and apical lesions. Among the fourteen state-of-the-art deep learning approaches, the results demonstrate that deep learning models, such as U-Net, AlexNet, and You Only Look Once (YOLO) version 8 (YOLOv8) are commonly employed for dental lesion detection. These deep learning models have the potential to serve as integral components of decision-making processes by improving detection accuracy and supporting clinical workflows. Furthermore, we found that among twelve types of data augmentation techniques, flipping, rotation, and reflection methods played an important role in increasing the diversity of the datasets. We also identified six challenges for dental lesion detection, and the main issues were identified as data integration, poor data quality, limited model generalization, and overfitting. Proposed solutions against the aforementioned challenges include the integration of larger datasets, model optimization, and diversification of data sources. Conclusions: This study provides a comprehensive overview of current methodologies and potential advancements in dental lesion detection using deep learning. The findings indicate that possible solutions against the challenges of AI-based diagnostic methods in dental lesion detection need to be more generalizable regardless of image type, the number of data, and data quality.
Deployment and Communication Patterns in Microservice Architectures: A Systematic Literature Review
(Elsevier Science Inc, 2021) Aksakalli, Isil Karabey; Celik, Turgay; Can, Ahmet Burak; Tekinerdogan, Bedir
Context: Microservice is an architectural style that separates large systems into small functional units to provide better modularity. A key challenge of microservice architecture design frequently discussed in the literature is the identification and decomposition of the service modules. Besides this, two other key challenges can be identified, including the deployment of the modules on the corresponding execution platform, and adopted communication patterns. Objective: This study aims to identify and describe the reported deployment approaches, and the communication platforms for microservices in the current literature. Furthermore, we aim to describe the identified obstacles of these approaches as well as the corresponding research directions. Method: A systematic literature review (SLR) is conducted using a multiphase study selection process in which we reviewed a total of 239 papers. Among these, we selected 38 of them as primary studies related to the described research questions. Results: Based on our study, we could identify three types of deployment approaches and seven different communication patterns. Moreover, we have identified eight challenges related to the deployment and six challenges related to the communication concerns. Conclusion: Our study shows that in addition to the identification of modules, the deployment and communication approaches are equally crucial for a successful application of the microservice architecture style. Various deployment approaches and communication patterns appear to be useful for different contexts. The identified research directions in the literature study show that still more research is needed to cope with the current challenges. (C) 2021 Elsevier Inc. All rights reserved.
Enhancing Indoor Localization with Room-To Transition Time: A Multi-Dataset Study
(MDPI, 2025) Aksakalli, Isil Karabey; Bayindir, Levent
With the rapid advancement of network technologies and the widespread adoption of smart devices, the demand for efficient indoor localization and navigation systems has surged. Addressing the navigation challenge without requiring additional hardware is critical for the broad adoption of such technologies. Among various fingerprint-based systems-such as Bluetooth, ZigBee, or FM radio-Wi-Fi-based indoor positioning stands out as a practical solution, due to the pervasiveness of Wi-Fi infrastructure in public indoor spaces. This study introduces an ESP32-based data-collection tool designed to minimize offline training time for Wi-Fi fingerprinting, and it presents a novel dataset incorporating room-to-room transition time, which represents the time taken to move between rooms, alongside Wi-Fi signal strength data. The proposed approach focuses on room-level localization, leveraging Machine Learning (ML) models to predict the most likely room rather than precise (x, y) coordinates. To assess the effectiveness of this feature, three datasets were collected from different residential environments by three different individuals, enabling a comprehensive evaluation across multiple spatial layouts and movement patterns. The experimental results demonstrate that incorporating room-to-room transition time consistently enhanced localization performance across all the datasets, with accuracy improvements ranging from 1.17% to 12.47%, depending on the model and dataset. Notably, the Wide Neural Network model exhibited the highest improvement, achieving an accuracy increase from 82.37% to 94.77%, while the Ensemble-based methods such as Ensemble Bagged Trees also benefited significantly, reaching up to 93.17% accuracy. Despite varying gains across the datasets, the results confirm that integrating room-to-room transition time improves Wi-Fi-based indoor positioning by leveraging temporal movement patterns to enhance classification.
Exploring the Effect of Image Enhancement Techniques with Deep Neural Networks on Direct Urinary System X-Ray (DUSX) Images for Automated Kidney Stone Detection
(Wiley, 2023) Kilic, Ugur; Aksakalli, Isil Karabey; Ozyer, Gulsah Tumuklu; Aksakalli, Tugay; Ozyer, Baris; Adanur, Senol
In diagnosing kidney stone disease, clinical specialists often apply medical imaging techniques such as CT, X-Ray and US. Among these imaging techniques, X-Ray is frequently chosen as the primary examination method in emergency services due to its low cost, accessibility, and low radiation levels. However, interpreting the X-Ray images by inexperienced specialists can be challenging due to the low image quality and the presence of noise. In this study, we propose a computer-aided diagnosis system based on deep neural networks to assist clinical specialists in detecting kidney stones using Direct Urinary System X-Ray (DUSX) images. Firstly, in consultation with clinical specialists, we created a new dataset composed of 630 DUSX images and presented it publicly. We also defined preprocessing steps that incorporate image enhancement techniques such as GF, LoG, BF, HE, CLAHE, and CBC to enable deep neural networks to perceive the images more clearly. With these techniques, we considered the noise reduction in the DUSX images and enhanced the poor quality, especially in terms of contrast. For each preprocessing step, we created models to detect kidney stones using YOLOv4 and Mask R-CNN architectures, which are common CNN-based object detectors. We examined the effects of the preprocessing steps on these models. To the best of our knowledge, the combination of BF and CLAHE which is called CBC in this study, has not been applied before in the literature to enhance DUSX images. In addition, this study is the first in its field in which the YOLOv4 and Mask R-CNN architectures have been used for the detection of kidney stones. The experimental results demonstrated the most accurate method is the YOLOv4 model, which includes the CBC preprocessing step, as the result model. This model shows that the accuracy rate, precision, recall, and F1-score were found as 96.1%, 99.3% 96.5%, and 97.9% respectively in the test set. According to these performance metrics, we expect that the proposed model will help to reduce the unnecessary radiation exposure and associated medical costs that come with CT scans.
A Hybrid Patchnet-Attention Based Deep Learning Architecture for Multi-Type Fabric Defect Classification in Textile Manufacturing and Quality Control
(Elsevier - Division Reed Elsevier India Pvt Ltd, 2025) Aksakalli, Isil Karabey; Demir, Kubra; Sokmen, Ozlem
Accurate and timely detection of defects that may occur on fabric surfaces is a critical requirement for ensuring sustainable production quality in the textile industry. Due to human resource, time, and cost limitations, there is a growing interest in advanced image processing and deep learning-based automatic defect detection systems to improve the accuracy and efficiency of quality control in fabric manufacturing processes. In this study, we propose a novel hybrid PatchNet-Attention architecture that integrates patch-based feature extraction with an attention mechanism to improve defect localization and recognition. To evaluate the generalizability of the proposed architecture, its performance was tested on three public datasets using different class structures. Specifically, four classification scenarios were conducted: (i) classification with baseline models, (ii) patch-based classification, (iii) classification with a Convolutional Block Attention Module (CBAM)-enhanced model, and (iv) the proposed hybrid PatchNet-Attention architecture. Initially, 15 pre-trained Convolutional Neural Network (CNN) architectures were evaluated using transfer learning on the ZD001 dataset. The best-performing models, ResNet101V2 and Xception, were then selected as the foundation for constructing the hybrid PatchNet-Attention model. The experimental results demonstrate that configurations incorporating the attention mechanism consistently achieve the highest performance across all evaluated datasets. Specifically, the hybrid PatchNet-Attention model attained superior outcomes on the ZD001 dataset, with an F1-score of 99.15% and a Receiver Operating Characteristic-Area Under the Curve (ROC-AUC) of 99.5% in the three-class setting, and an F1-score of 97.28% with a ROC-AUC of 99.74% in the nine-subclass configuration. In the TILDA data set, the proposed model produced an F1 score of 87.74% and an ROC-AUC of 98.09%, while in the FDD data set it achieved an F1 score of 98.95% and a ROC-AUC of 99.50%. The source code of the proposed method can be accessed from the Data Availability section.
Micro-Ide: A Tool Platform for Generating Efficient Deployment Alternatives Based on Microservices
(Wiley, 2022) Aksakalli, Isil Karabey; Celik, Turgay; Can, Ahmet Burak; Tekinerdogan, Bedir
Microservice architecture (MSA) is a paradigm to design and develop scalable distributed applications using loosely coupled, highly cohesive components that can be deployed independently. The applications that realize the MSA may contain thousands of services that together form the overall system. Microservices interact with each other by producing and consuming data. Deploying frequently communicating services to the same physical resource would reduce network utilization, which is vital for reducing costs and improving scalability. Since the physical resources have limited capacity, it is not always possible to deploy communicating services to the same resource. Therefore, automated efficient deployment alternatives need to be generated for MSA in the design phase. To address this problem, we proposed an algorithmic approach to generate efficient microservice deployment configurations to available cloud resources in our previous study. In this study, a tool (Micro-IDE) has been proposed to realize and evaluate this approach. The Micro-IDE tool has been validated using a case study inspired by the Spotify application.
A Model-Driven Architecture for Automated Deployment of Microservices
(MDPI, 2021) Aksakalli, Isil Karabey; Celik, Turgay; Can, Ahmet Burak; Tekinerdogan, Bedir
Microservice architecture consists of a collection of loosely coupled, self-contained services that can be deployed independently. Given the limited capacity of the resources for a large number of services, the deployment of the services does not scale well and leads to operational complexity and runtime overhead. This paper proposes a model-driven approach for the automated deployment of microservices to minimize the execution cost and communication costs among the microservices. The identification of the feasible deployment is defined at the architecture design level based on the provided capacity of the resources and the collection of microservices. The corresponding tool support using Eclipse Modeling Environment is described, and a case study on book shopping is used to illustrate the approach.
Prediction of Spontaneous Distal Ureteral Stone Passage Using Artificial Intelligence
(Springer, 2024) Aksakalli, Tugay; Aksakalli, Isil Karabey; Cinislioglu, Ahmet Emre; Utlu, Adem; Demirdogen, Saban Oguz; Celik, Feyzullah; Karabulut, Ibrahim
Purpose Identifying factors predicting the spontaneous passage of distal ureteral stones and evaluating the effectiveness of artificial intelligence in prediction. Materials and methods The files of patients presenting with distal ureteral stones were retrospectively evaluated. Those who experienced spontaneous passage were assigned to Group P, while those who did not were assigned to Group N. Demographic and clinical data of both groups were compared. Then, logistic regression analysis was performed to determine the factors predicting spontaneous stone passage. Based on these factors, a logistic regression model was prepared, and artificial intelligence algorithms trained on the dataset were compared with this model to evaluate the effectiveness of artificial intelligence in predicting spontaneous stone passage. Results When comparing stone characteristics and NCCT findings, it was found that the stone size was significantly smaller in Group P (4.9 +/- 1.7 mm vs. 6.8 +/- 1.4 mm), while the ureteral diameter was significantly higher in Group P (3.3 +/- 0.9 mm vs. 3.1 +/- 1.1 mm) (p < 0.05). Parameters such as stone HU, stone radiopacity, renal pelvis AP diameter, and perirenal stranding were similar between the groups. In multivariate analysis, stone size and alpha-blocker usage were significant factors in predicting spontaneous stone passage. The ROC analysis for the logistic regression model constructed from the significant variables revealed an area under the curve (AUC) of 0.835, with sensitivity of 80.1% and specificity of 68.4%. AI algorithms predicted the spontaneous stone passage up to 92% sensitivity and up to 86% specifity. Conclusions AI algorithms are high-powered alternatives that can be used in the prediction of spontaneous distal ureteral stone passage.
Systematic Approach for Generation of Feasible Deployment Alternatives for Microservices
(IEEE-Inst Electrical Electronics Engineers Inc, 2021) Aksakalli, Isil Karabey; Celik, Turgay; Can, Ahmet Burak; Tekinerdogan, Bedir
Microservice architectures rely on the development of modular and independent software units, which typically address a single task and communicate with other microservices via well-defined interfaces. This has several benefits such as easier maintenance and update of services. However, deploying a microservice-based application is often more complicated than a monolithic application. While a monolithic application can be deployed one at a time on a group of similar servers behind a load balancer, a microservice-based application consists of different microservices and each microservice usually has more than one runtime instance that needs to be configured and deployed. For a small number of microservices and applications, the deployment could be done manually. However, a large number of microservices are frequently observed in practice. In such cases, the deployment becomes cumbersome and error-prone and does not scale with the increased number of services. To cope with this problem, we present a systematic approach and the corresponding tool support for enabling the deployment of microservices to resources that have limited capacity. Hereby, we model and define the design space given the deployment parameters and automatically derive the feasible deployment solution. The approach is validated using a taxi-hailing system case study inspired by Uber which has spread all over the world in recent years.