OKOnur Karasoy

Onur Karasoy

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Onur Karasoy has a total of 13 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Automated periodontal assessment in orthodontic patients: a dual CNN framework

    by Ebru Yurdakurban, Ali Batuhan Bayırlı, Mehmetcan Uytun, Öznur Suçeken, Onur Karasoy, Osman Özkaraca, Kübra Gülnur Topsakal

    Objective: The aim of this study was to develop convolutional neural network (CNN)-based systems to diagnose calculus, plaque, gingival hyperplasia and gingival inflammation in intraoral images from orthodontic patients. Materials and methods: A dataset of 1,000 lateral and frontal intraoral images from orthodontic patients was used to develop CNN-based models. Periodontology specialists annotated areas of dental calculus, plaque, gingival inflammation, and gingival hyperplasia on the teeth and gingiva. The dataset was divided into training (80%), validation (10%), and test (10%) sets for model development. The YOLOv8 and hybrid U-Net + ResNet50 models were examined. Their performance was evaluated on the basis of accuracy, precision, recall, F 1 score, Tversky loss, intersection over union, mean average precision, Dice coefficient, and Cohen’s kappa. Results: The mean classification accuracy was 0.96 for the YOLOv8 model and 0.93 for the U-Net + ResNet50 model. On the basis of the Dice coefficient, the models performed best in detecting gingival hyperplasia (YOLOv8: 0.78, U-Net + ResNet50: 0.79) and worst in detecting dental calculus (YOLOv8:0.48, U-Net + ResNet50:0.53). Cohen’s kappa coefficient was highest for classifying gingival hyperplasia (YOLOv8: 0.785, U-Net + ResNet50: 0.790). The precision exceeded 0.72 across all the classifications, with the greatest precision in classifying gingival inflammation. Conclusion: Deep learning-based systems can serve as decision support tools by offering rapid and objective evaluations of dental calculus, plaque, gingival inflammation, and gingival hyperplasia. Nonetheless, the definitive diagnostic conclusion should be based on the clinician’s specialized expertise and professional judgment. Clinical relevance: The integration of CNN-based diagnostic models into clinical workflows has the potential to facilitate early periodontal diagnosis and improve accessibility to periodontal assessments in orthodontic patients.


  • Open Access

    CompCorona: A web application for comparative transcriptome analyses ofcoronaviruses reveals SARS-CoV-2-specific host response

    by Rana Sali̇hOğlu, Fati̇h Saraçoğlu, Mustafa Si̇bAi̇, Tali̇p Zengi̇n, Başak Abak Masud, Onur Karasoy, Tuğba Önal Süzek

    Background/aim: Understanding the mechanism of host transcriptomic response to infection by the SARS-CoV-2 virus is crucial, especially for patients suffering from long-term effects of COVID-19, such as long COVID or pericarditis inflammation, potentially linked to side effects of the SARS-CoV-2 spike proteins. We conducted comprehensive transcriptome and enrichment analyses on lung and peripheral blood mononuclear cells (PBMCs) infected with SARS-CoV-2, as well as on SARS-CoV and MERS-CoV, to uncover shared pathways and elucidate their common disease progression and viral replication mechanisms. Materials and methods: We developed CompCorona, the first interactive online tool for visualizing gene response variance among the family Coronaviridae through 2D and 3D principal component analysis (PCA) and exploring systems biology variance using pathway plots. We also made preprocessed datasets of lungs and PBMCs infected by SARS-CoV-2, SARS-CoV, and MERS-CoV publicly available through CompCorona. Results: One remarkable finding from the lung and PBMC datasets for infections by SARS-CoV-2, but not infections by other coronaviruses (CoVs), was the significant downregulation of the angiogenin (ANG) and vascular endothelial growth factor A (VEGFA) genes, both directly involved in epithelial and vascular endothelial cell dysfunction. Suppression of the TNF signaling pathway was also observed in cells infected by SARS-CoV-2, along with simultaneous activation of complement and coagulation cascades and pertussis pathways. The ribosome pathway was found to be universally suppressed across all three viruses. The CompCorona online tool enabled the comparative analysis of 9 preprocessed host transcriptome datasets of cells infected by CoVs, revealing the specific host response differences in cases of SARS-CoV-2 infection. This included identifying markers of epithelial dysfunction via interactive 2D and 3D PCA, Venn diagrams, and pathway plots. Conclusion: Our findings suggest that infection by SARS-CoV-2 might induce pulmonary epithelial dysfunction, a phenomenon not observed in cells infected by other CoVs. The publicly available CompCorona tool, along with the preprocessed datasets of cells infected by various CoVs, constitutes a valuable resource for further research into CoV-associated syndromes.


  • Open Access

    Development of content‐based SMS classification application by using Word2Vec‐based feature extraction

    by Serkan Ballı, Onur Karasoy

    While mobile instant messaging applications such as WhatsApp, Messenger, Viber offer benefits to phone users such as price, easy usage, stable, collective and direct communication, SMS (short message service) is still considered a more reliable privacy‐preserving technology for mobile communication. This situation directs the institutions that want to perform the product promotion such as advertising, informing, promotion etc. to use SMS. However, spam messages sent from unknown sources constitute a serious problem for SMS recipients. In this study, a content‐based classification model which uses the machine learning to filter out unwanted messages is proposed. From the selected dataset, the model to be used in the classification is created with the help of Word2Vec word embedding tool. Thanks to this model, two new features are revealed for calculating the distances of messages to spam and ham words. The performances of the classification algorithms are compared by taking these two new features into consideration. The random forest method succeeded with a correct accuracy rate of 99.64%. In comparison to other studies using the same dataset, more successful correct classification percentage is achieved.