ÇAÇiğdem Gülüzar Altıntop

Çiğdem Gülüzar Altıntop

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

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Çiğdem Gülüzar Altıntop has a total of 18 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Decoding Dyslexia Through EOG Complexity: A Machine Learning Driven Analysis of Reading Behavior in Children

    by Çiğdem Gülüzar Altıntop, Ramis İlEri, Fatma Latifoğlu, Esra Demirci

    1. Dyslexia is a common learning disability that specifically affects the acquisition of reading skills despite normal intelligence, adequate educational opportunity, and the absence of sensory imp...


  • Open Access

    Automated Monkeypox Disease Classification Using Texture and Focus-Based Image Features

    by Tuğba Şentürk, Çiğdem Gülüzar Altıntop, Fatma Latifoğlu

    Accurate diagnosis of monkeypox is challenging because it is a viral disease that causes skin lesions resembling other dermatological conditions. This study aims to present a machine learning-based approach for the classification of monkeypox skin lesions using image-based feature extraction techniques. In this study, texture-based features were extracted from skin lesion images using the Gray-Level Co-Occurrence Matrix (GLCM), Gray-Level Run Length Matrix (GLRLM), Local Binary Patterns (LBP), and Focus Measures for image sharpness analysis to improve classification performance. To address class imbalance and enhance model reliability, the Synthetic Minority Oversampling Technique (SMOTE) was applied. Four different classifiers Random Forest (RF), K-Nearest Neighbors (KNN), Naive Bayes (NB) and Decision Tree (DT) were evaluated using accuracy, precision, recall, specificity, F1-score and area under the curve (AUC) metrics. The results demonstrated that the highest classification performance was achieved by the Random Forest classifier, with an accuracy of 90.02% and an AUC of 96.10%, followed by the KNN classifier. Among the extracted features, LBP and Focus Measures provided the most discriminative information for monkeypox lesion classification. Overall, feature extraction and data balancing significantly improved classification performance. The findings indicate that machine learning-based classification combined with texture feature extraction and data balancing is effective for detecting monkeypox skin lesions. RF and KNN emerged as the most suitable classifiers among those tested. To the best of our knowledge, this is the first study to investigate monkeypox lesion classification using the specified extracted features. These results highlight the potential of image-based machine learning methods in supporting early diagnosis, while future studies should focus on larger datasets and hybrid deep learning approaches to further improve accuracy and generalizability.

  • Open Access

    EEG-Based Assessment of Stress Levels Using Time–Frequency Features and Machine Learning

    by Sevde Samsa, Çiğdem Gülüzar Altıntop

    Stress from everyday living and exam times can have a major impact on cognitive function. So, identifying the degree of stress can be crucial in choosing preventive measures. In order to categorize stress levels (binary and multi-class) using machine learning techniques, the Empirical Wavelet Transform was used in this study to decompose EEG signals into subbands. Accuracy, sensitivity, specificity, precision, and F-score measurements were used to evaluate the performance of machine learning algorithms. The findings demonstrated that binary classification yielded higher accuracy than other classification tasks, especially between low and high stress levels. After feature selection, the Random Forest classifier produced the best results, with an accuracy of 86.25%. The K-Nearest Neighbor algorithm produced the best accuracy of 66.67% in multi-class classification, demonstrating the greater challenge of differentiating stress levels. The most instructive features, as determined by statistical analysis, were associated with signal power, relative energy, and amplitude, which were mostly obtained from frontal and temporal EEG channels. This study demonstrates that stress levels can be effectively classified using a limited number of EEG channels and simple features, providing a practical approach for EEG-based stress assessment.

  • Binary and Multi-Class Chest X-Ray Classification for COVID-19 and Pneumonia Detection

    by Çiğdem Gülüzar Altıntop, Tuğba Şentürk, Fatma Latifoğlu

    The Lung disorders, encompassing conditions such as pneumonia and COVID-19, represent significant challenges to global health, impacting millions annually. This study investigates the application of advanced deep learning architectures, specifically AlexNet and SqueezeNet, for the automated classification of chest X-ray (CXR) images from patients diagnosed with pneumonia, COVID-19, and healthy individuals. We achieved an impressive accuracy of 99.85% in binary classification tasks with SqueezeNet and 97.72% for multi-class classification. The results indicate that SqueezeNet outperformed AlexNet in sensitivity, specificity, and accuracy, particularly in distinguishing between COVID-19 and pneumonia. This highlights SqueezeNet's computational efficiency and effectiveness in rapid diagnostic applications. Our findings underscore the importance of timely diagnosis in improving outcomes, especially in resource-limited settings. The use of computer-aided diagnosis (CAD) technologies can aid in early detection and appropriate treatment. Future work will explore deep neural network models on heterogeneous and unbalanced datasets and apply these methods to CT images to enhance generalizability.

  • Effects of Visual Color Stimuli on Brain Activity: An EEG-Based Study

    by Çiğdem Gülüzar Altıntop, Emre Erdoğan, Tolga Demren, Ahmet Calp, Furkan Kavaklı

    Color perception is a multifaceted phenomenon due to the combined functioning of eye and brain. It has been shown that colors affect emotion and behavior. This study investigated the effects of different color stimuli on the brain using electroencephalography (EEG) signals. Twelve healthy individuals participated in the study and EEG was obtained using an 18-channel system while they were exposed to eight basic color stimuli. Features were extracted from the signals, and then several statistical analyses and topographic mapping were performed. Participants were divided into two groups and significant differences were observed between the groups, specifically in the temporal, parietal, and frontal regions. The cooler (blue, cyan) and neutral colors (black, white) demonstrated the most significant changes, while the warm colors (red, yellow, violet) produced widespread frontal-parietal activation. In the current analysis, complexity emerged as the indicator that produced the most significant differences. These findings suggest that color stimuli modulate cortical activation in a distinct way, and are subject to individual differences. Future studies will be conducted to further support these findings with larger samples, different intensities of color, and multimodal methods.

  • Open Access

    Beyond Conventional Blood Parameters: Novel Hematologic Indices for Interpretable Artificial Intelligence in Acute Myocardial Infarction

    by Çiğdem Gülüzar

    Objective: Acute myocardial infarction (AMI) is a leading cause of morbidity and mortality worldwide, emphasizing the need for timely and accurate diagnosis. This study evaluates the diagnostic potential of newly derived hematological indices—mean platelet volume–to–platelet count ratio (MPR), platelet–to–white blood cell ratio (PWR), red cell distribution width–to–lymphocyte ratio (RLR), red cell distribution width–to–platelet ratio (RPR), Systemic Immune–Inflammation Index (SII), Systemic Inflammation Response Index (SIRI), and derived neutrophil–to–lymphocyte ratio (dNLR)—combined with conventional blood parameters. Materials and Methods: An open-access dataset with 981 participants (477 AMI patients and 504 healthy controls) was analyzed. Feature selection was performed using the ReliefF algorithm to identify the most informative features. At the same time, class imbalance was addressed using the Synthetic Minority Oversampling Technique (SMOTE) for multi-class classification (control, ST-elevation myocardial infarction [STEMI], and non–ST-elevation myocardial infarction [NSTEMI]). Random Forest (RF) and Adaptive Boosting (AdaBoost) classifiers were utilized for both binary and multi-class tasks. Results: In binary classification (Control vs. AMI), RF achieved 84.92% accuracy and 91.25% area under the curve (AUC), while balanced multi-class classification reached 80.75% accuracy with RF, demonstrating consistent and robust performance across cross-validation folds. Conclusion: The findings highlight the added value of derived indices that reflect systemic inflammation, immune response, and platelet activity in enhancing the diagnosis of AMI. The proposed framework offers a reliable, interpretable, and clinically applicable approach to support early detection.


  • Open Access

    DyslexiaNet: Examining the Viability and Efficacy of Eye Movement-Based Deep Learning for Dyslexia Detection

    by Ramis İleri, Çiğdem Gülüzar Altıntop, Fatma Latifoğlu, Esra Demirci

    Dyslexia is a neurodevelopmental disorder that impairs reading, affecting 5–17.5% of children and representing the most common learning disability. Individuals with dyslexia experience decoding, reading fluency, and comprehension difficulties, hindering vocabulary development and learning. Early and accurate identification is essential for targeted interventions. Traditional diagnostic methods rely on behavioral assessments and neuropsychological tests, which can be time-consuming and subjective. Recent studies suggest that physiological signals, such as electrooculography (EOG), can provide objective insights into reading-related cognitive and visual processes. Despite this potential, there is limited research on how typeface and font characteristics influence reading performance in dyslexic children using EOG measurements. To address this gap, we investigated the most suitable typefaces for Turkish-speaking children with dyslexia by analyzing EOG signals recorded during reading tasks. We developed a novel deep learning framework, DyslexiaNet, using scalogram images from horizontal and vertical EOG channels, and compared it with AlexNet, MobileNet, and ResNet. Reading performance indicators, including reading time, blink rate, regression rate, and EOG signal energy, were evaluated across multiple typefaces and font sizes. Results showed that typeface significantly affects reading efficiency in dyslexic children. The BonvenoCF font was associated with shorter reading times, fewer regressions, and lower cognitive load. DyslexiaNet achieved the highest classification accuracy (99.96% for horizontal channels) while requiring lower computational load than other networks. These findings demonstrate that EOG-based physiological measurements combined with deep learning offer a non-invasive, objective approach for dyslexia detection and personalized typeface selection. This method can provide practical guidance for designing educational materials and support clinicians in early diagnosis and individualized intervention strategies for children with dyslexia.


  • Open Access

    Enhancing Diagnostic Accuracy of Neurological Disorders Through Feature-Driven Multi-Class Classification with Machine Learning

    by Çiğdem Gülüzar Altıntop

    Background/Objectives: Neurological disorders (ND) are a global health challenge, affecting millions and greatly reducing quality of life. Disorders such as Alzheimer’s disease, mild cognitive impairment (MCI), schizophrenia, and depression often share overlapping symptoms, complicating diagnosis and treatment. Early detection is crucial for timely intervention; however, traditional diagnostic methods rely on subjective assessments and costly imaging, which are not universally accessible. Addressing these challenges, this study investigates the classification of multiple ND using electroencephalography (EEG) signals. Methods: Various feature extraction methods were employed, and the Least Absolute Shrinkage and Selection Operator (Lasso) algorithm was utilized for effective feature selection. Two-class (disease–disease and healthy control–disease), three-class (healthy control and two ND, as well as three ND), and four-class (healthy control and three ND) classifications were conducted using different machine learning algorithms with the selected features. An EEG dataset comprising 40 Alzheimer’s patients, 43 healthy controls, 42 schizophrenia patients, 28 MCI patients, and 28 depression patients served as the experimental benchmark. Results: The Linear Discriminant Analysis (LDA) classifier achieved the highest accuracy, distinguishing between healthy controls and Alzheimer’s with 100% accuracy and demonstrating strong performance in other comparisons. Multi-class classification reached 84.67% accuracy for distinguishing depression, MCI, and schizophrenia, while four-class classification achieved 57.89%, highlighting the complexity of differentiating among multiple ND. The frequent selection of frontal lobe channels across ND indicates their critical role in classification. Conclusions: This study contributes to the literature by emphasizing disease-to-disease classification over the traditional control-versus-patient framework, highlighting the potential for more effective diagnostic tools in clinical settings.


  • Open Access

    Kalp Ritim Bozukluklarının Çok Sınıflı Sınıflandırılmasında ReliefF Yöntemi ve Makine Öğrenimi Tabanlı Yaklaşım

    by Çiğdem Gülüzar Altıntop

    Kardiyovasküler hastalıklar (KVH), dünya genelinde ciddi sağlık tehditleri arasında yer almakta ve tüm ölümlerin yaklaşık %32'sine neden olmaktadır. Bu nedenle, KVH'ların erken tanısı ve uygun tedaviye başlanması hayati önem taşımaktadır. Elektrokardiyografi (EKG), kalbin elektriksel aktivitesinin kaydedilmesiyle elde edilen önemli bir tanı yöntemidir. Ancak, aritmi gibi kalp rahatsızlıklarının tanısı, uzman klinisyenlerin gözle incelemesine dayanmakta ve bu süreç zaman alıcı ve zahmetli olabilmektedir. Bu çalışmada, EKG sinyallerinden otomatik aritmi tespiti için bir bilgisayar destekli sistem geliştirilmesi amaçlanmıştır. Normal sinüs ritmi, pacemaker ritmi ve 15 farklı aritmi olmak üzere toplam 17 farklı kardiyak aktivite; EKG sinyallerinden çok çeşitli öznitelik çıkarımı, ReliefF kullanılarak öznitelik seçimi ve farklı makine öğrenimi algoritmalarını kullanılması ile sınıflandırılmıştır. Elde edilen sonuçlar, K-En Yakın Komşu ve Rastgele Orman algoritmalarının %93,4 ve %99,23 doğruluk oranları ile en yüksek başarıyı gösterdiğini ortaya koymuştur. Bu çalışma, aritmilerin çok sınıflı sınıflandırmasında EKG sinyallerinden morfolojik, zaman, frekans, entropi ve karmaşıklık özniteliklerini bir arada çıkararak, farklı makine öğrenimi algoritmaları kullanarak 17 farklı kardiyak aktivite sınıfını yüksek doğruluk oranlarıyla sınıflandırmıştır. Böylece, literatürdeki çalışmalardan farklılaşarak aritminin otomatik sınıflandırılmasına önemli bir katkı sağlamıştır.

  • Open Access

    Histopathological Image Analysis Using Machine Learning to Evaluate Cisplatin and Exosome Effects on Ovarian Tissue in Cancer Patients

    by Tuğba Şentürk, Fatma Latifoğlu, Çiğdem Gülüzar Altıntop, Arzu Yay, Zeynep Burçin Gönen, Gözde Özge Önder, Özge Cengiz Mat, Yusuf Özkul

    Cisplatin, a widely used chemotherapeutic agent, is highly effective in treating various cancers, including ovarian and lung cancers, but it often causes ovarian tissue damage and impairs reproductive health. Exosomes derived from mesenchymal stem cells are believed to possess reparative effects on such damage, as suggested by previous studies. This study aims to evaluate the reparative effects of cisplatin and exosome treatments on ovarian tissue damage through the analysis of histopathological images and machine learning (ML)-based classification techniques. Five experimental groups were examined: Control, cisplatin-treated (Cis), exosome-treated (Exo), exosome-before-cisplatin (ExoCis), and cisplatin-before-exosome (CisExo). A set of 177 Local Binary Pattern (LBP) features were extracted from histopathological images, followed by feature selection using Lasso regression. Classification was performed using ML algorithms, including decision tree (DT), k-nearest neighbors (KNN), support vector machine (SVM), and Artificial Neural Network (ANN). The CisExo group exhibited the most homogeneous texture, suggesting effective tissue recovery, whereas the ExoCis group demonstrated greater heterogeneity, possibly indicating incomplete recovery. KNN and ANN classifiers achieved the highest accuracy, particularly in comparisons between the Control and CisExo groups, reaching an accuracy of 87%. The highest classification accuracy was observed for the Control vs. Cis groups (approximately 91%), reflecting distinct features, whereas the Control vs. Exo groups demonstrated lower accuracy (around 68%) due to feature similarity. Exosome treatments, particularly when administered post-cisplatin, significantly improve ovarian tissue recovery. This study highlights the potential of ML-based classification as a robust tool for evaluating therapeutic outcomes. Additionally, it underscores the promise of exosome therapy in mitigating chemotherapy-induced ovarian damage and preserving reproductive health. Further research is warranted to validate these findings and optimize treatment protocols.