Electrical - Electronic Engineering / Elektrik - Elektronik Mühendisliği

Permanent URI for this collectionhttps://hdl.handle.net/11147/11

Browse

Filters

2023 - 20242

Settings

Access Right: Closed AccessAuthor: search.filters.author.Düzyel, Okan

Search Results

Now showing 1 - 2 of 2
  • Article
    Citation - WoS: 27
    Citation - Scopus: 34
    A New Method for Gan-Based Data Augmentation for Classes With Distinct Clusters
    (Pergamon-Elsevier Science Ltd, 2024) Kuntalp, Mehmet; Düzyel, Okan
    Data augmentation is a commonly used approach for addressing the issue of limited data availability in machine learning. There are various methods available, including classical and modern techniques. However, when applying modern data augmentation methods, such as Generative Adversarial Neural Networks (GANs), to a class specific data, the resulting data can exhibit structural discrepancies. This study explores a different use of GANs as a data augmentation method that solves this problem using the electrocardiogram (ECG) signals in the MITBIH arrhythmia dataset as the example. We begin by examining the cluster structure of a specific class using tDistributed Stochastic Neighbor (t-SNE) method. Based on this cluster structure, we propose a new method for applying GANs to augment data for that class. We assess the effect of our method in a classification task using 1-D Convolutional Neural Network (CNN), Support Vector Machine (SVM), One vs one classifier (Ovo), K-Nearest Neighbors (KNN), and Random Forest as the classifiers. The results demonstrate that our proposed method could lead to better classification performance if a specific class has distinct clusters when compared to normal use of GANs.
  • Article
    Citation - WoS: 5
    Citation - Scopus: 7
    Adaptive Resizer-Based Transfer Learning Framework for the Diagnosis of Breast Cancer Using Histopathology Images
    (Springer, 2023) Düzyel, Okan; Çatal, Mehmet Sergen; Kayan, Ceyhun Efe; Sevinç, Arda; Gümüş, Abdurrahman
    Breast cancer is a major global health concern, and early and accurate diagnosis is crucial for effective treatment. Recent advancements in computer-assisted prediction models have facilitated diagnosis and prognosis using high-resolution histopathology images, which provide detailed information on cancerous tissue. However, these high-resolution images often require resizing, leading to potential data loss. In this study, we demonstrate the effect of a learnable adaptive resizer for breast cancer classification using the BreakHis dataset. Our approach incorporates the adaptive resizer with various convolutional neural network models, including VGG16, VGG19, MobileNetV2, InceptionResnetV2, DenseNet121, DenseNet201, and EfficientNetB0. Despite producing visually less appealing images, the learnable resizer effectively improves classification performance. DenseNet201, when jointly trained with the adaptive resizer, achieves the highest accuracy of 98.96% for input images of 448x448 resolution. Our experimental results demonstrate that the adaptive resizer performs better at a magnification factor of 40x compared to higher magnifications. While its effectiveness becomes less pronounced as image resolution increases to 100x, 200x, and 400x, the adaptive resizer still outperforms bilinear interpolation. In conclusion, this study highlights the potential of adaptive resizers in enhancing performance for medical image classification. By outperforming traditional image resizing methods, our work contributes to the advancement of deep neural networks in the field of breast cancer diagnostics.