GCRIS

Now showing 1 - 2 of 2

Tree-net: Biyomedikal Görüntü Segmentasyonu için Tree-net: Darboğaz Özellik Süpervizyonu Kullanılan Yapay Sinir Ağı Modeli
(Abdullah Gül Üniversitesi, Fen Bilimleri Enstitüsü, 2024) Demirci, Orhan; Yılmaz, Bülent
In this thesis, we introduce Tree-NET, a novel approach for medical image segmentation utilizing bottleneck feature supervision. This method enhances traditional segmentation algorithms by keeping supervision between bottleneck features of the network. The primary goal is to improve the model's ability to learn discriminative and robust features while simultaneously reducing computational costs. Bottleneck feature supervision involves compressing the input and label data using Autoencoders and then supervising the bottleneck features with a segmentation network named 'Bridge-Net,' which can be any segmentation model of choice. We applied Tree-NET to two critical medical image segmentation tasks: skin lesion segmentation and polyp segmentation. Our experiments demonstrate significant improvements in segmentation accuracy and efficiency. For instance, the U-NET backboned Tree-NET uses only 154.43 MB for executing and storing the model, which is almost 3.5 times smaller than the original U-Net while having a close number of trainable parameters. In skin lesion segmentation, Tree-NET achieved dice, Intersection-over-Union (IoU), and accuracy scores of 0.893, 0.751, and 0.977 respectively. For polyp segmentation, the scores were 0.856, 0.795, and 0.923 for dice, IoU, and accuracy respectively. Compared to traditional segmentation models, the empirical results show that Tree-NET achieves higher accuracy with reduced training time and computational cost, thus representing a significant advancement in medical image analysis by providing more reliable and efficient tools for clinical applications.
Derin Öğrenme Yöntemleri Kullanarak Dermatoskopik Görüntülerden Otomatik Cilt Kanseri Tespiti ve Sınıflandırılması
(Abdullah Gül Üniversitesi / Fen Bilimleri Enstitüsü, 2023) Kalaycı, Serdar; Yılmaz, Bülent
Early detection of skin cancer is crucial for successful treatment and improved patient outcomes. The most prevalent form of cancer is skin cancer and if left undetected, it can spread and become more difficult to treat. A dangerous and frequently fatal type of skin cancer is melanoma. Regular skin examinations and self-examinations can help identify suspicious moles or lesions, which can then be evaluated by a dermatologist. In addition, advances in technology and artificial intelligence have enabled the development of tools for automated skin cancer screening, providing a convenient and efficient means of early detection. This can lead to more efficient diagnosis, reduced healthcare costs and improved patient care. By evaluating skin lesions from images, deep learning techniques have shown considerable potential in increasing the precision of melanoma detection. By using large datasets and complex neural networks, deep learning algorithms can effectively distinguish between benign and malignant skin lesions with high accuracy. Ensemble of CNN models helps improve the performance and reliability of the classification task. By combining the predictions of multiple CNN models lead to more accurate and robust predictions. In this thesis, for melanoma classification problem, many different data augmentations techniques applied and different convolutional neural networks architectures evaluated, applied vignetting effect filter and hair noise in accordance with the dataset and results of ensemble of the best CNN models are promising. This thesis attempts to produce a reliable model for the classification of melanoma by conducting experiments on two combined publically accessible data sets, ISIC 2019 and ISIC 2020. On the testing sets in our studies, the proposed solution attained 95.75% AUC.

Yüksek Lisans Tezleri

Browse

Filters

Settings

Sort By

Results per page

Search Results