Özet:

An abnormal proliferation of human cells due to excessive division is called a tumor. Tumors, which can form in many parts of the body, have a degree of danger according to where they occur. The brain is one of the most dangerous areas of tumor formation. Intense studies have been carried out in recent years for the detection of tumors in the brain region. Artificial intelligence-based methods are at the forefront of these studies. Convolutional neural networks (CNN), a deep learning method, are used for classification, feature extraction and transfer learning purposes. In this study, CNN method was used for feature extraction from brain MR images. In this context, DarkNet53 model, one of the pre-trained CNN models, was selected for feature extraction. The feature extractor layers of the DarkNet53 model are conv52, res23, avg1, and conv53, respectively. After feature extraction, feature selection process was applied. Relief and Chi-Square Test methods were chosen as feature-selective methods. After feature extraction, the support vector machine algorithm, which is one of the classical machine learning methods, was determined as the classifier method. The proposed method has been tested on the “Brain MRI Images for Brain Tumor Detection” dataset. According to the experimental results, the best result was obtained with the proposed method in which the res23 layer was determined as feature extractor and the Chi-Square Test method as feature selective.

Anahtar Kelimeler: