Effects of Some Deep Learning Models with Power Law Transformation on Brain Tumor Classification

Brain tumors are benign or malignant tumors that occur as a result of abnormal growth of various cell groups within the brain. These tumors are often a serious threat to human health and are widely prevalent worldwide. Magnetic Resonance Imaging (MRI) is the most common technique in medical imaging for the detection of brain tumors. MRI enables detailed examination of the complex structure of the brain and facilitates the detection of tumors. However, due to the complexity of brain tissue, manual diagnoses may often be inaccurate and adversely affect patients’ treatment processes. Therefore, technologies such as image processing and deep learning models have become increasingly popular in recent years. In this study, Power Law transformation is used to enhance Magnetic Resonance (MR) images. The transformed images are then trained with deep learning models. The four different deep learning models used in this study are VGG16, VGG19, ResNet50, and MobileNet. The aim of the study is to determine the best-performing deep learning model by comparing the results obtained from untreated and treated MR images. The results of all the training sessions showed that the Power Law transformation had a positive impact on all deep learning models, and the pair with the highest performance was determined to be Power Law + MobileNet.