Özet:

Human deaths from brain tumors are increasing today. As the brain tumor grows very quickly, it can double its normal size. Therefore, experts should quickly perform the process of examining magnetic resonance (MR) images. This step is crucial in the diagnosis of cancer, in the planning of treatment and in the evaluation of the result of treatment. If the patient with a tumor in the brain is not properly and quickly treated, the patient's chances of survival are reduced and this situation can lead to the patient's death. In this article, a computer-backed automatic tumor detection system developed from brain MR images that can easily detect the tumor and indicate the place of the tumor, which can help experts. The developed system is based on the Regional-based Evolutionary Neural Networks (BESA), a deep learning architecture. BESA is a structure that uses the Evolutionary Neural Networks (ESA) architecture, but it can be considered a structure where the area concerned is also given as an entrance in addition to the entrance image. In the proposed method, different BESA architectures were designed and tested on Benchmark, Rembredant and Harvard data sets. The highest accuracy gained is 99.10% from the BESA4 architecture and Benchmark data set. The highest average accuracy is estimated at 98.66% with the BESA4 architecture. In addition, the success of the recommended method is compared with some methods in literature. The comparisons also showed that the proposed method was more successful.

Anahtar Kelimeler: