Segmentation of brain tumors with image processing in artificial neural networks and an application

Brain tumors are the general name for abnormal cell and mass growth in the skull. In order to diagnose a brain tumor, the most common examination is an MR (magnetic resonance) image that shows foreign masses in the brain tissue and tissue. After the diagnosis is made, one should quickly plan a course of treatment. After the MR images are taken, it may take time for the images to be examined and reported by expert radiologists. In recent years, thanks to the rapidly developing deep learning technologies and innovations in the field of medicine, various studies are being conducted to diagnose diseases early and accurately. Minimizing human-caused errors has an important place in these studies. In this study, a new convolutional neural network model has been trained by using artificial intelligence techniques to help experts by marking MR images. At the training stage, 80% of the BRAST dataset was used by using the U-Net model. 20% of the samples in the dataset were used to evaluate the performance of the model. When the findings obtained as a result of the training and testing procedures are examined, it is seen that the trained model has been trained as a model that can successfully mark the entire tumor, tumor nucleus and expanding tumor sites with a similarity ratio of 0.908, 0.807 and 0.877 (BO, Dice Coefficient Score), respectively.