Prediction of Brain Tumor Evolutionary Process from Segmented MR Images Using YOLOv8
AbstractGlioma is one of the most common and aggressive types of brain tumors. This study aims to predict the glioma type and its evolutionary process using the YOLOv8 deep learning model based on segmented MR images. The segmented MR images obtained from the BraTS dataset were processed into 2D slices using a custom algorithm. The model was trained using transfer learning, and the Adam optimizer was employed for optimization. The model's performance was evaluated using YOLOv8's standard metrics, including mAP, IoU, Precision, and Recall. The results demonstrate that the YOLOv8 model trained on 2D data derived from segmented images achieved 98.5% accuracy, 98.5% F1 score, and 88% sensitivity, effectively classifying glioma types and reliably predicting the evolutionary process.