Brain Tumor Detection Using Fine-Tuned YOLO Model with Transfer Learning

The main motive of this research is to propose a transfer learning (TL) with fine-tuning for a deep learning model to detect brain tumor (BT) from MRI scans dataset. The brain tumor has obtrusive properties leading to a high demise rate but it is curable if the diagnosis is performed early. For this purpose computer aided diagnosis is used which is fast and reliable as compared to traditional method and this paper presents a YOLO based model for detection of tumor in MRI. The procedure is to use YOLO (version-4) model which is trained using a dataset which contain 3064 T1 weighted, contrast-enhanced (CE) MRI which were preprocessed and labeled using tool. By the help of TL we had the advantage of pre-trained loads of the COCO dataset which boosts the learning and provided the required features for detecting tumor in the brain. The work presented is prepared with the 29-layer YOLO Tiny and fine-tuned to work efficiently and perform task productively and accurately in most cases with solid execution. The outcome of the model is the highest like precision, recall and F1-score beating other previous results of earlier versions of YOLO and other studies like Fast R-CNN.