Automated Brain Tumor Detection And Localization Using YOLO-Based Deep Learning Algorithm

Abstract

Classifying brain diseases is a very difficult task because of their complexity and sensitivity. Developing an efficient treatment plan calls for early detection and diagnosis since brain tumors are severe and occasionally fatal. A key medical imaging tool, magnetic resonance imaging (MRI) enables the thorough, non-invasive view of the internal structures of the brain. Magnetic resonance imaging (MRI) is essential for diagnosing and treating brain tumors. Beginning with dataset preprocessing, the approach is applicable to MRI scans and clinical data from individuals with various brain disorders, including tumor and non-tumor cases. The dataset consists of training and testing sets. Detecting MRI tumors calls for several procedures including feature extraction, classification, and image post-processing. The system uses the YOLO (You Only Look Once) algorithm with the CNN model, a pre-trained model using the concept of transfer learning, for classifying and detecting brain tumors. The proposed framework not only improves the performance of training a better model by using the pre-trained model but also refines the dataset for better accuracy by using thresholding and increases the number of images in the dataset by using data augmentation. Early results indicate that the family of YOLO models outperforms earlier architectures since it uses a compound coefficient with a constant ratio to scale all depth, width, and resolution of an image. The findings also showed that by increasing the baseline architecture, the model can capture complex features, and thus the overall performance of the system is significantly improved.