Problem

Detecting and segmenting brain tumors from MRI scans is challenging due to variations in tumor size, shape, and location. Traditional methods require manual review by radiologists, which is time-consuming and prone to human error.

Method

The study employed various machine learning models, including logistic regression, CNNs, ResNet, U-Net for semantic segmentation, and EfficientDet for object detection, to classify and segment brain tumors from MRI scans.

Results

The results showed that CNNs and ResNet achieved high accuracy in classification tasks, while U-Net performed well in semantic segmentation with an IoU of 70%. Object detection using EfficientDet was less effective, with a validation IoU of 0.075.

Implications

The study demonstrates the potential of deep learning techniques in improving the accuracy and efficiency of brain tumor diagnostics, with implications for enhancing clinical outcomes through more precise and timely diagnosis.