Performance Analysis of Brain Tumor Prediction with ML Techniques

Brain tumors are critical and life-threatening conditions that require accurate and timely diagnosis for effective treatment. Manual interpretation of MRI scans is not only time-consuming but also highly dependent on the radiologist’s expertise, which can introduce variability and delays. To address this, our study presents a comprehensive comparative analysis of machine learning (ML) and deep learning (DL) approaches for automated brain tumor classification using MRI scans. The dataset comprises four classes: glioma, meningioma, pituitary tumor, and no tumor.We evaluate six models—two ML-based (Support Vector Machine and Random Forest using Gabor and HOG features) and four DL-based (VGG19, ResNet50, DenseNet121, and a Custom Convolutional Neural Network). Among these, the Custom CNN achieved the highest performance with 96.03% accuracy, while Random Forest was the top ML performer at 89.40%. The DL models demonstrated their ability to learn intricate features automatically, unlike ML models which rely on hand-crafted features. This comparative study reinforces the superiority of DL techniques in medical imaging tasks, particularly brain tumor diagnosis. Moreover, by automating the diagnostic process, such systems reduce manual workload, enhance diagnostic reliability, and support sustainable healthcare delivery—especially in low-resource or remote regions with limited access to specialized medical professionals.