eng Oriental Scientific Publishing Company Biomedical and Pharmacology Journal 0974-6242 2025-11-06 18 4 68886 article Caroline Lall 1 Anushka Varshney 1 Sanjay Roka 1 Ankur Maurya 2 Chandrakala Arya 3 Manoj Diwakar 1 CSE Department, Graphic Era Deemed to be University, Dehradun, Uttarakhand, India School of Computer Science, Engineering & Technology, Bennett University, India Graphic Era Hill University, Dehradun, Uttarakhand, India The brain is the main organ for regulating as well as coordinating all body processes. A brain tumor is defined as a growth of tissue that arises due to abnormal and uncontrolled cell division and is either malignant (cancerous) or benign (non-cancerous). The overall survival rate for primary brain tumors is approximately 75.2% so accurately and quickly diagnosing a brain tumor is essential. Depending on the method of diagnosis, traditional diagnosticians use a number of approaches that are often time-consuming and prone to human error. Therefore, computer-aided detection (CAD) approaches are becoming more appealing as a way to decrease human error. The ability to take medical Magnetic Resonance Imaging (MRI) scans and use independent computational models has evolved into a second approach to aid in brain tumor detection. In this research, deep learning techniques are used to classify brain tumors into three main categories like glioma, meningioma, and pituitary. A custom Convolutional Neural Network (CNN) model is created, and its performance is compared with two established architectures, that is, VGG-19 and ResNet-50. The VGG-19 model is appreciated for its simple yet powerful structure in image classification, whereas ResNet-50 is known for effectively managing deep networks through residual connections, preventing vanishing gradient issues. The classification accuracies were Custom CNN: 88.6%, VGG-19: 95.83%, and ResNet-50: 97.91%, demonstrating that ResNet-50 is most effective for automated brain tumor detection. ResNet-50 became famous for its capabilities to handle deep networks containing residual connections, a feature that helps mitigate vanishing gradient problems. The outcomes demonstrate how these frameworks, as well as special modifications, assist in attaining high diagnostic efficiency in detecting brain tumors https://biomedpharmajournal.org/vol18octoberspledition/comparison-of-cnn-vgg-19-and-resnet-50-algorithm-for-brain-tumor-detection/ Brain Tumor CNN Deep Learning Detection Machine Learning MRI Prediction ResNet-50 VGG-19