J-Aamir/Breast_MNIST_CNN

BreastMNIST ResNet18 with Grad-CAM

Overview

This repository contains a fine-tuned ResNet18 model for binary breast ultrasound image classification using the BreastMNIST dataset from MedMNIST.

The project demonstrates:

• Transfer Learning
• Medical Image Classification
• Explainable AI (Grad-CAM)
• Error Analysis

Dataset

BreastMNIST (MedMNIST v2)

Task: Binary Classification

Classes

• 0 → Benign
• 1 → Malignant

Model

Architecture:

• ResNet18

Framework:

• PyTorch

Input Size:

• 224×224

Output:

• Binary Classification

Explainability

Grad-CAM visualizations are included in the GitHub repository.

GitHub Repository

https://github.com/J-Aamir/BreastMNIST-ResNet18-GradCAM

Intended Use

Educational and research purposes only.

Not intended for clinical diagnosis.