Detecting and Localizing Lesions in Medical Images by using Faster R-CNN Model

Early and accurate identification of lesions in medical images plays a crucial role in ensuring proper diagnosis and timely treatment. However, analyzing large volumes of radiological data manually is both time-consuming and prone to errors, often relying heavily on the expertise of radiologists. To overcome these challenges, this project presents an automated framework for lesion detection using the Faster Region-Based Convolutional Neural Network (Faster R-CNN). The proposed system utilizes annotated medical image datasets to train a deep learning model that can accurately detect and localize lesions by generating bounding boxes. The process begins with essential data preprocessing steps such as image normalization, resizing, and organizing annotation data into formats suitable for model training. A custom data pipeline built using PyTorch is implemented to efficiently handle data loading, augmentation, and batching, ensuring effective training. The Faster R-CNN model, combined with a ResNet backbone, is fine-tuned to capture detailed lesion-specific features while optimizing both classification and localization tasks. The model’s performance is evaluated using standard object detection metrics, including mean Average Precision (mAP), precision, and recall, to ensure its reliability and accuracy. The experimental results show that the model performs well across different imaging conditions, highlighting its potential for practical clinical use. Overall, this automated framework is designed to assist radiologists by reducing their workload, improving diagnostic consistency, and enabling faster image analysis. It also provides a foundation for future improvements, such as detecting multiple lesions, integrating with existing medical imaging systems, and supporting real-time decision-making in healthcare environments.