Comparative Study of Deep Learning Approaches for Object Detection in Medical Imaging
Comparative Study of Deep Learning Approaches for Object Detection in Medical Imaging
Dr.K.Saraswathi1 , Mr. M. Shanmugapriyan2
1Associate Professor, Department of General Engineering, Annai Mathammal Sheela Engineering College, Erumapatty, Namakkal (DT), Tamil Nadu, India.
1saraswathimuruganams@gmail.com
2Undergraduate Scholar, PG & Research Department of Computer Science, Nehru Memorial College (Autonomous), Puthanampatti, Tiruchirappalli (DT), Tamil Nadu, India
2shanmugapriyan1806@gmail.com
Abstract— Object detection in medical imaging has emerged as a critical application of deep learning, enabling automated identification of pathological features across diverse modalities such as fundus photography, OCT, CT, MRI, and X-ray. This study presents a comparative analysis of state-of-the-art deep learning models—including Faster R-CNN, YOLOv5, RetinaNet, and EfficientDet—evaluated on benchmark datasets for tasks such as retinal abnormality detection, tumor localization, and lesion segmentation. Performance metrics including accuracy, precision, recall, inference time, and model complexity are systematically compared. The results highlight trade-offs between speed and diagnostic accuracy, offering insights into model suitability for real-time clinical deployment versus offline analysis. This comparative framework aims to guide researchers and practitioners in selecting optimal architectures for specific medical imaging tasks.
Keywords: Object Detection, Deep Learning, Convolutional Neural Networks (CNN), Computer Vision, YOLO, Faster R-CNN, Real-time Detection