Work

Deep learning-based automation framework for detecting multiple human disabilities in secure medical imaging systems

Year: 2025

Type: article

Abstract: Deep learning has become an influential tool in the machine learning domain, enabling artificial intelligence systems to identify complex patterns in two-dimensional (2D) images and present precise predictions from massive datasets. One of the deep learning’s promising uses is human disabilities detection, a diagnostic field for enhancing the identification and classification of people with physical and sensory limitations. In this research, a deep learning framework is proposed for identifying and transferring six primary human disabilities, namely, blindness, Down syndrome, dwarfism, cerebral palsy, prosthetic arms, and prosthetic legs in a secure medical imaging system. The proposed framework is evaluated based on broadly accepted performance metrics to accurately measure human disability detection in images with complex interfering objects where various visual elements are partially blocking or overlapping the person of interest, which makes it challenging for the deep learning models to determine relevant human disability features. The performance of the proposed approach is evaluated by implementing the current state-of-the-art object detection models, RetinaNet, Single Shot MultiBox Detector (SSD), Faster Region Convolutional Neural Network (FR-CNN), YOLOv11, and EfficientNet. A robust security technique is developed to verify the privacy, security, and integrity of human disabilities in medical imaging systems. The experimental results demonstrated that the proposed framework enables human multi-disabilities detection accuracy exceeding 97% and proves potential image transmission in secure medical imaging systems.

Source: PeerJ Computer Science

DOI: 10.7717/peerj-cs.3381