A survey on artificial intelligence and robotic systems in biomedical applications: challenges and future prospects

Abstract

The rapid emergence of new disease variants, ongoing health issues, various types of accidents, and the continuous challenges associated with aging pose significant burdens that the healthcare sector must address daily. Historically, these challenges were managed through manual methods until the advent of smart technology. The integration of smart technologies, such as AI-driven robotics, has created a paradigm shift in healthcare. However, the growth and global acceptance of these innovative systems face several challenges, including interpretability, uncertainty, and clinical trust; human-robot interaction in clinical settings; robustness, adversarial safety, and cybersecurity; hardware limitations; effective sensing and multimodal data integration; and the need for outcome-oriented, long-term clinical trials, among others. This review presents enhanced approaches to tackle these challenges through the application of advanced AI-powered robotics. Key strategies include the development of sophisticated AI models, replacing single-sensor systems with sensor fusion (incorporating vision, force, and biosignals) to prevent failures, secure model execution through encrypted inference and hardware enclaves, implementing shared and adjustable autonomy frameworks, enhancing anomaly detection and input validation, and utilizing adversarial training and robust optimization with certified bounds.