From pixels to practice: a scientometric and systematic review of computer vision for clinical robotics

Recent developments in computer vision and artificial intelligence (AI) can transform healthcare, from supporting patients to enhancing surgery. While extensive research exists, a comprehensive overview of how vision and AI facilitate collaborative, situationally aware, and evolving robotic capabilities is still lacking. We address this gap by detailing state-of-the-art contributions in recognition using spatial visual features, deep learning, and other intelligent applications. We concentrate on core computer vision techniques such as object detection, human posture assessment, and semantic mapping, emphasizing their significance across three primary application domains: (1) AI-guided surgical robotics; (2) AI-powered robotic rehabilitation and prosthetics; and (3) AI-enhanced robotics for clinical monitoring and intervention. While existing literature surveys explore vision or robotics independently, our review uniquely focuses on their integrated application within clinical and healthcare settings, highlighting the potential for practical, real-world deployment. We highlight recommendations, difficulties, and important metrics for success. Finally, we recommend future research directions, including sensor fusion and ethical considerations, to ensure safe, reliable, and efficient human-robot collaboration in healthcare in the near future.