LIDAR-based gesture recognition system for interactive navigation of 2D and 3D medical data

The integration of advanced imaging technologies in surgery has significantly improved intraoperative decision-making. However, conventional interaction methods, such as touchscreens and keyboards, present challenges in sterile en vironments, requiring additional personnel and increasing the risk of contamination. This thesis proposes a LiDAR-based gesture recognition system to enable contactless manipulation of medical images and 3D anatomical volumes. By leveraging the Intel RealSense L515 depth camera, the system detects and interprets surgeon hand gestures, allowing for intuitive navigation of DICOM images and volumetric reconstructions without physical contact. The developed algorithm comprises hand detection, gesture recognition, and interactive control mechanisms. The system maps hand movements to commands such as zooming, rotation, and scrolling, ensuring real-time responsiveness and ergonomic usability. Experimental validation in a controlled setting assessed hand tracking accuracy, gesture recognition reliability, and robustness against occlusions and distance variations. The results demonstrate high precision at short ranges (up to 3 meters), with decreasing accuracy at greater distances or extreme hand orientations. Despite its limitations, such as tracking stability at extended distances and differentiation of users in multi-person environments, the system presents a viable approach to hands-free interaction in surgical settings. Future improvements include optimizing recognition algorithms, integrating real-time occlusion handling, and exploring direct control of surgical fluoroscopes. The proposed solution represents a step forward in human-machine interaction for medical applications, enhancing workflow efficiency, surgeon autonomy, and patient safety