A Computer Vision enabled Smart Tray (CVST) was designed for use in medical training for Central Venous Catheterization (CVC). The effects of background color on the ability of the computer vision algorithm to distinguish between tools and the tray was investigated. In addition, the computer vision algorithm was evaluated for accuracy in tool detection. Results indicate that a white monochromatic background is the most useful for segregating background from medical tools, and the algorithm was successfully able to detect 5 different CVC tools both individually and as a group in various arrangements, even when tools overlapped or touched. When the system was in error, it was nearly always due to one tool which has a color similar to that of the background. The CVST shows promise as a CVC training tool and demonstrates that computer vision can be used to accurately detect medical tools.

A machine learning (ML) object detection algorithm was developed to replace the original color-based image detection algorithm for the Dynamic Haptic Robotic Trainer Plus (DHRT+). This image recognition system was used for medical training in Central Venous Catheterization (CVC). This image tracking allows for the training system to provide accurate performance feedback to the user during the training process. The ML object detection algorithm was developed and evaluated using training data. The results indicate that increasing the training data set improves the detection system’s accuracy. The system was found to have an overall precision rate of 90.9% and a recall rate of 81.69%. This new ML model will be implemented into the DHRT+ system and used to train medical residents.