Combining RGB-D Sensing With Adaptive Force Control in a Robotic Ultrasound System for Automated Real-Time Fistula Stenosis Evaluation

Fistulas are often referred to as the lifeline for hemodialysis (HD) patients because it is crucial for their treatment; thus, physicians use ultrasound imaging to evaluate these fistulas and ensure they are free from obstruction. However, this evaluation process is typically time-consuming and heavily reliant on the operator's skill, and current methods lack generalizability. To address these limitations, this study presented a novel robotic ultrasound system (RUS) integrating an RGB-D sensor and an adaptive force sensor for more generalized and efficient fistula obstruction evaluation. A composite AI model, combining YOLOv5 for real-time fistula recognition and U-Net++ for precise lumen segmentation, was integrated into the system. Subsequently, the degree of stenosis (DOS) was automatically calculated, and Doppler ultrasound was applied to the most stenotic point for detailed blood flow analysis. The experimental results demonstrated that the improved RUS achieved high accuracy, with an average path planning error of less than 1.8 mm, a mean absolute error of less than 1% in stenosis calculations, and a Doppler ultrasound scanning position error of less than 0.5 mm. The system exhibited greater accuracy and reliability in evaluating fistula obstruction compared to existing approaches. This sensing solution offered an effective method for real-time stenosis evaluation in clinical HD settings.