This paper develops an efficient vision-based real-time vein detection algorithm for preclinical vascular insertions. Mouse tail vein injections perform a routine but critical step in most preclinical applications. Compensating for poor manual injection stability and high skill requirements, Vascular Access System (VAS) has been developed so a trained technician can manually command the system to perform needle insertions and monitor the operation through a near-infrared camera. However, VAS’ vein detection algorithm requires much computation and is, therefore, difficult to reflect the real-time tail movement during an insertion. Furthermore, the detection performance is often disturbed by tail hair and skin pigmentation. In this work, an effective noise filtering algorithm is proposed based on convex optimization. Effectively eliminating false-positive detections and preserving cross-sectional continuity, this algorithm provides vein detection results approximately every 200 ms at the presence of tail hair and skin pigmentation. This developed real-time tail vein detection method is able to capture the tail movement during insertion, therefore allow for the development of an automated Vascular Access System (A-VAS) for preclinical injections.

This content is only available via PDF.
You do not currently have access to this content.