Automatic Patency Discrimination in the Pig Bilateral Femoral Veins for Biomedical Implants

Abstract

Free flap surgeries require hourly monitoring to detect vascular compromise. If not caught promptly, the flap can be lost. Monitoring free flaps using the gold standard requires experienced operators to interpret blood flow signatures, which are often difficult to distinguish from background noise. Previously reported hardware-only automatic patency classification showed a high sensitivity, specificity, and a low false-positive rate, but it was demonstrated using bulky discrete electronics and a syringe pump to generate the expected flow rates. In this paper, we investigate automatic hardware-only patency classification on blood flow data collected from the bilateral femoral veins during flow and occluded states using SPICE simulations in an IBM 130-nm CMOS process with a 1-V supply voltage and a 200-ms window length. Experimental results show a very high sensitivity (99.45%), specificity (99.93%), and very low false-positive rate (0.07275%) at just 8.715 $\mu \text{A}$ . This paper shows that automatic hardware-only patency classification is effective for monitoring patency on real pig blood flow data. The demonstrated classifier’s performance makes it suitable for integration as part of a wirelessly-powered biomedical patency monitor.