Computer Vision Quality Assurance for Medical Linear Accelerators

Abstract

Each year, millions of patients undergo radiation treatment utilizing medical linear accelerators (linacs) due to their noninvasiveness and submillimeter precision. Although linacs offer innovative and advanced treatment plans, some mechanical QA procedures remain subjective due to human error. One of these being the verification of the optical distance indicator and vertical offset between the linac head and treatment couch. In order to address this clinical need, StriX has worked on further advancing an automated Computer Vision Quality Assurance (C.V.Q.A.) device designed by the Mayo Clinic. The team’s newly integrated time-of-flight laser distance sensor is able to provide millimeter distance measurements through innovative C++ code run through a Raspberry Pi 5 module that wirelessly computes and transmits data through WiFi to a server accessed using a personal laptop or phone. Through the resources available, StriX’s solution has provided repeatable and precise distance measurements. This includes average variation in measurement range of 0.576 mm, coefficients of variability in repeatability as low as 0.05%, and variance in reproducibility of 0.15%. Despite the findings being slightly off from the ideal 1 mm value, calibrations can be made to the prototype to achieve maximum proficiency. Although further design and developments await this device, it provides accurate millimeter readings that do not rely on human subjectivity. This prototype can also reduce the QA session times and provide a cost-friendly solution to current devices. StriX’s mission with this new prototype is to ultimately ensure patients receive safe and precise radiation treatment.

Video