SU‐G‐JeP4‐01: An Assessment of a Microsoft Kinect V2 Sensor for Voluntary Breath‐Hold Monitoring in Radiotherapy

Purpose: To determine whether the Microsoft Kinect Version 2 (Kinect v2), a commercial off‐the‐shelf (COTS) depth sensors designed for entertainment purposes, were robust to the radiotherapy treatment environment and could be suitable for monitoring of voluntary breath‐hold compliance. This could complement current visual monitoring techniques, and be useful for heart sparing left breast radiotherapy. Methods: In‐house software to control Kinect v2 sensors, and capture output information, was developed using the free Microsoft software development kit, and the Cinder creative coding C++ library. Each sensor was used with a 12m USB 3.0 active cable. A solid water block was used as the object. The depth accuracy and precision of the sensors was evaluated by comparing Kinect reported distance to the object with a precision laser measurement across a distance range of 0.6m to 2.0 m. The object was positioned on a high‐precision programmable motion platform and moved in two programmed motion patterns and Kinect reported distance logged. Robustness to the radiation environment was tested by repeating all measurements with a linear accelerator operating over a range of pulse repetition frequencies (6Hz to 400Hz) and dose rates 50 to 1500 monitor units (MU) per minute. Results: The complex, consistent relationship between true and measured distance was unaffected by the radiation environment, as was the ability to detect motion. Sensor precision was < 1 mm and the accuracy between 1.3 mm and 1.8 mm when a distance correction was applied. Both motion patterns were tracked successfully with a root mean squared error (RMSE) of 1.4 and 1.1 mm respectively. Conclusion: Kinect v2 sensors are capable of tracking pre‐programmed motion patterns with an accuracy <2 mm and appear robust to the radiotherapy treatment environment. A clinical trial using Kinect v2 sensor for monitoring voluntary breath hold has ethical approval and is open to recruitment. The authors are supported by a National Institute of Health Research (NIHR) Career Development Fellowship (CDF‐2013‐06‐005). Microsoft Corporation donated three sensors. The views expressed in this publication are those of the author(s) and not necessarily those of the NHS, the National Institute for Health Research or the Department of Health.