A statistical analysis of external respiratory motion using Microsoft Kinect

Abstract

External respiratory motion has been extracted from a set of normal volunteers (14 males and 6 females) using Kinect for Windows - a low cost 3D depth camera, which is non-invasive and where there is no marker placement requirement. Such motion was captured on three separate occasions for each individual. We present the first analysis of this respiratory motion. All volunteers were registered to a common reference using a 2D affine transformation so that inter- and intra-session analysis could be successfully completed. A map representing the standard deviation of the depth displacement across the chest was obtained to categorise the dominant mode of breathing. To investigate this hypothesis, the first Eigen image obtained was segmented. Then statistical characteristics of this motion were extracted for each individual including amplitude, period, end exhale, as well as baseline drift and duty cycle, as represented by binning the data into 10 phases as commonly used in dynamic CT. We also demonstrate the intrinsic relationship between respiratory frequency and respiratory amplitude. This work demonstrates for the first time the effectiveness of Kinect in acquiring external respiratory motion data across a significant cohort of volunteers without the need for marker placement. Moreover this analysis offers insight into the inter- and intra session variations in respiratory motion. Such information may be used to inform the development of motion correction and motion prediction strategies in diagnostic and therapeutic imaging.