Abstract
Movement disorder diagnosis often relies on expert evaluation of patientvideos, but sharing these videos poses privacy risks. Current methods forde-identifying videos, such as blurring faces, are often manual, inconsistent,or inaccurate. Furthermore, these methods can compromise objective kinematicanalysis - a crucial component of diagnosis. To address these challenges, wedeveloped SecurePose, an open-source software that simultaneously providesreliable de-identification and automated kinematic extraction from videosrecorded in clinic settings using smartphones/tablets. SecurePose utilizes poseestimation (using OpenPose) to extract full body kinematics, track individuals,identify the patient, and then accurately blur faces in the videos. Wevalidated SecurePose on gait videos recorded in outpatient clinic visits of 116children with cerebral palsy, assessing both the accuracy of itsde-identification compared to the ground truth (manual blurring) and thereliability of the intermediate steps of kinematics extraction. Resultsdemonstrate that SecurePose outperformed six existing methods in automated facedetection and achieved comparable accuracy to robust manual blurring, but insignificantly less time (91.08% faster). Ten experienced researchers alsoconfirmed SecurePose's usability via System Usability Scale scores. Thesefindings validate SecurePose as a practical and effective tool for protectingpatient privacy while enabling accurate kinematics extraction in clinicalsettings.