“Free to Fly”: Development and evaluation of a novel exergame with a low-cost 3D tracking method for post-stroke rehabilitation

Abstract

Stroke is one of the main causes of long-term disability worldwide. Conventional upper limb physiotherapy programs can be tedious, expensive and require physical transportation. Video games can help solve these problems. In fact, recent studies show that health professionals are increasingly interested in using computer games for post-stroke rehabilitation. However, commercial possibilities can be inaccessible and inadequate to the needs of patients and therapists. Also, among academic research solutions, there is a prevalence of: exoskeletons, inertial, magnetic and ultrasonic sensors, all electronics of significant cost and difficult to obtain (imported). In Brazil, it is important to remember that financial conditions can be minimal. In this work, a solution for the following problems: time (for the proper set up of every session; lack of time of the therapist to accompany the rehabilitation sessions), space (for equipment to be stored; for some camera systems to capture the full movement properly), and cost (for private clinicians and patients themselves, some systems can be inaccessible, because they can require sophisticated electronic devices) was attempted, trying to develop a very accessible post-stroke rehabilitation exergame alternative. Thus, a video game with an innovative alternative for tracking the 3D movement of the upper limb was made, after reviewing the bibliographic and patent literature and identifying the needs directly from an occupational therapist and her stroke patients, which uses optical capture with a regular camera and colored sphere markers, while maintaining lightweight real-time processing on mobile devices. The game has 2D and 3D versions, for both Windows and Android, simplified interface and progress monitoring. The fact that the controller can be handcrafted by the users makes the game very low-cost, possible to be distributed worldwide, reaching a large number of people, and possible to be played and monitored remotely. The proposed system was tested in order to find out how accurate it can be, compared to a gold standard system (a goniometer). It has been found that the system has limitations, such as low accuracy at obtuse angles, illumination variation, small spheres (the type of marker utilized), occlusions, camera distortions and motion blur. Still, it can be faster than using fiducial markers or Deep Learning, since it is a simpler algorithm, leading to a higher frame rate, which was demonstrated in this work, and can be very accurate when respecting the aforementioned problems. In conclusion, the system seems promising, due to its accessibility, very low cost, customization to the needs of patients and therapists and good tests results, as a complementary alternative for post-stroke rehabilitation.