Table 4.

Summary of articles on mobile OA motion monitoring tools.

Reference	Subjects	Data collection settings	Monitored objects	Measures	Mobile devices	Results
Li et al.54	20 trainers	Experiments under the guidance of doctors (controlled)	Non-standard knee movements recognition rate	Insufficient holding time Bent-leg raise Rapid leg movement	Acceleration sensors (MMA7361) Android smartphone (KOA)	89% for insufficient holding time 89.7% for bent-leg raise 89.4% for rapid leg movement
Kim et al.56	13 TKA patients	Manual data input by patients using the app (uncontrolled): Pre-surgery (14 days) Post-surgery (30 days)	Adherence rate	Pre-surgery: educational class; medication and activity protocols Post-surgery: quality-of-life questions; physical therapy exercises	Apple iPad mini (iGetBetter)	Pre-surgery: 3.54 out of 6 occasions (59%) on average; ranged 0–6 occasions Post-surgery: 17.77 out of 30 days (59.2%) on average; ranged 4–30 days
Majumder et al.52	15 adults (aged 20–35 years)	Lab environment (controlled)	Gait recognition rate	Normal walking Standing still Simulated peg leg Simulated leg length discrepancy	smartPrediction system using: Apple iPhone (prototype app) Smartshoe	91% for all the four movement patterns
Lu et al.53	47 adults for walking detector training 12 adults for supervised data 8 adults for unsupervised data	Supervised activity sessions (controlled) Daily life activities (uncontrolled)	Gait recognition from other activities	Stationary Biking Running Vehicle Random movements	Android smartphones: Samsung Galaxy S3 and 4 Google Nexus 5 Intel Xolo	Accuracy was improved by increasing training data Accuracy was higher (at least 5%) with the supervised data than the unsupervised data Accuracy was higher when placed in pant pocket than in hands
Mazilu et al.51	9 Parkinson’s disease patients (6 males & 3 females; mean age = 68.3 years)	Gait-training exercise (controlled) Daily life activities (uncontrolled)—data collected from 5 out of 9 patients.	Gait	User satisfaction FoG duration	GaitAssist system using: Wearable sensors attached to ankles Smartphone (GaitAssist) Earphones	User ratings (5-point scale) on average: system operation = 4; wearability = 4; exercise content = 3.8; subjective opinions = 3.8 FoG duration was decreased: in the gait-training session (3 out of 5 patients); in daily life setting (4 out of 5 patients)
LeMoyne et al.55	1 subject with trans-tibial amputation	Gait analysis in an indoor environment (controlled)	Gait	Stance to stance temporal disparity Time-averaged acceleration from stance to stance	Apple iPhone (accelerometer app) 3D printed adapter	Measures were consistent: Temporal disparity measure, mean = 1.10 (seconds), SD = 0.02 with a 96% CI Acceleration measure, mean = 1.47 (g’s), SD = 0.02 with a 96% CI
Shin and Wuensche50	10 adults (8 males and 2 females) 10 golf players (3 males and 7 females; mean age = 37.7 years)	Golf shots measuring session (controlled) Virtual 3-hole golf course in an outdoor golf club (uncontrolled)	Golf game-related movements	Driving distance Angular velocity User perception s of usability, enjoyment, effectiveness, realism	Android smartphones (G-Swing app): Huawei U8100 Samsung s5620 and I9000 HTC One X809 Motorola Droid 2	Formula for short distance estimation: (a) Driver = 15.2 × angular velocity – 19.1; (b) Iron = 10.7 × angular velocity – 6.0 The game concept, the required motions, and the duration of the game were perceived as suitable for arthritis patients
Chandra et al.58	3 physicians (mean age = 30 years) 11 patients (age ranged 16–75 years)	Interview		Prescribed exercise at home: Comments during interviews	N/A	Design guidelines identified: Target patients’ understanding of exercises and the correctness of their performance Present simple visualizations Focus on accurate and informative depiction of data Include reminder and scheduling systems for exercise Enhance communication channels between patients and therapists Keep patients’ data private