| Chae et al [29] |
Machine learning (ML) algorithm implemented by a convolutional neural network ascertaining what types of sensor data can detect home exercise activities most accurately via a cross-validation test |
Activity monitoring using wearables (smartwatch, smartphone, and apps [Android Studio 2.3, Google]) |
The AI-based rehabilitation system connected end users and therapists at a distance and made it possible to share end users’ home exercise data with therapists at remote locations. The system assisted participants in the intervention group to record their exercise time, obtain their own home exercise results, and communicate with a clinician, yielding visible improvement and acting as a motivation—features that were not available to those in the control group. |
| Aharon et al [28] |
A real-time ML algorithm processing the digital questionnaires |
Well-Beat platform, which created a profile for each end user by processing a digital questionnaire assessing end users’ initial state |
After processing the questionnaires, the system created a profile for each end user and presented it on the health care providers’ toolbar. The toolbar included the end user’s persistence level, readiness for change level (maturity), self-efficacy level, main motivational driver, and barrier, as well as what the health care provider should watch out for in communication with them. Based on this information, the engine recommended the personalized end-user dialogue to use, in terms of the tone, the style, and motivation factors, as well as what to avoid saying. |
| Burns et al [33] |
Supervised ML approach |
Exercise recognition using wearables, that is, smartwatch device with embedded inertial sensors (Apple Watch [series 2 and 3] with the PowerSense app, sampling at fs=50 Hz) |
The AI-based system assessed the feasibility of performing shoulder physiotherapy exercise recognition with inertial sensor data recorded from a wrist-worn device to enable objective measurement of home shoulder physiotherapy adherence. |
| Thiengwittayaporn et al [30] |
Rule-based and AI techniques (eg, determining the disease stage based on decision tree) |
Education and assessment for the stage of disease via Love-Your-Knee mobile app (Android) |
The AI-based system was used as a personalized solution and recommending the appropriate exercise types and the number of sets for each end user by assessing the stage of the disease, monitoring disease progression, and promoting physical therapy and rehabilitation exercise. |
| Capecci et al [31] |
Algorithms of AI (patent pending), a neural network was used to recognize the performed exercise and segment the data into single repetitions |
The ARC Intellicare system (an AI-powered and inertial motion unit-based mobile platform) consisting of a set of 5 inertial sensor inserted in slap supports, a tablet with a dedicated app, and a charging station |
The AI-based system allowed rehabilitation professionals to prescribe exercises according to specific therapeutic needs and to monitor end users’ performances and progresses remotely. The counting of the number of exercise repetitions correctly performed was the output of the developed AI algorithm. Real-time feedback was provided to the end user through the app user interface. |
| Ramkumar et al [32] |
ML algorithms |
Activity monitoring using wearables (knee sleeve, personal iPhones [Apple], and mobile app termed TKR [Focus Ventures]) |
The system transmitted the wearable knee sleeve motion data to the smartphone, then transmitted these and all other data to the dashboard, then analyzed these data by the ML algorithms to actively record and check daily compliance in order to provide automated reminder notifications based on the end user’s compliance to the program. |
