Gains:
- Secure, auditable deployment of AI in remote assessment and exercise tracking
- Ability to manage data security, informed consent and technical limits in remote sessions
- Ability to maintain human control and emergency planning while transforming remotely collected data into clinical decisions
Physiotherapy is no longer done only in the clinic. Tele-rehabilitation (remote rehabilitation; physiotherapy service that reaches the patient at home via video call, application and sensors) removes geographical barriers, facilitates transportation for the patient and ensures continuous follow-up. Artificial intelligence (AI) powers this model: processing remotely collected video and data, automating exercise tracking, and managing the flow of reminders and feedback. But being remote does not diminish clinical responsibility; On the contrary, it makes privacy, security and emergency management even more critical. In this unit you will learn how to set up AI-assisted tele-rehabilitation in a safe, ethical and auditable way.
Opportunities and risks of the remote model
The opportunities of tele-rehabilitation are clear: access increases, patient comfort increases, follow-up becomes more frequent, and data flows constantly. AI processes this data (exercise videos, compliance records, pain logs) and provides the physiotherapist with a rich picture. But the remote model magnifies three new risks: examination limit (can't touch, can't test by hand; some of the findings are invisible), security/emergency (the patient is exercising alone and no one is around if something goes wrong), and data security (video and health data flow over the internet and need to be protected). AI helps manage these risks but does not eliminate them.
Caution: Remote assessment cannot provide some components of an in-person examination (palpation, manual muscle testing, end-feel). Remote data collected with AI cannot fully fill this gap. Clinical decision should be made with awareness of this limit; In case of doubt, face-to-face evaluation should be requested.
Üç temel gereklilik
Every telerehabilitation flow, whether using AI or not, must pre-define three things:
- Aydınlatılmış onam. The patient must understand and clearly approve what remote service is, its limits, data usage and the role of AI.
- Data security. Image, audio and health data must be processed with encrypted, secure and KVKK-compliant tools; should not be recorded without permission.
- Emergency plan. What to do in case the patient deteriorates or in case of an emergency (who to call, 112 for which symptom) should be written in advance.
necessity
Why is it critical?
Application
Informed consent
Legal and ethical basis
Written consent, explanation of limits, notification of AI use
Data security
Health data is sensitive
Encrypted platform, no unauthorized recording, secure storage
emergency plan
The patient works alone
Written protocol, emergency numbers, red flag list
human control
AI does not decide
Critical data falls to the physiotherapist, the decision belongs to the person
technical spare
The connection may break
Alternative communication method, session rules
Step by step: Set up AI-powered remote tracking
- Clarify consent and boundaries. Explain the model, limits, and AI role to the patient and obtain written consent.
- Choose a safe vehicle. Encrypted, KVKK compliant platform; AI tool that protects personal data.
- Write your emergency plan. Red flags, emergency numbers, when 112.
- Set up the data collection flow. Exercise video, pain diary, compliance record; approved and safe.
- Position AI as an assistant. Let it summarize the data and highlight the signals; Don't let him decide.
- Maintain human control. Let the physiotherapist see the critical data; Let the human make the clinical decision.
- When in doubt, ask face to face. If distance is not enough, refer to the clinic/another specialist.
Three mini cases (in numbers)
Case 1 — Gaining access. A low back pain patient lives in a rural area, 90 minutes away from the clinic. 1 video session per week + AI-powered daily exercise tracking installed. Attended 95% of sessions at 8 weeks (previous trial at clinic had 50% attendance, due to transportation). Consent was obtained and the data was kept in an encrypted platform.
Case 2 — The value of an emergency plan. The 62-year-old patient, who was under remote follow-up, wrote "pressure in my chest and shortness of breath" on the application outside the session. The predefined emergency plan flagged this as a red flag and immediately displayed a 112 message to the patient and sent a push notification to the physiotherapist. Lesson: in the remote model, emergency planning saves lives; AI carried the signal, the decision was driven by the human.
Case 3 — Inspection limit. In the remote evaluation, the AI summarized the shoulder movement as “normal” over the video. But the physiotherapist noticed the night pain and non-mechanical pattern described by the patient and requested a face-to-face evaluation; A different finding was detected upon palpation and the patient was referred to a physician. Lesson: remote data is not a substitute for examination.
Weak prompt / Strong prompt
Weak: "Remotely evaluate the exercise video sent by the patient and tell him what to do."
It puts clinical judgment on AI and ignores boundaries and safety.
Strong: "Summarize the pose estimation output of this remote exercise video (estimated, must be verified). List possible errors and points of caution observed. Highlight separately if there are any red flags. Recommend no clinical judgment; summarize for physical therapist review only. Specify limits (2D, camera)."
Four copyable templates
Task: Write a DRAFT informed consent text for remote monitoring. Scope: [video session + AI-assisted monitoring]Must include: description of the service, boundaries (no tactile examination), data use, AI role, privacy, what to do in case of emergency, consent statement. Plain language, at a level the patient can understand. Do not use without approval.
Task: Produce DRAFT emergency and escalation plan. Context: [patient group, remote monitoring] I want: Red flag list, patient's action for each (call 112 / us), automatic notification rule to physiotherapist. The clinical decision is mine.
Task: Summarize remotely collected exercise data (decision making).Data: [pose printout + patient pain log + compliance record]I want: Brief summary for the physical therapist, signals that stand out, "predict/must be verified" tags, highlight any red flags. Not a suggestion, but a summary.
Task: Produce technical/privacy checklist. Context: [platform and AI tool to be used] I want: Encryption, no unauthorized recording, data retention period, KVKK compliance, backup communication, bullet point checklist for patient consent.
Tip: “Human control” is non-negotiable in the remote model. AI can collect and summarize data; But every clinical decision, every interpretation of deterioration and every referral decision should fall on the physiotherapist. Position automation as an assistant, not a decision maker.
Patient selection: who is suitable for remote monitoring
The most skipped step in tele-rehabilitation is deciding which patient fits this model. Not every patient is suitable for remote follow-up. Factors that determine suitability include the patient's ability to use technology, a safe exercise space in their home, the stability of their condition (acute/unstable conditions are less amenable to remote monitoring), and red flag risk. For example, a chronic low back pain patient who is stable, cooperative and has access to technology is ideal for remote monitoring; A patient with a new and complex neurological condition at high risk for falls requires initial in-person evaluation. AI can help provide a checklist for this selection, but the decision on suitability is a clinical judgment and rests with the physical therapist. Monitoring the wrong patient remotely creates risks in terms of both safety and treatment quality.
Tip: Think of remote monitoring not as “all or nothing” but as a hybrid model: critical assessments face-to-face, follow-up and training sessions remotely. This hybrid approach best balances safety and access for most patients.
Common mistakes
- Skipping consent. Remote service and AI use requires explicit consent.
- Starting without a contingency plan. The patient works alone; The plan saves lives.
- Forgetting the inspection limit. Remote data is not a substitute for palpation and manual testing.
- Unsafe driving. Processing health data on an unencrypted platform is a violation of KVKK.
- Let AI decide. Clinical decision and interpretation of deterioration is human.
- To insist on doubt. If distance is not sufficient, face-to-face evaluation should be requested.
In summary
Tele-rehabilitation enhances access and follow-up; AI processes data remotely and provides a rich picture to the physiotherapist. But being remote does not reduce responsibility. Three things are mandatory: informed consent, data security, and an emergency plan. Be aware of the inspection limit, drive safely, position AI as an assistant rather than a decision maker, maintain human supervision, and when in doubt, seek a face-to-face evaluation.
Application task
Design a remote monitoring scenario. Produce an informed consent draft with the first template and a contingency plan with the second template (at least five red flags). Then, with the fourth template, create a technical/privacy checklist and evaluate a platform you're considering using against it.
checklist
- [ ] I obtained informed consent and explained the boundaries to the patient.
- [ ] I chose an encrypted, KVKK compliant platform and secure AI tool.
- [ ] I prepared a written emergency and escalation plan.
- [ ] I considered the limits of remote examination in my clinical judgment.
- [ ] I positioned AI as an assistant; I made the clinical decision myself.
- [ ] I applied the rule of asking for a face-to-face evaluation when in doubt.