2nd/10_Wiki/Topics/DevOps_and_Security/Remote-Rehabilitation.md

id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack

id	title	category	status	canonical_id	aliases	duplicate_of	source_trust_level	confidence_score	verification_status	tags	raw_sources	last_reinforced	github_commit	tech_stack
wiki-2026-0508-remote-rehabilitation	Remote Rehabilitation	10_Wiki/Topics	verified	self	telerehabilitation telerehab digital rehabilitation	none	A	0.9	applied	rehabilitation telehealth devops monitoring healthtech		2026-05-10	pending	language framework typescript nextjs-supabase-webrtc

Remote Rehabilitation

매 한 줄

"매 clinic 의 walls 의 dissolving — 매 patient 의 home 의 becoming 의 PT studio". Remote rehabilitation (telerehab) 의 PT/OT/cognitive therapy 의 delivering 의 video, sensors, gamified exercises 의 via. 2026 의 standard care 의 stroke recovery, post-op orthopedics, chronic pain — 매 reimbursement (CPT 98975-98981) 의 mainstream 의 making.

매 핵심

매 modalities

• Synchronous: 매 live video PT session — 매 therapist 의 form correction 의 real-time.
• Asynchronous: 매 patient 의 records exercise 의 video 의, 매 therapist 의 reviews 의 later.
• RPM (Remote Patient Monitoring): 매 wearables 의 ROM, gait, HR 의 streaming 의 dashboard 의.
• DTx (Digital Therapeutics): 매 prescription apps — 매 Akili EndeavorRx, 매 Pear reSET (deprecated).

매 tech stack 의 typical

• Video: WebRTC (Daily, Twilio Video, Zoom SDK) — 매 HIPAA BAA 의 require.
• Pose estimation: MediaPipe Pose, 매 Apple Vision Pro Body Tracking, 매 Google ML Kit.
• Wearables: Apple Watch, Whoop, IMU patches (BioStamp).
• Backend: FHIR R5 의 EHR integration 의, 매 HL7 Bulk Data API.

매 응용

1. 매 stroke recovery — 매 mirror therapy 의 VR 의.
2. 매 post-ACL 의 ROM tracking 의 IMU 의.
3. 매 chronic low back pain 의 Hinge Health-style 의 daily exercises.

💻 패턴

Pose-based form scoring (MediaPipe + TS)

import { PoseLandmarker, FilesetResolver } from '@mediapipe/tasks-vision';

const vision = await FilesetResolver.forVisionTasks(
  'https://cdn.jsdelivr.net/npm/@mediapipe/tasks-vision@0.10/wasm'
);
const pose = await PoseLandmarker.createFromOptions(vision, {
  baseOptions: { modelAssetPath: '/pose_landmarker_full.task' },
  runningMode: 'VIDEO',
  numPoses: 1,
});

function squatDepthScore(landmarks: any[]): number {
  const hip = landmarks[24], knee = landmarks[26], ankle = landmarks[28];
  const angle = Math.atan2(hip.y - knee.y, hip.x - knee.x) -
                Math.atan2(ankle.y - knee.y, ankle.x - knee.x);
  const deg = Math.abs((angle * 180) / Math.PI);
  return deg < 90 ? 1.0 : Math.max(0, 1 - (deg - 90) / 30);
}

WebRTC 의 HIPAA-compliant session 의

import Daily from '@daily-co/daily-js';

const call = Daily.createCallObject({
  audioSource: true,
  videoSource: true,
  dailyConfig: { useDevicePreferenceCookies: true },
});
await call.join({
  url: signedRoomUrl, // server-issued, BAA-covered
  token: patientJWT,
});
call.on('recording-started', (e) => logToFHIR(e.recordingId, encounterId));

IMU streaming 의 ROM tracking

const device = await navigator.bluetooth.requestDevice({
  filters: [{ services: ['battery_service', 'heart_rate'] }],
  optionalServices: ['0000fff0-0000-1000-8000-00805f9b34fb'],
});
const server = await device.gatt!.connect();
const svc = await server.getPrimaryService('0000fff0-0000-1000-8000-00805f9b34fb');
const ch = await svc.getCharacteristic('0000fff1-0000-1000-8000-00805f9b34fb');
await ch.startNotifications();
ch.addEventListener('characteristicvaluechanged', (e: any) => {
  const dv = e.target.value as DataView;
  const quat = [dv.getFloat32(0), dv.getFloat32(4), dv.getFloat32(8), dv.getFloat32(12)];
  pushROM(quaternionToEulerDeg(quat));
});

FHIR Observation 의 exercise log

const obs = {
  resourceType: 'Observation',
  status: 'final',
  category: [{ coding: [{ system: 'http://terminology.hl7.org/CodeSystem/observation-category', code: 'activity' }] }],
  code: { coding: [{ system: 'http://loinc.org', code: '82290-8', display: 'ROM knee flexion' }] },
  subject: { reference: `Patient/${patientId}` },
  effectiveDateTime: new Date().toISOString(),
  valueQuantity: { value: maxFlexionDeg, unit: 'deg', system: 'http://unitsofmeasure.org' },
};
await fetch(`${FHIR_BASE}/Observation`, { method: 'POST', headers, body: JSON.stringify(obs) });

Adherence nudging (server cron)

export default async (req: Request) => {
  const { data: due } = await sb
    .from('patients')
    .select('id, phone, plan_id')
    .lt('last_session_at', new Date(Date.now() - 86400_000).toISOString());
  await Promise.all(
    due!.map((p) => twilio.messages.create({
      to: p.phone,
      from: TWILIO_FROM,
      body: '오늘 의 5분 의 PT routine 의 done?',
    }))
  );
  return new Response('ok');
};

매 결정 기준

상황	Approach
post-acute stroke	hybrid (sync video 2x/wk + async daily)
chronic pain (>3mo)	async-first DTx (Hinge, Sword)
post-op week 1-2	sync-heavy + RPM continuous
medicare reimbursement 의 target	RTM/RPM (CPT 98975-77 + 98980-81)

기본값: hybrid sync+async + IMU/wearable RPM, FHIR-backed.

🔗 Graph

• 변형: Pose Estimation
• 응용: Stroke Recovery
• Adjacent: WebRTC

🤖 LLM 활용

언제: exercise plan generation, session note summarization, patient-facing Q&A (with guardrails). 언제 X: clinical diagnosis, dosage decisions, medical advice 의 unsupervised 의.

❌ 안티패턴

• Consumer Zoom 사용: BAA 없음 — HIPAA violation.
• PHI 의 client-side log: console.log 의 patient name — 매 audit fail.
• Pose model 의 cloud-only: latency >200ms — 매 form correction 의 useless.
• Adherence ignore: 매 70%+ patients drop off by week 3 — nudging 없으면 ROI zero.

🧪 검증 / 중복

• Verified (CMS RTM/RPM 2025 final rule, Hinge Health 의 BMJ 2024 RCT, MediaPipe Pose docs).
• 신뢰도 A.

🕓 Changelog

날짜	변경
2026-05-08	Phase 1
2026-05-10	Manual cleanup — telerehab tech stack + FHIR/WebRTC/pose patterns