CareLoop — AI Health Diary & Clinical Note Assistant (SOAP)

CareLoop is an AI-powered assistant that summarizes, cleans, and structures health-related information for both patients and clinicians—combining an AI Health Diary Summarizer with a Clinical Note Cleaner (Voice → SOAP) into one closed-loop prototype.

What the Prototype Does

Patient Mode — Health Diary Summarizer (Text/Voice)

• Log symptoms, food, mood, sleep via text or voice
• Get temporal visual summaries, trend highlights, and gentle, non-medical guidance prompts
• Generate a Pre-Visit Summary (last N days) for faster clinical context

Clinician Mode — Clinical Note Cleaner (Voice → SOAP)

• Record clinician dictation → transcribe → generate structured SOAP notes:
  • Subjective, Objective, Assessment, Plan
• Optional “evidence linking”: SOAP sentences can reference the relevant transcript spans for traceability

Closed Loop — Plan Back to Patient

• Convert SOAP Plan into a simple patient checklist (follow-ups, measurements, reminders) to support continuity of care

---

2-Minute Demo Flow (What the Jury Should Try)

1. Patient enters or records a short diary entry (text/voice)
2. App displays trend visualization + a concise summary
3. Click Generate Pre-Visit Summary
4. Clinician records a short dictation
5. App outputs a clean SOAP note
6. Click Send Plan to Patient to generate a follow-up checklist
7. Open Privacy Panel to verify synthetic-data + PHI controls

---

Architecture (High-Level)

• Speech → Text for diary entries and clinician dictation
• Medical entity extraction (symptoms/meds/durations) to improve structure
• LLM structuring into:
  • patient summaries + trend highlights
  • clinician-ready SOAP output (schema-based)

See: docs/ARCHITECTURE.md

---

Azure Services Used

• Azure Speech-to-Text — transcription (patient + clinician)
• Azure OpenAI / AI Foundry — summarization + SOAP structuring
• Text Analytics for Health — medical concept extraction / NER
• (Optional) Azure ML — deployment of the API as a managed endpoint

---

Privacy & Compliance (Hackathon Safe-by-Design)

• No real personal data / no PHI is used in this repository.
• All demos run on synthetic or anonymized content only.
• Optional PHI redaction is available for extra safety in demos.