HaruMD — Pressure Ulcer AI
A clinical documentation app that turns a single wound photo into a structured pressure-ulcer assessment, running on-premise and usable offline at the bedside.
Overview
HaruMD is a clinical documentation app for hospital medical staff that turns a single wound photo into a structured pressure-ulcer assessment. An on-premise AI engine estimates the ulcer stage, size, depth and necrotic-tissue ratio, and overlays the wound boundary directly on the image. Clinicians review and correct the AI result, track each patient's lesions on a calendar, and generate before/after comparison reports that flag improvement or deterioration. Built with React Native (Expo) and a FastAPI backend running a local MLX model, HaruMD keeps patient images inside the hospital network and stays fully usable offline — cached lesion photos and analyses remain viewable in airplane mode, which matters at the bedside. HaruMD is an assistive tool and does not replace clinical judgment.
Highlights
- • Single-photo AI staging: pressure-ulcer stage, size, depth and necrotic-tissue ratio
- • Wound-boundary polygon overlay rendered on the original image
- • Clinician review and correction workflow with a record of medical edits
- • Per-patient calendar tracking and before/after comparison reports (improved / worsened)
- • On-premise FastAPI + local MLX AI — patient images stay within the hospital network
- • Offline-first: cached lesion photos and analyses remain viewable without connectivity
Details
- • Role-based access for super-admin, hospital admin and clinicians (doctors), scoped per hospital
- • Presigned-URL image upload with analysis-status polling
- • Persistent on-device image cache (LRU) and offline data persistence
- • SaMD (Software as Medical Device) Class 2 certification in progress
- • Assistive tool — does not replace clinical judgment