smartphone urine dipstick reader for pediatric nephrology.
Running live at AIIMS Patna. Three-stage CNN pipeline reads urine dipstick from a smartphone camera — κw 0.87 agreement with lab, 70–80% lower cost per test. APIIC + NIDHI-DST funded · CDSCO TRL 4→6 · HIPAA-architecture · WHO IMCI-aligned · transferable to any LMIC or academic clinical network.
- Three-stage CNNStage 1 object detection on-device. Stage 2 segmentation in secure cloud. Stage 3 classification with confidence.
- Confidence-floor safetyAnything under 70% confidence outputs 'Indeterminate' — never a guess on a child's diagnosis.
- BLE vitals integrationBluetooth BP cuffs + weighing scales stream vitals to the dashboard live.
- PIN anonymisation6-digit PIN per patient; PIN-to-identity registry isolated to the PI only.
- India residency by designPer-tenant CMK · private endpoints · ap-south-1 (Mumbai) only · ICMR 5-year retention.
- REDCap clinical CRFBi-directional REDCap connector for clinical capture and regulatory submission.
- Per-brand subgroup evalModel robustness across Multistix · Combur · Uristix · other brand dipsticks.
Mobile app captures dipstick image under standardised LED lighting. Stage 1 CNN on-device detects the strip. EXIF metadata stripped. TLS 1.3 upload to ap-south-1. Stage 2 segmentation in secure cloud. Stage 3 classification (NEG / 1+ / 2+ / 3+ / 4+) with confidence. BLE BP/weight streams in parallel. REDCap clinical CRF stores anonymised by 6-digit PIN.
↓ tls 1.3
stage 1 detection → stage 2 segmentation → stage 3 classification
↓
confidence floor · pin anonymisation
↓
redcap crf · ap-south-1 only
start with a diagnose.
Two to three weeks. Five to ten lakh. A written scorecard with topology recommendation, cost ranges, and remediation plan. No commitment to build.
request diagnose