Patient Case-Taking Software
Ministry of Ayush · Smart Automation · Software
The best-argued statement on the portal hiding behind a three-word title, with a demo the judge participates in — build the conversational history module properly and be upfront that the handwritten prescription OCR is partial, because pretending otherwise is the one thing that will sink you.
What it actually is
In a busy government hospital a doctor gets two to five minutes with each patient, and in that time has to hear the whole story, read whatever papers the patient brought, examine them and prescribe. The history is the part that gets cut, and that is the part that produces the diagnosis. The ask is a kiosk that takes the history from the patient by conversation before they go in, reads their old prescriptions and reports, and hands the doctor a complete structured summary.
What to build
A patient-facing intake platform with four modules as specified: a conversational history engine that interviews the patient by voice in their own language with every question also answerable by tapping, branching adaptively on the chief complaint the way a clinician would and probing along a framework such as SOCRATES, with an extended AYUSH mode capturing the Dashavidha Pariksha parameters the statement enumerates and a red-flag detector that escalates emergency presentations to triage instead of the queue; a document module that scans the patient's prior prescriptions, lab reports and discharge summaries, extracts diagnoses, medications with dosages and investigation values with reference ranges, orders them into a timeline and highlights out-of-range results; a summary generator producing a standard-format physician-ready history the doctor can edit or reject; and a consent and identity layer authenticating via ABHA, taking granular audio-explained consent and pushing the record to the hospital system over FHIR.
Smallest thing that wins the room
Have a judge describe a symptom out loud in Hindi, watch the system ask the right follow-up questions unprompted, then show the structured history appearing on the physician screen with the red-flag path triggering when they mention chest pain with breathlessness.
How crowded this one gets
A guess, projected from the 2025 statements — the last year where both the submission counts and the winners were published.
Quieter than 57% of the 226 · #98 of 226 by expected field
A normal-sized field. Your idea has to be good, not miraculous.
Why: central ministry statements sat below the average.
This is a guess, not a fact
Nobody has published 2026’s numbers yet. This is an analysed estimate from last year’s pattern, so please do not take it as the truth — check the live counter on the SIH portal before you decide anything. The range covers the middle half of likely outcomes, so one statement in two lands outside it. Entry closes at 500 ideas per statement, so no range goes past that — a statement that reaches the cap fills and shuts rather than drawing an unlimited crowd. The model reads only three things a team can see before choosing — software or hardware, the theme, and what kind of body posted it — and those explain about a quarter of the variation in last year’s field sizes (R² 0.25 on held-out statements). Trust the band more than the number, and the ordering more than either. It cannot see how good your idea is, which is the part that actually decides it.
The scores
The number is the shorthand. The line under it is the reason.
Acceptance potential
4/5A strong hidden gem — the title is three words and the specification behind it is enormous, so teams browsing titles will skim past it, the demo is participatory, the two-minute consultation problem needs no justification, and the AYUSH Dashavidha Pariksha mode is a differentiator essentially no competing team will build.
Feasibility
4/5The enabling pieces the statement identifies genuinely exist and are free — Indian-language speech recognition through national language infrastructure, an LLM constrained by a dialogue ontology, and an openly documented ABDM sandbox for the identity and FHIR side — so the conversational core is buildable now, with only the handwritten prescription OCR sitting well below what the statement assumes.
Innovation scope
3/5Four modules, the clinical frameworks, the patient journey and even the failure modes of alternatives are all laid out, but the genuinely open question — how you constrain a language model to elicit history without ever drifting into diagnosis — is not answered anywhere in the statement and is the design problem that matters.
Clarity
5/5At close to twelve thousand characters this is the most thoroughly specified statement on the portal, working through the problem quantitatively, explaining precisely why each existing alternative fails, naming the clinical frameworks for both allopathic and Ayurvedic history, and walking the entire patient journey step by step.
Effort
MassiveMultilingual speech in and out, an adaptive clinical dialogue manager, handwritten multilingual document OCR with entity extraction, timeline construction, clinical summarisation, ABHA authentication and FHIR push, and an accessible kiosk interface for low-literacy users is four large products in one statement.
Demo-ability
EasyThe judge talks to it and it interviews them back, which puts them inside the product rather than watching it, and the structured history appearing on the doctor's screen closes the loop in a way anyone in the room immediately understands.
In its favour
- Green flag: A three-word title over a twelve-thousand-character specification is a genuine competitive advantage — many teams filter statements by how substantial the title sounds and will never open this one
- Green flag: The problem is quantified for you with consultation-time figures and OPD volumes, so the motivation is evidence-backed rather than asserted
- Green flag: The dual-mode requirement that every question be answerable by speaking or tapping is both an accessibility win and a demo safety net — if speech recognition struggles on stage you fall back to touch without breaking the flow
- Green flag: The Dashavidha Pariksha mode is a real differentiator: an Ayurvedic history framework has a defined parameter set that no general-purpose intake product implements
Against it
- Red flag: Handwritten Indian prescriptions are among the hardest OCR targets in existence, and the statement treats reliable extraction from them as a solved enabling technology when it is not — scope this module honestly or it will collapse in front of a clinician
- Red flag: A conversational medical system must never appear to diagnose, and an unconstrained language model asking follow-up questions will drift there naturally — the ontology constraint and the physician-edits-before-saving design are safety requirements, not features
- Red flag: Hospital OPDs are loud and crowded, and speech recognition tested in a quiet lab behaves very differently at a registration counter with five hundred people in the hall
- Red flag: Four modules with multilingual speech, document AI, summarisation and ABDM integration is far more than one team finishes — decide which module is your demo and let the others be honest stubs
What you will be writing
- IndicWhisper / Bhashini ASR for multi-accent Indian speech
- ontology-constrained LLM dialogue manager
- handwritten prescription OCR with clinical entity extraction
- ABDM sandbox ABHA authentication and FHIR push
- icon-driven dual-mode kiosk UI
- red-flag symptom rule layer over the dialogue
- Clinical documentation
- Conversational AI
- Digital health infrastructure
Prior art to read before you start
conversational clinical history elicitation · medical document digitisation and timelining · multilingual accessible kiosk interface
Analysed by Claude Opus. Every score above is a judgment call with its reasoning attached — kindly cross-check this against the official statement on the SIH portal before your team commits to it.