Phone cough recordings that tell camp health workers who needs spirometry
TL;DR Coughsort lets a trained health worker record three coughs on a basic smartphone and get a COPD or asthma risk band in under a minute. Anyone in the high band gets booked for a breathing test. CSR lung-health drives, pharma screening camps and hospital outreach teams pay per person screened. It triages and does not diagnose.
India had an estimated 5.53 crore people with COPD and 3.79 crore with asthma in 2016, and chronic respiratory disease caused 10.9% of deaths[1]. The country carried 32% of the world's burden of these diseases with about 18% of its population, and air pollution drove more than half of COPD[1].
Most of it gets caught late. COPD is confirmed by spirometry, which needs equipment and a trained technician and costs ₹600–1,500 a test[2]. Few rural clinics have one, so years of breathlessness get written off as a smoker's cough.
So why not test everyone? It is too costly for a screening drive. A cheap first filter that sorts likely-normal from needs-a-test would let camps spend their spirometry budget on the right people.
What it does, in regulatory terms: screening and triage only. Coughsort flags people who may have COPD or asthma and refers them for spirometry and a doctor. A negative result does not rule out disease. CDSCO's July 2026 software guidance says screening software can't be sold as wellness, and a tool that flags early signs for a clinician is likely Class B[3]. Consumer use without a professional could raise the class.
Here's how it works at a camp or clinic. The worker asks a few questions on smoking, cooking fuel and job, then records three coughs, and a model built on a pre-trained health audio foundation and tuned on Indian recordings returns a risk band. The app works offline and tracks whether flagged people actually got their test.
The seed proposed passive voice analysis. Every accuracy result I found came from deliberate, prompted coughs, and an always-on microphone records relatives who never agreed to it. So Coughsort uses prompted coughs in trained hands. It leads with COPD and asthma because the public TB programme already has a free cough app[4].
The people screened are adults with long exposure to smoke, dust or fumes - smokers, women cooking on wood fires, construction and mine workers, truck drivers. Health workers use the app. CSR foundations, pharma respiratory brands and hospital outreach teams pay, and the leading Indian cough app already counts corporate foundations and a big hospital chain among its partners[5]. It charges ₹100–350 per test[6], well below a ₹600–1,500 spirometry test.
A big tech lab released a health audio model in 2024, trained on about 300 million clips (a third of them coughs) which researchers can request[7]. That cuts the cost of building a cough model by a good chunk. The leading Indian cough app reports more than 4 lakh assessments[6], so Indian buyers have seen the format. And CDSCO's 2026 guidance sets out the class and the Indian validation an AI screening app needs[3].
No published figure exists for cough-based respiratory screening in India. As a ceiling, government TB case-finding screened 2.23 crore people in 2021[8]. At ₹100–350 a test that would be ₹220–780 Cr a year by my arithmetic, but that channel uses a free tool. The reachable market is CSR, pharma and private clinic screening, and nobody publishes those volumes. Adjacent global vocal-biomarker estimates differ about 3.7-fold, so I'd leave them aside.
A Hyderabad cough-AI startup is the closest comparable. Its phone cough tool covers TB and lung health, reportedly holds a Class B licence[2], and is funded by ₹4 Cr of government grants and about US$1.2M of investment[6]. A nonprofit's free cough app holds the public TB channel[4]. A US app counts coughs passively on the phone and keeps no audio. A global drugmaker bought an Australian cough-diagnosis startup in 2022[9]. The open lane is COPD and asthma triage for paid camps.
| Difficulty | Hard (57/100) |
| Starting capital | ₹3.1–5.5 Cr |
| Time to launch | 18–24 months |
Data is the hardest part. You need thousands of cough recordings paired with spirometry, across many phones and sites, gathered under ethics approval. Then comes a Class B licence with an Indian clinical investigation and change control on every model update, which is slow and costly work. The app itself is the easy bit.
Yes. Coughsort is screening software, so it can't be sold as wellness and needs a licence before any camp uses it.
Building this outside India? The categories above usually have a local equivalent, so check with your own regulator.
Don't let AI rule disease in or out. Coughsort screens and refers, a negative result never clears anyone, and a doctor with spirometry makes the diagnosis.
Want to work out the AI layer for Coughsort? Talk to Yash →
Here's what can go south -
Prompted cough triage for COPD and asthma is a credible business, and passive voice detection is not ready. I'd back a founder with clinical research links and the patience for data collection and licensing. If you want a consumer app that listens all day then drop it, because evidence and privacy law both cut against you. Prove first that your model holds up on phones and sites it never trained on.
My estimate is ₹3.1–5.5 Cr, covering data collection with spirometry, model development, the clinical study and a CDSCO licence. The kit breaks it down line by line.
Yes. Screening software can't be sold as wellness, and a tool that flags COPD or asthma for a clinician is likely Class B under CDSCO's 2026 guidance.
Yes, but models trained in one country often don't hold up in another, so plan local validation and check your regulator's rules first.
The free Coughsort validation kit has the line-by-line capital breakdown, the effort scoring, the CDSCO and data-protection checklist, a 14-day validation sprint with pass/fail targets, a 90-day launch plan, 10 interview questions for CSR, pharma and hospital buyers, and 5 assumptions to test. Download the kit →
Thinking of building Coughsort or something like it? The make-or-break bit is planning a Class B validation study and collecting cough data that holds up across phones and sites, and that's the kind of thing I help founders with. If you need a hand, you know where to find me. Talk to Yash →
For illustration only. Names are fictional and ideas hypothetical; sourced facts are real but may be out of date. Nothing here is legal, financial or professional advice, and Yash Khatri is not personally liable for any use of it. Read the full disclaimer.