ElyCia Lifesciences Pvt. Ltd. · Eureka! 2026 Finals · EU2698236
A handheld, non-sputum breath test for tuberculosis triage.
Dr. Ruwise E. A.
Founder, Executive Director & Principal Investigator
Problem
Many patients cannot produce sputum: children, people living with HIV, early disease
Primary health centres: often no X-ray and no lab
Screening stalls where the burden is highest
Source: WHO Global Tuberculosis Report 2025
Solution · how the device works
TBot reads a patient's breath in minutes: no sputum, no X-ray, no lab. It pairs a multi-sensor breath array with a TB-marker-specific electrochemical cartridge, capturing both the broad breath signature and the specific markers the TB bacterium itself releases.
What makes TBot different
A disposable, TB-marker-specific electrochemical cartridge gives molecular specificity a generic e-nose cannot, inside a reusable, low-cost handheld reader with UV-C self-sterilisation.
Reusable reader + swappable disease-specific cartridge. The same platform extends to other respiratory diseases after TB.
Two decades of breath-VOC science, low-cost printed-sensor manufacturing and on-device AI now make a cheap, specific, field-deployable test possible, and no breath-based TB test has been cleared anywhere in the world yet. The category is open.
How it works · one breath, four simple steps
Disposable mouthpiece; a filter blocks germs. 1–2 min, no training.
Capnography isolates the deep-lung fraction where TB markers concentrate.
Phase-III CO₂ plateau = the only air sent to the sensors
Multi-sensor array plus the TB-marker EC cartridge read the sample.
Edge-AI returns TB / not-TB; UV-C sterilises the path between patients.
Virtual test run · simulation
System architecture
Single-use mouthpiece · bacterial/viral filter · one-way valve
Heated above dew point; humidity held at a set-point that stabilises both sensor types
In-array breath-phase & reference: flow · CO₂ capnography · temperature · humidity · pressure
Detects the four mycobacterial marker markers (lead: the lead marker): molecular specificity a generic e-nose cannot give. Developed as a standalone module (CSIR-NIIST); not built into the MSA embedded system in the current phase.
Capnography-synchronised alveolar windowing · environmental compensation · ML pattern classifier
refer if positive
Between-patient decontamination: disposable consumables + UV-C decontaminable flow path.
Broad VOC pattern + TB-specific markers, fused on-device. Capnography windows the deep-lung fraction; conditioning holds humidity where MOX and electrochemical sensors are stable.
ElyCia LifesciencesThe science & what's new
TB releases four specific marker markers into breath. A proprietary recognition film captures them. Validated in solution at CSIR-NIIST for the lead marker, the lead marker.
limit of detection
linear range
Characterised by XRD / SEM and cyclic + differential-pulse voltammetry.
Electrochemistry needs a liquid; breath is a gas, so every prior group had to trap breath into liquid first. Our electrolyte film provides the liquid and concentrates the signal in a single step. No trapping, no condensation.
reaches children, HIV & extrapulmonary cases
names a TB molecule, not just a pattern
no lab, X-ray or connectivity
3 patents + PCT; FTO cleared, 9 jurisdictions
EC cartridge · 3D exploded view · drag to rotate
Single-use cartridge, reusable reader. DPV/SWV resolves each marker at its own potential against an internal redox reference; the control electrode is subtracted.
AI / ML · two channels, one decision
The MSA learns the breath-pattern shift · the EC module scores the TB marker · a fixed layer fuses both. GC-MS is the reference truth both models train against.
alveolar gating · R/R₀ baseline · per-litre volume-norm · humidity control · PCA drift-correction → BME688 step-response (slope · AUC · step-ratios · EMA) · PID peak/AUC · capnogram etCO₂ / phase-III
Calibrated fusion of p₁ + p₂ into one score, with a confidence estimate and an explicit indeterminate (repeat) zone rather than a forced call.
Task C: deterministic validity gating guards every sample before fusion.
baseline + NIP control-electrode subtraction · internal-redox (ferrocene-class) referencing → per-marker DPV/SWV peak current · peak potential · integrated charge · redox kinetics
Unit economics
Most diagnostics are costly because sensing uses antibodies or DNA. We detect small breath molecules, so the recognition chemistry is bulk, printable and cheap. The cost sits in plastic and packaging, not biology.
Reusable handheld: BOM target <₹50,000, sale ~₹1,00,000. Priced near cost to drive placements; profit compounds on the cartridge annuity.
Indicative modelled costs at scale, not quoted prices. Biodegradable mouthpiece and reduced-foil packaging are committed design goals.
Market · one platform, many indications
beachhead
Global respiratory dx & monitoring: multi-billion-USD, chronic + recurring
Large global screening markets: early detection & monitoring
Each new disease = a new disease-specific cartridge + tuned classifier on the same reader. Both filed specifications claim extension to other diseases by substituting the marker panel. No new hardware.
combined breath-diagnostics opportunity globally, addressable on one reader with disease-specific cartridges
Adjacent-market figures are indicative global estimates; TBot enters each indication only after TB validation. Extensions are per the filed provisional specifications.
Business model & revenue
Reusable handheld: sensor platform, capnography, display, UV-C, software
Single-use TB-marker cartridge: the recurring margin engine
NTEP / Nikshay-ready dashboards and fleet management
>70%
at scale (₹135 vs ~₹40 COGS)
Installed base
× tests per device / yr
Low
institutional / government sales
≈ ₹293 Cr (US$35M) revenue by the 5th commercial year (base case)
Development roadmap · click a track or phase
A 60-person breath study answers the biggest question by month 6, without waiting for a prototype, and sets the targets for the EC strip and the reader. Fusion and the pivotal trial stay gated on results.
Go-to-market
Decides & buys
Sets the national testing algorithm and procures at scale, as it did for Truenat & Xpert.
Deploys
Roll out community active case-finding once a technology is adopted nationally.
Operates
Run TBot at the point of care, designed for minimal training.
Competitive landscape
| Segment | Players | Why it isn't TBot |
|---|---|---|
| Molecular confirmatory India, NTEP | Xpert MTB/RIF, Truenat, AG Chitra | Sputum + lab; confirmatory. TBot feeds them, never replaces |
| Imaging / acoustic triage | qXR, Genki (CAD-CXR); Swaasa (cough AI) | Needs an X-ray unit, or reads only a cough pattern; not pathogen-specific |
| International breath platforms | Owlstone (FAIMS); Breathomix (SpiroNose) | Built for oncology / asthma-COPD, not TB; centralised or MOX-only |
| Breath-VOC TB & patent-stage | TSI-3000 (IMS); VIT, G. Rani, CVR filings | Non-specific total-volatilome, or applications; none a cleared product |
No player holds TBot's full combination: a TB-marker-specific EC channel + MOX array + capnography-gated alveolar sampling + disposable cartridge. No breath-based TB test is cleared anywhere; FTO cleared across 9 jurisdictions.
Competitive benchmark · the test TBot must match
AI X-ray: multi-country head-to-head of 5 CAD products (medRxiv 2024). TBot: Monte Carlo of 2,400 trials from literature-anchored priors, not clinical data. Median 54–61%, 80% range 37–87%, driven mainly by how many TB patients emit the markers, which Track A measures.
Phase 2 study (n ≈ 270): every participant also gets an AI X-ray read, so the first accuracy result is a direct comparison. The previous-TB subgroup is pre-registered. To clearly beat X-ray, the marker cartridge alone must reach AUC ≈ 0.88–0.90.
IP & regulatory pathway
Trademark
"TBot", App 7219027, Class 10, filed
Freedom to operate cleared, 9 jurisdictions (TIPS@SCTIMST-TIMed)
A new IVD with no CDSCO predicate device, reviewed by ICMR MedTech Mitra (51st TAC). Building to the conservative Class C standard, seeking Class B where risk justifies.
Test Licence: manufacture the device for evaluation
Clinical Performance Evaluation: permission for human-subject testing
Manufacture permission for a no-predicate device
Manufacturing Licence: Central Licensing Authority; ISO 13485
Team
Founder, Exec. Director & PI
Orthopaedic surgeon & device innovator. Owns architecture, IP & coordination; sole inventor on all patents.
Co-founder · Regulatory & Quality
Pharmacologist. Leads regulatory affairs, QMS, risk documentation and the CDSCO interface.
Co-founder · Analytical Science
MD Forensic Medicine, DM Toxicology. Analytical toxicology & breath-biomarker interpretation.
Sensing chemistry (Dr Rakhi R.B)
MSA engineering (MeitY)
Prototyping, EMI/safety, DFM
The ask · 24 months · three tracks, then the exploratory study
Raising ₹2.1 crore seed equity; the remaining ₹1.2 Cr of the ₹3.3 Cr plan is targeted from non-dilutive grants (ICMR, BIRAC, PRIP). The first ₹31 L answers the biggest question by month 6.
A · M6 · TB biomarkers measured in the breath of 60 patients and controls by GC-MS, before any prototype is built
B · M15 · A TB biomarker EC strip validated for gas-phase reading at the breath levels Track A measured
C → Ph 2 · M24 · An interface-frozen MSA reader and a locked triage model from the ~270-participant exploratory study
Alongside the seed · co-fund M1–24
This ask · 3 tracks + exploratory study
Equity + DoP/ICMR · fusion MVP + pilot
Equity + clinical · pivotal, MD-9, launch
Milestones bought: breath-marker readout (M6) · interface freeze (M12) · EC gas-phase (M15) · MD-12/24 (M15) · study locked + first sens/spec (M24)
Thank you
A sub-₹150, sputum-free TB triage test in every primary health centre.
Dr. Ruwise E. A. · Elycia Lifesciences Pvt. Ltd.
ruwiseea@elycialifesciences.com · +91 70127 49565
Appendix A1 · AI / ML · validation & calibration
Subject-level nested CV · leave-one-site-out; metric = specificity @ 90% sensitivity (WHO TPP-3)
Beats a confound-only benchmark (age · sex · BMI · fever · HIV · diabetes · site); algorithm locked for v1 (IEC 62304)
n ≈ 270 → feature discovery + model locking (3–6 features) · pivotal validation ~1,000–1,200
MSA: onboard zero/span cartridge (carbon scrub + permeation) · reference-gas drift span
EC: internal-redox standard + per-lot carrier (EEPROM/NFC) · strip authentication
EC working range 0.1–110 µM (LOD ≈ 0.1 µM); out-of-range flagged indeterminate
Appendix A2 · How much breath is enough
1–2 min
adult · ~15 normal breaths
2–3 min
small child · smaller breaths
Log-scale design calculation: 1 µL gel, ~20% capture, read at 3× the detection limit. Not yet measured.
Appendix A3 · EC sensor · layer architecture
The electrolyte film supplies the electrolyte and captures gas-phase markers in one step: no trapping, no condensation. recognition film binds the TB biomarkers (validated in solution for the lead marker); DPV/SWV resolves each marker against an internal redox reference.
Single-use cartridge, reusable reader. The strip carries a per-lot calibration and authentication.
Appendix A4 · Q&A · evidence
Lesson: pattern-only devices lose specificity when controls are sick. Targeted chemistry is promising but small and unreplicated.
Sick controls by design. ~165 of ~270 participants are symptomatic and TB-negative, not healthy volunteers.
Two independent channels. A broad breath pattern plus a TB biomarker mycobacterial panel, so they fail in different ways.
Deep-lung air only, with a Repeat zone instead of a forced call.
Pre-registered go / no-go against GC-MS and Xpert / culture; results reported per subgroup.
Saktiawati et al., EBioMedicine 2019 · Coronel Teixeira et al., PLOS ONE 2023 · Jeppson et al., medRxiv 2024 (not peer-reviewed). WHO triage TPP 2014 (≥90% / ≥70%); WHO screening TPPs updated Aug 2025.
Appendix A5 · Q&A · confounders
Appendix A6 · Q&A · competition and adoption
| Tool | Status | What it needs |
|---|---|---|
| AI chest X-ray | WHO-recommended; qXR screened >7 M for TB in FY24–25 | X-ray unit, power, radiation safety |
| Cough-sound AI (Swaasa) | CDSCO Class B SaMD; 90% / 85% in validation | Phone; pattern only |
| Tongue swab NAAT | WHO-recommended sample (Mar 2026) | Molecular instrument, cartridge |
| Breath-aerosol PCR (Avelo) | Late-stage; 71% / 92% | Molecular lab step |
| TBot | Pre-clinical (TRL 3) | Nothing but the reader: no radiation, sample or lab |
people screened in the 2024–25 100-day TB Mukt Bharat campaign; ~500 AI handheld X-ray units deployed, 1,000 more planned
The pre-filter before X-ray and NAAT in camps, homes and PHCs without imaging. It complements X-ray rather than replacing it.
The route Truenat took (validated at NIRT, AIIMS, NITRD, JALMA).
WHO TB guidelines Module 3 (2025, Mar 2026 update) · Qure.ai FY24–25 impact report · Yellapu et al., Sci Rep 2023 · TAG Pipeline Report 2025 · MoHFW / PIB 2025.
Appendix A7 · Q&A · risk and de-risking
Molbio (Truenat) became India's first medtech unicorn in 2022; Qure.ai raised a US$65 M Series D in 2024. Both scaled after validation and programme adoption.
Seed is matched by non-dilutive money: BIRAC, ICMR, PRIP and India Health Fund–type grants.
Appendix A8 · Development plan · overview
| Stage · led by | Question it answers | Output | Direct cost | Go / no-go |
|---|---|---|---|---|
| M0 · now TRL 3 · ElyCia + CSIR-NIIST | Does the chemistry work? | Lead marker on recognition film in solution (LOD ≈ 0.1 µM, 0.1–110 µM); IP filed | ₹4.5 L spent | — |
| A · M1–6 Clinical validation · ElyCia | Are the markers in TB breath, and at what level? | n = 60 at a Delhi TB centre, GC-MS lab: emitter share, breath levels, AUC against symptomatic controls; no prototype needed | ₹0.31 Cr | ≥ 80% emit ≥ 2 markers and AUC ≥ 0.80 → go; 50–80% → re-scope; < 50% → stop the cartridge |
| B · M1–15 EC TB biomarker strip · CSIR-NIIST | Can the strip read them from humid breath? | Per-marker vapour detection at Track A levels; cross-reaction grid; confounder challenges | ₹0.32 Cr | ≥ 3 markers read at breath levels; tobacco does not raise the anchor markers |
| C · M1–12 MSA reader · engineering partner | Does the reader deliver deep-lung air reliably? | Gated reader with trap, heated path and UV-C; breath volume set from Track A | ₹0.60 Cr | Interface frozen; study units built |
| Ph 2 · M13–24 Exploratory study · ElyCia | Does it work in sick patients? | n ≈ 270 at 2 NTEP-linked sites, both channels + GC-MS reference · expected AUC 0.80–0.91, specificity 50–72% at 90% sensitivity (±7 points) | ₹0.85 Cr | Marker AUC ≥ 0.85 and fused specificity ≥ 60% → proceed; < 50% → change course |
| Ph 3 · M25–36 Fused beta · ElyCia + partners | Does the whole device hold up? | Locked model and confounder logic in an integrated pilot | ₹1.6 Cr | ≥ 70% specificity at 90% sensitivity; ≤ 10% first-attempt Repeat → design freeze |
| Ph 4–5 · M37–60 Pivotal + licence · ElyCia + AMTZ | Does it meet WHO targets at scale? | CDSCO evaluation at 4–6 sites, n ≈ 1,000–1,200; results usually drop 5–10 points, so Phase 2 should show ≥ 75–80% specificity | ₹5.8 Cr | Endpoints met → MD-9 licence, NTEP entry |
₹2.08 Cr direct (Tracks A–C and Phase 2) + ₹1.23 Cr team, regulatory, IP and contingency. Seed ₹2.1 Cr + grants ₹1.2 Cr. Running the tracks together saves about six months; B and C carry a month-6 checkpoint, so about ₹1.0 Cr is committed before the first gate.
Month 6: first real breath data on the four markers. Month 15: the strip reads them from breath. Month 24: first accuracy in symptomatic patients, the step before a Series A.
If Track A shows few emitters, the cartridge pauses and B and C are re-scoped: condensate mode, a narrower intended use, or the reader as a platform for other markers. The filed IP and the gated breath interface keep their value either way.
Appendix A9 · Track A · initial clinical validation (M1–6)
| Group | n | Definition |
|---|---|---|
| Confirmed TB | 25 | Xpert Ultra or culture positive; sampled before or within 24 h of starting treatment |
| Symptomatic, TB-negative | 25 | Presumptive TB; Xpert and culture negative |
| Healthy controls | 10 | No symptoms, non-smokers |
At least 6 smokers in each of the first two groups, analysed separately. Observational: ethics approval and CTRI registration, no device licence.
Deep-lung breath only, verified by CO₂ (≥ 90% of end-tidal); 1 L on a Tenax TA tube; paired room-air tube
Thermal-desorption GC-MS with authentic standards; detection limits set before the first sample; analyst blinded
M1 method + ethics · M2–4 recruitment · M5–6 analysis and decision
Share of TB patients emitting ≥ 2 markers → expected fused specificity at 90% sensitivity (80% range below)
Monte Carlo, 2,400 trials from literature-anchored priors: a simulation, not data.
| TD-GC-MS analysis · 238 runs | ₹11.90 L |
| Tubes, sampling kit, capnometer, CO monitor | ₹5.95 L |
| Research coordinator · 6 months | ₹3.00 L |
| Marker standards, second-source check | ₹1.40 L |
| Untargeted workup (exploratory) | ₹1.00 L |
| Ethics, reimbursement, culture | ₹1.50 L |
| Statistics, report, manuscript | ₹1.50 L |
| Shipping and sundries | ₹0.50 L |
| Contingency (15%) | ₹4.01 L |
| Total | ₹30.76 L |
About ₹21 L if the GC-MS lab joins as co-investigator and charges consumables only.
Appendix A13 · Phase 2 exploratory study (M13–24)
Appendix A14 · Phases 3–5 · fused beta to market
| ISO 13485 | QMS, design controls, DHF / DMR |
| ISO 14971 | Risk file; confounder controls as risk controls |
| IEC 60601-1 / -1-2 | Electrical safety, EMC |
| IEC 62304 | Software lifecycle; algorithm locked for v1 |
| IEC 62366 | Usability; pre-test protocol, Repeat messages |
| IEC 62471 | UV-C photobiological safety |
| ISO 10993 | Biocompatibility of mouthpiece and strip |
| ISO 14155 | Clinical investigation practice |
Cartridge docks without changing the frozen airflow · locked combiner · integrated pilot · target ≥ 70% specificity at 90% sensitivity · design freeze
CDSCO evaluation, n ≈ 1,000–1,200 at 4–6 sites · likely Class C IVD · MD-28 (no-predicate) → MD-9 manufacturing licence · ISO 13485 line at AMTZ
Appendix A15 · Programme value
| Approach | Referred to NAAT | Cost | TB missed (of 150) |
|---|---|---|---|
| NAAT for everyone | 1,000 | ~₹10.0 L | 0 |
| TBot at 90% / 51% | ~550 | ~₹6.9 L (−31%) | 15 |
| TBot at 90% / 72% | ~370 | ~₹5.1 L (−49%) | 15 |
Assumes 15% TB among symptomatic people screened, NAAT ≈ ₹1,000 and TBot ₹135 per test.
In camps and PHCs the real comparison is usually no test at all. There, TBot finds ~135 of every 150 TB cases per 1,000 people, without X-ray, sputum or lab.
Up to half the NAAT spend saved per 1,000 screened, and cartridges directed to the people most likely to have TB.
Every screened person is a ₹135 cartridge: the recurring revenue engine behind the ~₹293 Cr year-5 base case.
Appendix A16 · Risk register · click a tab
H high M medium L low · L = likelihood, I = impact
ElyCia Lifesciences