User
Suite 02 · Chest X-ray screen
📸 For a clear photo
• Chest (PA) X-ray only — not other body parts
• Place the film on a lightbox / bright screen
• Fill the frame, keep it straight, avoid glare/reflections
• Screening aid only — not a diagnosis
• Place the film on a lightbox / bright screen
• Fill the frame, keep it straight, avoid glare/reflections
• Screening aid only — not a diagnosis
The photo is analysed over a Cloud AI and not stored. Confirm any positive with sputum/NAAT testing (NTEP). Screening aid, not a diagnosis.
Research instrument · AI
Find TB early — every breath matters
Screen lung sounds, cough, breathing and a chest X-ray with the phone already in your pocket — no extra hardware. TB F.I.R.S.T helps you find early signs of TB and other heart & lung problems, and refer at the right time.
Screening suites
🫁
PulmoScope
Lung sound · Percussion · EchoLat
Open
🫀
CardioScope
Heart sound · Rhythm · JVP
Open
❤️
VascAge
Fingertip PPG · risk factors
Open
PulmoScope · Percussion
Phone percussion
Map percussion notes across the chest against a within-patient reference.
Technique · step-by-step
How to tap the chest with the phone
The phone listens to the tapping sound for you — you don't hold it in the air.STEP 1Press the phone flat and firm on the skin, screen up, over the spot you want to check — hold it there with one hand.
STEP 2With the middle finger of your other hand, tap the chest firmly right next to the phone (about a finger-width away), 5–6 times, roughly one tap a second.
STEP 3Tap a normal, healthy spot first — this is the reference — then tap the spots you are worried about. The app compares them for you.
Quiet room. Tap with the same strength each time. Keep the phone still while you tap.
1) Tap a normal zone → record it as REFERENCE. 2) Tap each suspicious zone → record. App compares to the reference.
not done
recording
done
selected
reference
duller
Tap a zone on the body above.
Quiet room. Phone mic against skin. Tap firmly with ONE finger, 5–6× ~1s apart.
RESULT vs reference (heuristic ⚠)
—
Peak freq—Hz
Centroid—Hz
Low:High—
Decay—ms
Suite 03 · VascAge
Fingertip PPG + CVD risk
Estimates pulse, heart rhythm and cardiovascular-risk indicators from fingertip PPG.
Technique · step-by-step
How to record fingertip PPG
Cover the phone's rear camera lens AND flash together with the pad of the index fingertip — both fully covered, no gap.PRESSURELight and steady. Rest the finger on; pressing hard squeezes the blood out and kills the signal.
STILLRest the hand on a table and hold dead still for the whole capture. The flash stays on; no flash → sit in bright daylight.
Wait until the live trace shows regular pulse waves before trusting the heart rate.
Rest the fingertip lightly over the rear camera + flash
Cover the REAR camera + flash with a fingertip. Use bright light if no torch.
PulmoScope · EchoLat
Side-to-side breath comparison
Records each side sequentially and compares the breath sounds between the two sides.
Technique · step-by-step
How to record breath sounds
Seated in thin clothing, leaning slightly forward. Press the phone's bottom (mic) edge flat and firm against the lung base on the back.SAME SPOTRecord Right first, then Left at the mirror-image spot — the app compares the two sides.
BREATHEPatient takes slow, deep breaths through an open mouth through each 15-second recording.
Quiet room. Don't slide the phone during a recording.
not donerecordingdone
Tap a lung base on the figure (mic on posterior base; patient breathes deeply).
Record both sides to compare.
Suite 01 · CardioScope
Cardiovascular exam
On-demand heart-rhythm (phone accelerometer on the chest) + neck / JVP pulsation (camera).
Technique · step-by-step
How to read rhythm (seismocardiography)
Patient lying down or sitting still. Place the phone flat, screen up, on the centre of the chest (lower breastbone). Hands off, breathe normally, stay quiet for the countdown — the phone feels each heartbeat as a tiny vibration.Best lying down. Phone flat on the sternum, thin clothing or bare. No talking or moving during capture.
Lay the phone flat, screen up, on the centre of the chest (breastbone)
Phone flat on the chest, then Start.
—
Heart rate—bpm
Mean RR—ms
RR variability (SDNN)—ms
Signal quality—
Irregularity is a screen, not a diagnosis. Confirm any "irregular" with ECG / Holter.
Suite 02 · PulmoScope
Lung-sound screen
Records lung sounds with the phone mic and a cloud AI model screens for normal vs abnormal (crackles/wheezes). Tap and record 6 points on the back — the app averages them into one result.
Technique · step-by-step
How to record lung sounds
Seated in thin clothing, leaning slightly forward, quiet room. Tap a point on the body image below, press the phone's bottom (mic) edge firmly on the skin at that spot, ask the patient to breathe deeply through the mouth, then press record. Do all 6 points.LISTENRecords each point to a .wav and runs the cloud AI model. Averaging 6 points gives the most reliable normal/abnormal screen.
Thin clothing or bare skin. No talking during capture.
Tap a point on the back image → record 15 s of deep breathing. Do all 6 for the most reliable screen.
not done
recording
done
selected
Tap a point on the back above.
Quiet room, phone mic on skin, patient breathes deeply through the mouth.
—
AI model (multi-point average).
Mean abnormal probability—
Points recorded0/6
The AI lung screen is a screening tool, not a diagnosis. Confirm positives clinically. Threshold is provisional — calibrate with your own data.
Suite 02 · PulmoScope
Cough counter
Records 60 seconds with the phone mic. A cloud AI model (YAMNet) detects coughs and returns a total count and rate per minute. No audio is stored — only the number.
Technique · step-by-step
How to count coughs
Quiet room. Patient sits an arm's length from the phone. Press START and let the patient sit naturally for 60 seconds — do not ask them to force-cough. The app stops automatically and runs the cloud AI.COUNTThe YAMNet model detects and counts cough sounds. Audio is sent and only the number returns — nothing is stored.
Press START and wait 60 s. It stops automatically and shows the result.
Quiet room. Phone an arm's length from the patient.
Suite 02 · PulmoScope
Lung sound type
Records 15 seconds at one chest point. A cloud AI analyses the sound for crackles (pneumonia/fibrosis/edema) and wheeze (asthma/COPD/obstruction). No audio is stored.
Quiet room. Press the phone mic firmly on the chest, ask the patient to breathe deeply. Press START — recording auto-begins once the room is quiet and stops at 15 s.
Press firmly on the chest.
Suite 02 · PulmoScope
Respiratory rate (WHO fast-breathing)
Hold the phone near the patient's nose/mouth (not the chest) and ask them to breathe audibly. Press START — after a quiet moment it records 30 s and a cloud AI computes the rate. (In a noisy room, use the tap method below.)
Hold the phone near the nose/mouth.
Or manually: tap for every breath (60 s).
0
Suite 02 · PulmoScope
Blow test (Forced Expiratory Time)
Patient takes the deepest breath possible, then blows into the phone mic as hard and as long as possible until no more air comes out. The app times the audible breath. ≥6 seconds suggests airflow obstruction (COPD/asthma).
Hold the phone near your OPEN MOUTH. Press START, then blow out hard and long through your open mouth right after the beep, until no air is left.
Hold the phone close to the mouth.
Suite 02 · PulmoScope
Single-breath count test
Patient takes the deepest breath possible, then counts out loud "one, two, three…" at about two per second in a normal voice, until they must breathe again. The app measures how long they sustain it and estimates the count. Under 25 suggests reduced respiratory reserve.
Hold the phone close to the mouth. Press START, then after the beep take a deep breath and start counting.
Hold the phone close to the mouth.
Suite 02 · PulmoScope
Maximum phonation time ("Ahh" test)
Patient takes the deepest breath possible, then says "ahhhh" in a steady voice for as long as possible on one breath. The app times it. Under ~10 seconds suggests reduced respiratory/laryngeal function.
Hold the phone close to the mouth. Press START, then after the beep take a deep breath and start "ahhh".
Hold the phone close to the mouth.
TB F.I.R.S.T
About TB F.I.R.S.T
You're not logged in. Open Cardio, Pulmo, or Vasc to log in.
TB F.I.R.S.T turns the phone already in your pocket into a gentle first check for your heart and lungs — heart sounds, lung sounds, rhythm, pulse, and a chest X-ray read — wherever you are, not just at a hospital. It is here to help you notice things early and know when to see a doctor. A caring screening companion — not a diagnosis.
Why we built this
In community medicine we care for whole communities, not only the people who reach a hospital. Too often a heart murmur, a quiet wheeze or an early TB shadow is found late — simply because a trained ear or an X-ray reader wasn't nearby. TB F.I.R.S.T puts a helpful first look into anyone's hands, so more of these are noticed in time.
Built on real evidence
Every model here was trained and tested on large, expert-labelled datasets, and each result is shown honestly with its performance. AI analysis runs securely in the cloud, works offline when needed, and your recordings are kept — only with your consent — to keep making these tools better for everyone who comes after you.
Test performance & methods
AI models — performance
| Model | Dataset | AUC | Sens | Spec |
|---|---|---|---|---|
| Heart sound (CardioScope) | PhysioNet/CinC-2016 | 0.996 | 98.3% | 95.6% |
| Murmur (CirCor) | CirCor DigiScope | 0.875 | 72.8% | 93.3% |
| Lung sound (normal/abnormal) | ICBHI-2017 | 0.87 | 63.2% | 100% |
| Crackle | ICBHI-2017 | 0.671 | 64.1% | 61.3% |
| Wheeze | ICBHI-2017 | 0.734 | 63.7% | 70.7% |
Heart-sound and murmur models were trained on PhysioNet/CinC-2016 (3,153 recordings, 764 subjects) and CirCor DigiScope (3,003 recordings). Lung, crackle and wheeze models were trained on ICBHI-2017 (6,898 cycles, patient-level split, MobileNetV2 with SpecAugment): crackle points to pneumonia/fibrosis/edema, wheeze to asthma/COPD/obstruction. Figures are on held-out test sets; models are noise-augmented for real-world robustness.
Chest X-ray — findings
Screens a chest X-ray photo for tuberculosis and other heart/lung changes. Trained with transfer learning (MobileNetV2) on large, expert-labelled TB X-ray datasets (Shenzhen + Montgomery + Qatar DB). WHO recommends computer-aided detection for TB screening where radiologists are scarce. Screening only — confirm any positive with sputum/NAAT per NTEP.
| Metric | Chest X-ray |
|---|---|
| Training data | TB CXR (Shenzhen + Montgomery + Qatar DB) + VinDr-CXR |
| Method | MobileNetV2 transfer learning |
| Internal test AUC | TB ~0.99 · Cardiomegaly 0.96 · Effusion 0.96 · Pneumonia 0.84 · Consolidation 0.93 · Nodule/mass 0.94 · Pneumothorax 0.83 · Fibrosis 0.92 · Pleural thickening 0.95 |
For the clearest read, photograph the film straight and without glare. Screening aid — not a diagnosis.
Signal-analysis tests (validated acoustic measures)
| Test | Measures | Validation |
|---|---|---|
| Cough counter | Coughs + rate/min (YAMNet) | Detects cough events; no audio stored |
| Breaths/min (RR) | Resp. rate → WHO fast-breathing | ~0.2–0.8 breaths/min MAE vs RIP belts |
| Blow test (FET) | Forced expiratory time | ≥5 s: Sens 94% / Spec 81% (obstruction) |
| Breath count (SBCT) | Single-breath count | r≈0.55 vs FVC; <30 flags support (COVID) |
| Ahh time (MPT) | Max phonation time | Correlates FVC; inter-rater 0.94–0.95 |
These are deterministic signal-processing measures (not AI classifiers), use published clinical cutoffs, and run on the phone itself when offline. Screening only — not a diagnosis.
Cardiac timing (S1-S2 systolic interval)
CardioScope measures the S1-S2 systolic interval from the phone recording and reports it corrected for heart rate. The healthy reference was derived from EPHNOGRAM (PhysioNet): 69 simultaneous ECG-phonocardiogram recordings from 24 healthy adults.
| Metric | Cardiac timing |
|---|---|
| Reference dataset | EPHNOGRAM (PhysioNet v1.0.0) |
| Recordings / subjects | 69 rec · 24 healthy adults |
| Healthy band (@60 bpm) | 270–337 ms |
Rate-corrected S1-S2 values outside 270–337 ms are flagged for clinical correlation. Screening aid — not a diagnosis.
Rhythm / AF screen (VascAge)
VascAge screens for an irregular pulse (AF pattern) from the camera pulse (PPG). The validated McManus/Chon method (RMSSD/mean + Shannon entropy) showed 96.2% sensitivity and 97.5% specificity for AF in a 76-patient study. Screening only — confirm any positive with an ECG.
| Metric | Rhythm screen |
|---|---|
| Method | RMSSD/mean + Shannon entropy |
| Sensitivity / Specificity | 96.2% / 97.5% |
Motion or poor signal can cause false results — record still, in good light. Screening aid — not a diagnosis.
Other modules
EchoLat (lung asymmetry), Percussion, Rhythm (seismocardiography) and JVP add extra context alongside the main checks. Use them together with the other results, and confirm anything important with a clinician.
Before you act on a result
TB F.I.R.S.T helps you see clearly — your doctor helps you act. If something is flagged, treat it as a kind nudge to get checked, not a diagnosis, and share the result with a healthcare professional who can guide you.
Legal
TB F.I.R.S.T
Terms & Conditions
Last updated: July 2026. Please read these terms before using TB F.I.R.S.T. By ticking the consent box or using the app, you agree to them.
1. What TB F.I.R.S.T is
TB F.I.R.S.T is a smartphone screening and research tool that records heart and lung sounds and runs AI models and signal-processing to flag sounds that may be worth a clinician's attention. It is provided for research and educational use.
2. Not a medical device or diagnosis
TB F.I.R.S.T is not a diagnostic device and is not a substitute for a doctor, stethoscope, ECG, echocardiogram, or any clinical test. Results are indicative only and may be wrong (false positives and false negatives). Never start, stop, or change any treatment based on this app. Any positive or concerning result must be confirmed by a qualified clinician.
3. Not for emergencies
Do not use TB F.I.R.S.T in a medical emergency. If you have chest pain, severe breathlessness, fainting, or any emergency, contact local emergency services or go to the nearest hospital immediately.
4. Eligibility & consent
You must be 18 years or older to use TB F.I.R.S.T yourself. A recording may be taken for a person under 18 or a dependent adult only by a parent or legal guardian who consents on their behalf. By using the app you confirm you meet these conditions.
5. Proper use
Use the app only for lawful, personal, educational, or approved research purposes. Do not misuse it, attempt to reverse-engineer it, upload recordings of people who have not consented, or rely on it for clinical decisions about others.
6. Your data
Recordings and the details you provide are handled as described in the Privacy Policy. By using the app you consent to that handling, including use of de-identified data to build and validate screening models and for research publication.
7. No warranty & limitation of liability
The app is provided "as is", without warranty of accuracy, availability, or fitness for any purpose. To the maximum extent permitted by law, the investigator and contributors are not liable for any loss, harm, or damages arising from use of, or reliance on, the app or its results.
8. Changes & governing law
These terms may be updated from time to time; continued use means acceptance of the updated terms. These terms are governed by the laws of India, with jurisdiction in Telangana.
9. Contact
Dr. Jaideep Rao M., MBBS, MD (Community Medicine), Government Medical College, Maheshwaram, Telangana, India.
TB F.I.R.S.T
Privacy Policy
Last updated: July 2026. This policy explains what TB F.I.R.S.T collects, why, and your choices. It is written to align with India's Digital Personal Data Protection Act, 2023.
1. Who we are
TB F.I.R.S.T is a research project led by Dr. Jaideep Rao M., MBBS, MD (Community Medicine), Government Medical College, Maheshwaram, Telangana, India — the data fiduciary responsible for your information.
2. What we collect
• Account: your email address (via Google or email sign-in).
• Details you enter: name, age, sex, height, weight, smoking status, any known heart/lung condition, current symptoms, optional phone number, and your role.
• Recordings: short heart and lung sound clips (.wav) you choose to record.
• Results: the screening outputs computed from your recordings (e.g., rate, rhythm, probabilities, quality).
We do not collect precise location, contacts, or other data from your phone.
• Details you enter: name, age, sex, height, weight, smoking status, any known heart/lung condition, current symptoms, optional phone number, and your role.
• Recordings: short heart and lung sound clips (.wav) you choose to record.
• Results: the screening outputs computed from your recordings (e.g., rate, rhythm, probabilities, quality).
We do not collect precise location, contacts, or other data from your phone.
3. How it is collected
Recordings are captured on your device. AI screening runs on your device. When you are signed in, your recordings, details, and results are securely uploaded to our protected database so they can be saved to your account and used for research.
4. Why we use it
To provide and improve the screening tool, to save your results, and to build, train, and validate AI models for heart and lung screening — including characterising sounds in healthy people so the models can tell normal from abnormal. Health details are used to label recordings correctly and to remove confounders.
5. Legal basis — your consent
We process your data on the basis of the consent you give when you tick the consent box. Health information is sensitive personal data; we collect only what is necessary for the purposes above.
6. Storage & security
Data is stored on a secured, access-controlled cloud database (hosted in a Mumbai, India region) with authentication and row-level access rules, so each person's data is separated. Access is limited to the investigator and authorised research staff.
7. Sharing
We do not sell your data or show ads. We may share de-identified data (with direct identifiers such as name, email, and phone removed) for research collaboration, model development, and scientific publication. Aggregated or de-identified results may appear in papers and presentations.
8. Retention
We keep your data for as long as needed for the research and model validation, or until you ask us to delete it. De-identified research data may be retained to support published findings.
9. Your rights & choices
You can view your saved recordings in the app (About → My recordings). You can ask us to access, correct, or delete your data, or withdraw your consent at any time by contacting the investigator. Withdrawing consent stops further collection; some de-identified data already used in research may be retained.
10. Children
Recordings for anyone under 18 must be taken only by a consenting parent or legal guardian. We do not knowingly collect data directly from children.
11. Changes & contact
This policy may be updated; the "last updated" date will change. For any privacy request, contact Dr. Jaideep Rao M., Government Medical College, Maheshwaram, Telangana, India.
NOT a diagnostic device. Always confirm clinically.