Fingertip optical scan
Place a fingertip gently over the rear camera and follow the steps. This measures camera-derived optical signals for research; it cannot estimate cancer probability or diagnose cancer.
Check the acknowledgement, then open the camera. Camera permission is required.
Your optical measurement
Complete a 30-second scan to see optical ratios and signal quality. No reference upload is needed.
These measurements cannot detect, rule out or estimate the probability of cancer. The TETRA electromagnetic sensitivity model is a separate hypothetical simulation, not a patient prediction.
Cancer probability: Not available — clinical validation required.
No cloud upload is performed by this simple mode. The older research modules remain accessible in Advanced mode. Do not use this app for medical decisions.
Prospective case/control research
Recommended research sequence: consent → coded Study ID → independent case/control label → three accepted 5-minute scans → export → participant-level modelling → independent external validation.
Research measurement
Participant workflow
Use Research Scan (5 min) for pilot records. Enter only a coded Study ID; direct identifiers must remain in the approved clinical study system.
Automatic optical research analysis
Complete a 30-second quality-passing fingertip scan. This panel then calculates camera-space optical features automatically. No reference upload, manual reference values, or external dataset is required.
Interpretation: The channel percentages describe the measured RGB composition, not a percentage of optical abnormality, tissue pathology or cancer risk. The illustrative Maxwell–Garnett sensitivity uses assumed permittivities and is not fitted to this measurement. A single camera scan cannot determine cancer probability.
3. Compare optical scenarios
After a quality-passing fingertip scan, enter independently documented camera-space R/G and B/G reference values for two optical scenarios. No references are supplied or inferred by the app.
Reference scenario A
Altered optical scenario B
Interpretation: Displayed percentages are normalized mathematical similarity weights from a chosen distance formula. They are not calibrated probabilities, not cancer likelihood, and do not establish that either scenario corresponds to actual tissue pathology. Reference values are not derived from Maxwell–Garnett permittivity without a separately validated optical-camera transfer model. This tool is not a cancer test.
Fingertip acquisition
Quick Scan is for signal/QC development. Research Scan is intended for standardized prospective data collection; the first 30 s are treated as stabilization before research feature extraction.
Start camera and cover the active camera/flash area gently.
Camera diagnostics: waiting to start.
Phone camera: open the published HTTPS BioScan address directly in Safari/Chrome — not inside Canva/editor preview or a local file. Allow Camera access when prompted. Consent is required before recording a scan, but no longer blocks camera preview.
Raw intensity
Filtered pulsatile component
v1.2 preserves the signal-quality gate and adds an internal consistency gate comparing spectral HR with IBI-derived HR. READY is an engineering signal-quality decision only; a completed scan can still be flagged REVIEW if independent estimates disagree.
PPG validation descriptors
Internal consistency validation
Independent heart-rate estimates are compared after the scan. PASS ≤10 bpm difference; REVIEW >10 bpm; REPEAT if basic signal QC fails or an independent estimate is unavailable.
Clinical feature validity
Cancer analysis remains blocked unless signal QC and independent HR consistency both pass.
Peak validation
Dots mark pulse peaks accepted by the prototype peak detector. Inspect the overlay for missed or extra peaks before interpreting interval descriptors.
Reference comparison (optional)
*SDNN and RMSSD are experimental camera-derived pulse-interval descriptors, not clinical HRV measurements. They are marked valid for research export only in a 5-minute scan with at least 30 accepted intervals. Optical R/G and G CV are N/A when the green channel is effectively unavailable.
Model-readiness assessment
Research cohort metadata (optional; de-identified only)
TETRA biometric feature vector
Complete a validated scan to prepare research features.
Labelled cohort workspace
Import a de-identified CSV/JSON file. Minimum required fields: study_id, clinical_label. For model development, each BioScan record should also contain the exported feature fields and a label established independently of BioScan.
Model-development gate
The training gate remains locked until the imported cohort contains both independently labelled cases and controls plus usable BioScan features. Training/validation should use patient-level splits and an external test cohort to reduce leakage and overfitting.
Standardized clinical research record
Study metadata
The export includes protocol/acquisition metadata and BioScan signal descriptors. It remains a research record, not a diagnostic report.
Prospective participant visit
De-identified visit metadata
Three-scan visit workflow
After each 5-minute scan finishes, press “Record current scan in visit”. PASS + consistency PASS = ACCEPTED; signal failure = REPEAT; consistency failure = REVIEW/REPEAT. Failed attempts remain in the visit audit trail.
Local participant dataset
Only pilot-ready records can be added. Duplicate Study IDs are rejected.
Accumulation & readiness dashboard
Readiness checks
Import a cohort in the labelled cohort workspace above to calculate analytics.
Locked research classifier
Model checkpoint
Expected schema: TETRA_PPG_FEATURESET_v1_6core. The checkpoint must contain training-derived normalization and logistic-model parameters. Classification remains locked unless the checkpoint explicitly records completed independent external validation.
Research classification
A result requires three accepted 5-minute scans plus a compatible trained checkpoint. Positive/negative means model-associated pattern only and must not be used for screening, diagnosis, reassurance, or treatment decisions.
Pre-modelling validation gate
Frozen v1 feature set
spectral_hr_bpm · ibi_hr_bpm · hr_difference_bpm · mean_ibi_ms · spectral_snr_db · filtered_sd
Import a cohort to run the validation gate. Core frozen features are channel-independent. R/G and G CV remain optional metadata and are N/A when the green channel is unavailable. Outlier ranges are engineering plausibility checks, not clinical reference intervals.
Lung-cancer PPG/HRV evidence
BioScan's short smartphone fingertip recording must not be treated as equivalent to seven-day wearable HRV. LF/HF is intentionally not calculated from the short scan. A same-protocol prospective case/control cohort is required before cancer-associated inference can be evaluated.
Participant-level cancer-association modelling hand-off
AGGREGATE_PARTICIPANTS_V34.py, then run TRAINING_PIPELINE_V34.py. The bridge aggregates the three accepted scans into one participant-level vector and uses the six prespecified channel-independent features. No individual cancer output is enabled in the browser.Outputs include cross-validated ROC-AUC and balanced accuracy, held-out ROC-AUC, sensitivity, specificity, confusion matrix, coefficients, predictions and a machine-readable metrics file. These are exploratory research outputs only.
Fingertip Effective-Medium Research
Illustrative two-component Maxwell–Garnett calculation (dilute spherical inclusions in a uniform host). It does not represent the full layered, absorbing, scattering fingertip.
Run the forward model. Complete a validated scan to populate camera-derived proxies. These are independent outputs, not fitted to each other.
Fingertip three-layer sensitivity model
Outputs are simulated fractional transmission changes, not fitted to patient RGB. Diffusion-style effective attenuation is a toy approximation, not a validated layered-skin optical solution.
Can three RGB channels distinguish tissue changes?
Three broad overlapping camera channels cannot reliably identify arbitrary optical tissue properties without calibration and constraints.
Calibration readiness checklist
Before comparing phone readings with this model: record the phone model, illumination/flash mode, exposure and white-balance lock state; acquire dark/reference frames; use calibrated optical phantoms with known absorption and scattering; quantify repeated-scan and inter-device variability; document pressure and temperature; characterize the camera spectral response. No phantom calibration is claimed here.
Measured RGB reproducibility & forward-model comparison
Complete a camera scan before capturing. At least 3 QC-passing scans are recommended for repeatability. Simulation is separate and does not fit patient data.
Research-only model pattern comparison
Run the v6.1 layered simulation above, then analyze measured scans. The comparison is qualitative and non-identifiable.
RGB dark/reference normalization and repeatability
Use the same phone, illumination and exposure settings. Enter mean RGB channel values (0–255) from independently acquired dark and reference images. A blank/reference image should use an appropriate reproducible optical target, not a fingertip. No camera dark/reference image capture is automated in this version.
Formula: (sample − dark)/(reference − dark). Values outside [0,1] are retained and flagged; they do not imply physical reflectance or transmittance.
Phantom measurement validation laboratory
Enter a phantom ID, independently characterized absorption μa and reduced scattering μs′ (mm⁻¹) at a stated wavelength, then normalize the latest phantom camera scan with the v6.4 dark/reference values. Do not use patient identifiers or clinical scans in this module.
Requires a completed scan and valid v6.4 reference values. Repeated scans of the same phantom can quantify within-session reproducibility; distinct phantom properties allow exploratory association analysis only.
Literature-informed optical modelling and anomaly analysis
Literature catalogue (verified bibliographic sources)
Jacques (2013), Optical properties of biological tissues: a review, Phys. Med. Biol. 58 R37–R61. DOI 10.1088/0031-9155/58/11/R37. Optical absorption, scattering, chromophores and wavelength dependence.
Gabriel, Lau & Gabriel (1996), Dielectric properties of biological tissues III: parametric models, Phys. Med. Biol. 41 2271–2293. DOI 10.1088/0031-9155/41/11/003. Frequency-dependent dielectric spectra, not interchangeable with RGB measurements.
Illustrative optical forward model (not literature-calibrated)
Experimental optical deviation assessment
Use matched, independently established healthy-reference distributions for the same phone, acquisition protocol and feature. Demonstration numbers are synthetic placeholders and must not be interpreted as published clinical reference values. Enter your own mean and SD after obtaining appropriate data.
No assessment yet. Cancer status: insufficient evidence.
Interpretation: z = (observed R/G − reference mean)/reference SD. |z| ≥ 2 is an illustrative statistical flag only, not a clinically validated threshold. All outputs remain cancer-indeterminate.
Repeated fingertip measurements & optical trends
Complete a fingertip scan above, then capture it below. Use the same phone, finger, camera, lighting and pressure. Only scans passing both QC and consistency checks are included in baseline calculations.
Capture at least five distinct passing scans before setting a baseline. Results are held only in this browser session until exported.
Optical R/G trend
A baseline is a within-session camera reference, not a healthy-population reference. Small baseline SD can exaggerate z-scores. No fixed clinical thresholds are applied. No data is transmitted by this module.
Measured optical variation versus simulated effective medium
First set a v7.2 baseline and capture a new passing scan. Then run the illustrative v6.0 Maxwell–Garnett forward model above. This module will summarize both without fitting one to the other.
The measurement and simulation are reported in separate columns; they cannot be equated or correlated from a single paired observation. A physical mapping would require independent phantom measurements, spectral/illumination calibration, a specified forward optical model and external validation.