JetPatient JetVision

01Overview

JetVision is a planned smartphone-imaging pipeline that captures standardised photographs of skin and produces a structured report of derived skin metrics. The intended product is a companion tool inside three surfaces: Patient Passport (for cross-border surgical candidates), the Clinic Portal (for contracted surgeons), and the PostOp recovery module (for monitoring wound-adjacent skin).

JetVision is positioned as a workflow aid, not a clinical decision tool. Image acquisition, derived metrics, and clinical interpretation are decoupled by design: JetVision outputs numbers, a human clinician makes any clinical decision.

Intended use

• Pre-operative documentation of skin condition for cosmetic, plastic, and bariatric surgical candidates inside Patient Passport.
• Surgeon-side baseline capture and serial monitoring inside the Clinic Portal, under the supervising surgeon's name and license.
• Post-operative wound-adjacent skin tracking inside PostOp, with explicit clinician sign-off on capture protocol per procedure category.
• Clinical research collaboration with named academic partners, under IRB-approved protocols.

Out-of-scope use

• Cancer screening of any kind, including melanoma, basal-cell, or squamous-cell detection.
• Diagnosis of any skin condition, including dermatitis, psoriasis, eczema, acne severity, or rosacea staging.
• Triage decisions, urgent-care routing, or any decision that substitutes for in-person dermatologic examination.
• Consumer-facing self-assessment outside of a JetPatient-supervised surgical or recovery workflow.
• Use on populations or anatomic regions outside the v1.0 validated capture protocol (see §4).

02Pipeline details (v1.0 target)

03Biomarker schema (v1.0 launch set: 8–12 metrics)

The v1.0 launch set is intentionally narrow. Each metric below has a published reference standard, a defined unit, and a validation protocol. Marketing references to "25+ biomarkers" describe the long-term roadmap, not v1.0.

Additional biomarkers (laxity, wrinkle severity, hyperpigmentation map, scar-maturation index, vascularity index, etc.) are roadmap items. Each will be added to this card after its own validation study is complete and reviewed by the clinical advisory board.

04Capture protocol and device support

• Supported devices (v1.0): iPhone 12 and later, Samsung Galaxy S22 and later, Google Pixel 7 and later. Other devices return "device not supported."
• Anatomic regions: face (frontal, left, right), neck, décolletage, dorsal hand, forearm dorsum, abdomen (umbilical), thigh (anterior), shin.
• Lighting: ambient indoor lighting with the colour card visible. Direct sunlight, fluorescent flicker, and shadow on the colour card are rejected.
• Distance: device-specific, enforced via on-screen anchor. Capture rejected outside tolerance.
• Colour-calibration card: physical card shipped to the clinician at onboarding. Card barcode is read during capture for batch tracking and recalibration.

05Performance (v1.0 target spec)

Validation is per-biomarker, not aggregate. Each metric is held to the validation target stated in §3 on a held-out test set drawn from the v1.0 reference dataset (§6). No biomarker ships in v1.0 below its stated target.

06Reference dataset (v1.0 target)

• Minimum 2,500 distinct subjects, paired smartphone capture and reference-instrument measurement, across all v1.0 anatomic regions.
• Fitzpatrick I–VI representation across all six skin-tone categories, with at least 250 subjects per category. v1.0 release is blocked if any Fitzpatrick category is below this floor.
• Age distribution: 18–34 (≥ 25%), 35–54 (≥ 35%), 55–74 (≥ 25%), 75+ (≥ 10%).
• Sex distribution: at least 40% female and 40% male at minimum.
• Geographic representation: at least three IRB sites across two continents.
• All subjects under written informed consent and IRB-approved protocols.

07Fairness and skin-tone equity audit

Camera-based skin analytics have a well-documented history of degraded performance on darker skin tones. JetVision treats Fitzpatrick-stratified performance as a release-blocking criterion, not a post-launch fairness audit.

• Per-Fitzpatrick performance: no biomarker may ship if measured accuracy on any Fitzpatrick category is more than 10% below the aggregate target.
• Capture-quality false-reject rate must not vary by Fitzpatrick category by more than 3 percentage points.
• Device × Fitzpatrick interaction reported per cell. Any cell with insufficient sample is held out of the release.
• Quarterly fairness audit published to the Drift, Fairness & Audit Console with a named clinical reviewer.

08Limitations

• JetVision is not a medical device. It is not FDA-cleared, CE-marked, or ANVISA-cleared.
• JetVision does not detect, diagnose, or screen for skin cancer or any malignancy.
• Image quality is the dominant source of variance. Off-protocol captures are rejected, not silently degraded — but operator training is the rate-limiting variable in real-world deployment.
• Tattoos, scars, hair coverage, makeup, and active dermatologic disease can confound several biomarkers. The clinician-facing report flags these confounders when detected.
• Cross-device serial comparison is supported within the same device family, not across.

09Privacy, consent, and data handling

• Patient photographs are PHI. JetVision will collect, transmit, and store photographs only under an executed BAA with the operating clinic and an executed JetPatient patient-consent form naming photograph storage and use cases.
• Consent UI: capture is blocked until the patient consents on-device with a timestamped, audit-logged consent record. Revocation is one-tap and triggers deletion within 14 days.
• Transit: TLS 1.3, certificate-pinned. Storage: AES-256 at rest under customer-isolated KMS keys.
• Default retention: 24 months from capture or 90 days after revocation, whichever is earlier. Retention is configurable per BAA.
• No third-party model providers are used for production image inference. Inference runs on JetPatient-managed infrastructure inside the contracted BAA scope.

10Governance and version log

• Clinical advisory board sign-off required before any biomarker is added to or removed from the v1.0 launch set.
• Quarterly drift audit. Monthly published fairness audit on the Drift, Fairness & Audit Console.
• Every release is reversible. The prior version is retained for at least four prior generations.

↓ See Trust & Methods hub Back to Trust & Methods hub