Skip to main content
Skinpendium
Investigational device · pursuing FDA 510(k)

Skin cancer screening, seen clearly.

Skinpendium gives non-dermatologist physicians a calibrated malignancy probability and the closest confirmed cases for every suspicious lesion — in seconds, inside the visit.

  • Calibrated malignancy probability
  • Similar confirmed cases
  • Writes to the EHR
Skinpendium analysis result showing a calibrated malignancy risk and a differential diagnosis

What it does

Decision support that supports the decision.

Transparent, calibrated and integrated. Skinpendium informs the biopsy-or-refer call; the clinician makes it.

Built for skin cancer, not generic vision

A dual-head model on a frozen vision foundation model, trained on histopathology-confirmed dermoscopic images, scores malignancy and separates the common benign and malignant lesion classes.

  • Melanoma
  • Squamous Cell Carcinoma
  • Basal Cell Carcinoma
  • Actinic Keratosis

Confidence you can act on

Isotonic calibration keeps reported probabilities honest: a stated risk behaves like that risk, with no inflated confidence.

Show the evidence

Every result arrives with the most visually similar histopathology-confirmed cases, so the reasoning is inspectable.

Lives in the EHR

A SMART-on-FHIR plugin for Epic and Oracle Health (Cerner) writes DiagnosticReport, ServiceRequest, Task, Media back to the chart with server-side audit logging.

Native on iOS and Android

One shared clinical core: capture with quality gates, offline-first sync, per-patient galleries and PDF reports.

Designed for HIPAA

Images encrypted at rest, Safe Harbor de-identification, TLS everywhere and per-organization API keys with a full audit trail.

Validated performance

Built to the standard a regulator reads.

Melanoma sensitivity is the special control for this device category. Skinpendium is engineered to clear it while keeping specificity high enough to be useful in the room.

  • Histopathology ground truth

    Trained and validated on biopsy-confirmed lesions, not clinician impressions.

  • Tested on new institutions

    Evaluated on images from clinics and sources the model never saw in training.

  • Rigorous by design

    Every model change is registered before it is tested, and tested on its own, so an improvement is real rather than noise.

Workflow

Three steps, one visit.

  1. 01

    Capture

    Attach a dermoscope to your phone. On-device quality gates check focus, exposure and framing before anything is submitted.

  2. 02

    Analyze

    The AI engine returns a calibrated malignancy probability, a differential and the closest confirmed cases in seconds.

  3. 03

    Decide

    Review the result beside your own assessment, then biopsy, refer or follow up. The report writes straight to the chart.

Provider workspace

Run it across the whole practice.

Clinicians, devices and sites in one place, with every flagged lesion followed through to its pathology result.

Open the preview

Be first in the clinic.

Skinpendium is investigational, not FDA-cleared. Leave your email and we will tell you when it is available for clinical use.

For healthcare professionals only. We never share your information.