Skin cancer detection support that shows the evidence behind the score.
Dermascribe focuses on the lesion image itself: melanoma-risk probability, model confidence, calibration context, Grad-CAM attention, boundary visualization, and morphology proxy signals. It supports clinician review; it does not make a diagnosis.
Melanoma probability
Raw and calibrated risk are visible
Explainability gates
Grad-CAM appears when signal is meaningful
No imaging tower
Phone capture keeps the workflow lightweight
Risk support
73%
calibrated signal
Raw and calibrated probability are separated for review.
AI outputs
Skin lesion AI should show more than a score.
The Dermascribe app is strongest when it explains the image model output: calibrated melanoma-risk probability, decision thresholds, Grad-CAM attention, boundary controls, and morphology proxies.
Melanoma-risk probability
73%
Raw and calibrated risk probability
Dermascribe separates the model signal from calibrated probability so clinicians can see what changed after calibration.
What clinicians can inspect
AI role
support
Claim type
not diagnostic
Final call
clinician
Most skin AI tools stop at the risk label.
Dermascribe should compete by making the model output inspectable: probability, calibration, attention, boundary, and morphology proxies, while keeping final judgment with the clinician.
Inconsistent Lesion Photos
Clinics often rely on ad hoc phone photos, creating variable image quality before AI review even begins.
Unclear Risk Prioritization
Clinicians need a fast second signal for which lesions deserve closer review, biopsy consideration, or referral.
Opaque AI Scores
A risk label alone is not enough. Clinicians need probability, confidence, calibration, and threshold context.
Weak Explainability
Many tools give a result without showing attention coverage, boundary behavior, or morphology proxy signals.
Missing Boundary Detail
The estimated perimeter, asymmetry, circularity, and color variation are often hidden or not exposed at all.
Hardware Dependency
Premium dermoscopy and total-body imaging systems are powerful, but expensive and high-friction for smaller clinics.
The product differentiates on what the AI reveals.
The value is a lightweight skin lesion AI workflow that exposes risk probability, calibration, explainability, and proxy morphology without requiring a heavy imaging setup.
Melanoma-Risk Probability
Shows calibrated risk probability, model confidence, and a clinician-friendly state: elevated concern or benign-leaning.
Raw vs Calibrated
Technical view exposes raw probability, calibrated probability, calibration temperature, and operating mode.
Grad-CAM Attention
Attention coverage and concentration help clinicians understand which visual regions affected the model output.
Boundary Thresholding
The estimated perimeter can be inspected and tuned through threshold controls without treating it as anatomical size.
Morphology Proxy Signals
Surfaces lesion area, asymmetry, circularity, boundary complexity, and color variation as transparent non-diagnostic proxies.
Clinician-in-the-Loop
Every output is decision support. The app avoids presenting AI as a diagnosis or treatment decision.
From image to explainable AI evidence.
This flow is intentionally about skin lesion review support, not broad clinic operations. The core promise is to make the image model easier to inspect and challenge.
Capture the lesion image
Use a standard phone image to start the skin AI workflow without special imaging hardware.
Calculate melanoma-risk signal
Generate probability, confidence, and elevated-concern state for clinician review.
Expose thresholds and mode
Show raw vs calibrated probability, high-risk threshold, and operating mode.
Inspect Grad-CAM attention
Reveal where the model focused and how broad or concentrated the attention signal is.
Review boundary proxies
Inspect boundary threshold, lesion area proxy, asymmetry, complexity, and color variation.
Clinician makes the call
Use the AI output as decision support, with final interpretation left to the clinician.
Decision support, always
Dermascribe should be presented as a clinician-review tool. It can prioritize and explain image signals, but all clinical decisions remain with the treating clinician.
Positioned between consumer apps and hardware-heavy systems.
Competitors often win on regulatory maturity, consumer reach, or imaging hardware. Dermascribe should win the lightweight explainability wedge: melanoma-risk probability, calibration, Grad-CAM, boundary controls, and morphology proxies without a heavy setup.
Targets image-based lesion review without forcing a premium dermoscopy or total-body imaging setup.
Shows thresholds, calibration, and proxy metrics instead of hiding the model behind a single risk label.
Combines Grad-CAM-style attention, boundary visualization, and image comparison for clinician review.
Designed for clinicians to review and override. It avoids autonomous diagnostic positioning.
Phone-based lesion capture
No dedicated imaging tower for core workflow
Melanoma-risk probability output
Raw vs calibrated probability
Visible high-risk threshold/mode
Grad-CAM-style attention overlay
Adjustable lesion boundary threshold
Morphology proxy metrics
Clinician-in-the-loop wording
Comparison is directional and based on publicly available product positioning. Dermascribe is clinical decision support; final decisions remain with qualified clinicians.
Better evidence, clearer review decisions.
Dermascribe should be positioned around review quality and workflow clarity. Economic impact is plausible, but it should be validated in pilots before being treated as a claim.
Support biopsy triage
Clearer AI evidence may help clinicians distinguish cases that need closer review from lower-concern images.
Prioritize concerning lesions
Elevated risk signals can help teams review suspicious lesions sooner while keeping final decisions clinical.
Reduce avoidable friction
A lightweight image workflow may reduce manual photo handling, repeat review steps, and unnecessary follow-up effort.
Dermascribe is clinical decision support. It does not replace clinician judgment, confirm cancer, or determine treatment. Pilot results should drive any final clinical or savings claims.
About
Oleskiew Sharma Research
Dermascribe is developed by Oleskiew Sharma Research (OSR), focused on practical AI systems that reduce clinical workflow friction. OSR builds tools grounded in real clinical environments — prioritizing explainability, clinician trust, and meaningful workflow integration over novelty.
Interested in piloting Dermascribe?
We're working with select clinics to pilot Dermascribe in skin lesion review workflows. If you're interested in early access, a demo, or a research collaboration with OSR, we'd love to hear from you.
Email: contact@osresearch.ai