Kolena AI Agent
Skin-Image Measurement (Vision)
Turn a personal-care study's standardized skin photographs into clean, cited per-image measurements and change vs baseline - wrinkles, spots, pores, redness, texture, radiance, or perceived age - with a capture-quality gate on the front.
Impact
Manual, image-by-image measurement of each study's standardized photographs and re-keying into the master sheet - normally done by hand across many volunteers and timepoints - compressed into a single cited, quality-gated run that emits the change-vs-baseline read-out directly
per study image set
Volume fit
Works best for
R&D and imaging teams processing many-volunteer, multi-timepoint image studies where each study yields dozens to hundreds of standardized photographs to measure
Too small for
a one-off read of a single before/after image pair
Typical inputs
Documents
- Standardized skin study photographs (one volunteer's measurement area at one timepoint per image)
- Barcoded image filenames encoding volunteer, timepoint, area, and treatment code
- Cross-polarized, parallel-polarized, white-light, or UV facial images
- Any accompanying capture log or study protocol
Systems
- Facial-imaging system / capture station
- Contract-lab or imaging portal
- SharePoint
- Google Drive
Output
A cited measurement read-out - image inventory with barcode parse, a per-image capture-quality assessment and study-level quality gate, per-image skin-attribute measurements within a defined region of interest, per-volunteer change vs baseline (absolute and percent), a group-level change summary, an overall change call, data-integrity checks, and an analyst summary
Delivered to
- Efficacy-study data-extraction workflow
- Claims substantiation dossier / Product Information File
- R&D study database or tracker
What it extracts
· 10 fields
- Form
Study & Imaging Overview
The one-look roll-up read from the image set - study design, imaging modality and lighting, measurement area, product category, the attribute(s) measured, and the number of volunteers, timepoints, and images - so a reviewer can orient before reading detail.
- Table
Image Inventory & Barcode Parse
Every image with its verbatim filename and the volunteer, timepoint, measurement area, and treatment code parsed from its barcode; a mis-parsed barcode attributes a measurement to the wrong volunteer or timepoint and invalidates the read-out.
- Table
Image Quality Assessment
Each image scored against standardized-capture criteria - lighting consistency, pose/angle alignment, focus, and region-of-interest framing - with an overall Pass / Borderline / Fail call and an issue note, because repeatable capture is the precondition for a trustworthy before/after comparison.
- Classification
Study Image-Quality Gate
The study-level go/no-go on the imagery - all pass, minor issues usable with caveats, some fail and re-capture is recommended, or not analyzable - read from the per-image quality assessment.
- Table
Per-Image Skin Attribute Measurements
The core measurement - one numeric estimate per image and attribute (perceived age by default; also wrinkles/fine lines, spots/pigmentation, pores, redness/erythema, texture, radiance) within a defined region of interest, on a single consistent scale aligned to a grading reference where one exists.
- Table
Baseline & Timepoint Comparison
Each volunteer's change over time per attribute and follow-up timepoint - baseline and timepoint value, absolute change, and percent change vs baseline, with an improvement/worsening direction on the attribute's convention - the effect on the volunteer's own skin.
- Table
Group-Level Change Summary
The panel roll-up per attribute and timepoint - mean baseline, mean timepoint, mean percent change vs baseline, contributing N, and direction of effect - computed only from complete baseline-to-timepoint pairs, never by placing a group mean into an individual's cell.
- Classification
Overall Measured Change
The single descriptive call on the primary attribute - improvement, no meaningful change, mixed across attributes, or insufficient/gated-out imagery - read from the quality gate, per-volunteer comparison, and group summary.
- Table
Measurement Data-Integrity Validation
The consistency checks an analyst runs before trusting the read-out - every image parsed, failed-QC images excluded, a baseline present per volunteer, percent change ties to the values, one scale per attribute, no group mean in an individual cell, and consistent region of interest - each Pass / Fail / Review.
- Text
Study Measurement Summary
The analyst narrative - imaging setup and panel, whether the imagery cleared the quality gate, the key measured changes and their magnitude on the correct basis, the overall change call, and any integrity flags or caveats to resolve before relying on the results.
Prerequisites
- The full image set for the study (baseline plus follow-up timepoints)
- The parameter to measure (perceived age is the default; e.g. wrinkles, spots, pores, redness, texture, or radiance)
- The barcode/filename convention and, if separate, the treatment-code key
Human review
An image-analysis or claims scientist reviews the capture-quality gate, spot-checks per-image measurements against the images, and confirms the change-vs-baseline read-out and improvement conventions before the numbers feed a claim. Statistical significance is established in the downstream efficacy-study workflow, not here.
Where it fits
Cosmetic / personal-care instrumental image measurement for efficacy testing
Volunteers are photographed under standardized conditions - a multi-point-positioned facial-imaging system, controlled lighting (cross-polarized, parallel-polarized, white, or UV) - at baseline and one or more follow-up timepoints; the barcoded image set is delivered by portal, shared drive, or upload
This step
Computer-vision measurement of the skin images and baseline-to-timepoint change
After
Who uses it
Grounded in
- Regulation (EU) No 655/2013 - common criteria for cosmetic product claimsverified as of 2026-07-22
- Cosmetics Europe - Guidelines for Cosmetic Product Claim Substantiationverified as of 2026-07-22
- Standardized facial imaging (VISIA-style, e.g. Canfield) - de-facto method for objective skin-feature measurementverified as of 2026-07-22
- Validated photonumeric grading scales for skin attributes (e.g. wrinkle-severity and erythema scales)verified as of 2026-07-22
Related agents
Browse all agents →Clinical / Efficacy Study Data Extraction
Turn a personal-care efficacy study report into a clean, cited read-out - design, endpoints, statistics vs baseline and between treatments, and whether each claim is substantiated.
Product Claims & Concept Development
Pressure-test a consumer / personal-care product concept and its marketing claims: break the concept into insight, benefit, reason-to-believe, and positioning; inventory every express and implied claim; link each to a substantiation basis; test claim-to-evidence adequacy against FTC, NAD, and EU 655/2013; flag overclaiming; and return a cited claims-substantiation matrix, a readiness call, and a finalised concept summary.
Consumer Complaint Analysis
Triage a consumer personal-care product complaint into a clean, cited record - category, product and batch, adverse-event and seriousness assessment, a MoCRA / EU cosmetovigilance reportability screen, root-cause signals, and recommended actions.
See Skin-Image Measurement (Vision) on your documents
We'll run it against a file of yours and walk through every cited field.