Imaging Protocol
This page describes image capture, quality control and metadata requirements for cutaneous lupus erythematosus studies using Legit.Health.
Smartphone-based capture
Legit.Health uses standard smartphone cameras for image acquisition. No specialised photography equipment is required.
Traditional clinical photography often relies on systems like Canfield VISIA, which require per-site hardware, per-site calibration, and significant rental or purchase costs. Smartphone-based capture eliminates these costs while maintaining the image quality needed for AI scoring.
The Legit.Health mobile application guides investigators through the capture process with visual perspective silhouettes, real-time DIQA quality checks, and immediate feedback on image adequacy.
Traceability is a precondition, not a convenience
CLASI is a sum over independently scored anatomical regions, recorded per patient and per visit. Every photograph must therefore carry three pieces of metadata, and a fourth where the protocol designates target lesions:
| Metadata | Purpose | Consequence if absent |
|---|---|---|
| Patient | Pseudonymised study identifier | No per-patient score can be assembled |
| Visit | Scheduled timepoint or visit number | No change from baseline can be computed |
| Region | Which of the 13 CLASI regions the image shows | No region-level or total CLASI is possible |
| Lesion | Target lesion identifier, where the protocol designates one | No per-lesion trajectory can be followed |
When these are present, the platform assembles CLASI-A and CLASI-D and tracks both over the study. When any is missing, the device still returns valid per-image sign measurements, and those measurements cannot be assembled into a score. Nothing downstream recovers the missing attachment.
This matters most for retrospective datasets. Images collected outside a protocol designed for region-level scoring, however good the photography, commonly lack the region label. Such a dataset supports per-image sign quantification and comparisons across images, and does not support CLASI reconstruction or longitudinal analysis. Both cases are set out on the trial workflow page.
In a prospective study the platform supplies the metadata automatically: the investigator selects the patient and the visit, then the application guides capture region by region and tags each image as it is taken.
Identification and registration are different problems
It is easy to treat "matching images across visits" as one task. It is two, and they have different solutions.
Identification asks which images belong together: same patient, same visit, same region, same target lesion. This is a metadata problem. Solved at capture, it is exact and costs nothing. Solved afterwards on an unlabelled archive, it is guesswork over filenames and timestamps, and it is the usual reason a retrospective CLE set underdelivers.
Registration asks where the same lesion sits in two images that are already known to correspond. This is a geometric problem and it survives good metadata: two correctly labelled photographs of the same forearm taken at different distances and angles still need to be brought into a common frame before areas can be compared. Reproducible framing reduces the difference, and the in-frame marker resolves the scale component of what remains by converting both images into physical units.
The practical consequence is that metadata discipline and capture discipline are not substitutes. A study needs both: the first so that measurements can be assembled at all, the second so that the assembled measurements are comparable.
The capture set
The standard protocol captures one photograph per CLASI region, plus quadrant coverage of the scalp:
| Group | Captures | Notes |
|---|---|---|
| Scalp | 4 quadrant views | Supports the quadrant-based alopecia and scarring items |
| Head | Ears; nose and malar area; rest of the face | The malar area carries the characteristic CLE rash |
| Neck | V-area of the neck; posterior neck and shoulders | Photo-exposed sites, frequently involved |
| Trunk | Chest; abdomen; back and buttocks | Captured as configured by protocol |
| Limbs | Arms; hands; legs; feet | Hands and feet captured dorsally unless specified |
The region set is configurable. A study restricted to photo-exposed sites captures a reduced set; the CLASI total is then reported over the captured regions and its scope stated explicitly, rather than presented as a whole-body score.
Baseline and follow-up are rarely the same capture
Capturing all 13 regions plus scalp quadrants at every visit is the cleanest design and a heavy one, and most CLE protocols do not ask for it. The common pattern is asymmetric: screening and baseline capture comprehensively, covering the whole body and every region, while follow-up visits capture the regions with active disease plus the designated target lesion.
That is a reasonable trade of completeness against site burden, and it has one consequence that has to be faced at protocol design rather than at analysis. A CLASI total is a sum over regions, so a total assembled from eight regions at week 12 is not comparable with a total assembled from thirteen at baseline, and the difference between them is not a treatment effect.
Two designs resolve it, and a protocol should pick one explicitly:
| Design | Capture | Reported |
|---|---|---|
| Full CLASI at every visit | All regions at every timepoint | CLASI-A and CLASI-D totals, directly comparable across visits |
| Comprehensive baseline, focused follow-up | All regions at baseline, active regions plus target lesions afterwards | Per-region and per-lesion measures as the comparable endpoints; any total reported only over a fixed region set captured at both timepoints |
The second design is often the better study. It simply means the primary endpoint is a per-region or per-lesion measure rather than a whole-body total, which is also where the measurement is least noisy.
Target lesions
Where the protocol designates target lesions, each is assigned a stable identifier at baseline and captured at every subsequent visit, close enough to fill the frame and always with a marker in shot so that area is reported in physical units.
Some protocols also place a small physical label on the skin beside the lesion at capture. Where they do, it helps the photographer find the right lesion and helps a reviewer read the image, and it should sit on unaffected skin clear of the lesion margin, since anything resting on or overlapping the lesion occludes the skin being measured. The binding between an image and a lesion is the identifier in the metadata, not the label in the photograph.
Colour fidelity and physical scale
CLE is unusually demanding of a photograph, because the two measurements that matter most are both comparisons rather than readings.
CLASI erythema is graded by hue and saturation, from faint pink through red to violaceous, and it moves with the illumination the photograph was taken under. CLASI dyspigmentation compares affected skin against surrounding skin, and it moves the same way. Lesion extent, meanwhile, is a count of pixels, and pixels change size with working distance. An uncontrolled change in lighting or camera distance between two visits will move all three measurements with nothing having changed in the patient.
A marker in the frame addresses both problems at once, and they are genuinely separate problems:
Colour reference
Known colour patches in the frame let the analysis normalise for the illuminant rather than assume it was constant, which is what makes erythema grading and pigmentation comparison stable between visits and between sites.
Physical size reference
Known marker dimensions let the camera be calibrated so that segmented pixels convert into square millimetres. Lesion area then stops depending on how close the photographer stood.
Three controls apply, in increasing order of strength:
Consistency of capture conditions is the baseline requirement. The same lighting, distance, angle and patient positioning at every visit, specified in the study-specific investigator manual and reinforced by in-application guidance.
DIQA rejection removes captures whose exposure, focus or framing fall outside the acceptable range before any sign is measured, so a poorly lit image produces a recapture prompt rather than a wrong number.
Calibration markers supply the colour and size references above. For any CLE study where erythema, pigmentation, repigmentation or lesion area is a primary or key secondary endpoint, this is the recommended configuration, and without it lesion extent should be reported as a descriptive proportion rather than as a change from baseline. See calibration markers for specifications and placement.
DIQA: Dermatology Image Quality Assessment
What is DIQA?
DIQA (Dermatology Image Quality Assessment) is an AI-powered image quality assessment algorithm that evaluates every captured image in real time before it is accepted for analysis. It was developed by Legit.Health and published in the Journal of the American Academy of Dermatology (Hernández Montilla et al., 2023).
What DIQA evaluates
| Quality dimension | What it checks | Why it matters |
|---|---|---|
| Focus | Sharpness of the image; absence of motion blur | Out-of-focus images can obscure small lesions, leading to undercounting |
| Lighting | Adequate, even illumination; absence of harsh shadows or glare | Poor lighting creates shadows that mimic or hide lesions |
| Framing | Correct anatomical region captured at the required angle | Incorrect framing means the AI analyses the wrong area |
| Resolution | Sufficient pixel density for lesion detection | Low resolution makes small features undetectable |
How it works in the workflow
- The investigator or patient captures an image through the mobile application
- DIQA evaluates the image immediately (sub-second processing)
- If the image passes: it is accepted and queued for AI scoring
- If the image fails: the person capturing the image receives immediate feedback explaining the quality issue and must recapture
Configurable thresholds
The DIQA pass/fail threshold is configurable per study protocol. Sponsors can choose stricter thresholds for pivotal studies (rejecting more images to ensure the highest quality) or more lenient thresholds for real-world evidence studies.
Patient preparation
- Remove make-up and any camouflage cosmetics from the regions to be photographed, which matters particularly on the face, where cosmetic coverage of a malar rash is common
- Remove sunscreen residue, which can alter apparent skin tone
- Tie back or part hair to expose the scalp quadrants and the ears
- Remove jewellery and clothing obscuring a region to be captured
- Allow skin to return to a settled state after any rubbing, scratching or removal of dressings
Environmental conditions
- Lighting: even, diffuse illumination without harsh shadows or mixed colour temperatures. The same light source at every visit for a given patient wherever possible.
- Background: neutral and non-reflective, since a strongly coloured background casts a tint onto skin.
- Distance: consistent between visits, framing the full anatomical region with minimal surrounding background.
- Positioning: reproducible posture per region, especially for the scalp quadrants and the V-area of the neck.
Facial regions and anonymization
CLE differs from most indications the platform scores in that four of the 13 CLASI regions are on the head, and the nose and malar area carry the disease's most characteristic sign. Facial skin is not incidental to a CLE study; it is a substantial part of the scored surface.
This has to be reconciled with anonymization at protocol design rather than afterwards. The platform applies automatic region-aware anonymization to stored and exported images, and for a CLE study that configuration must remove identifying features while preserving the facial skin being scored. The two requirements are compatible, and they are only compatible if the question is settled before capture begins. An anonymization rule chosen for a trunk-and-limb indication and applied unchanged to CLE would obscure the regions the endpoint depends on.
Where a study's consent framework or data protection assessment restricts facial imagery outright, the protocol can exclude the head regions and report CLASI over the remaining regions, with the reduced scope stated in the analysis plan.
Consistency across visits
The governing principle is that a change in score should reflect a change in the patient. For CLE this means the same regions captured in the same order, under the same lighting, at the same distance, at every visit, with any deviation recorded. Where CLASI-D is an endpoint, this consistency has to hold across the entire study, which may span years and several rounds of site personnel.