Clinical evidence
This section compiles the evidence behind the automated acne scoring provided by Legit.Health: the peer-reviewed study of ALADIN (Acne Lesion And Density INdex) against expert dermatologists, the validation of the model that detects the lesions, the image quality gate, and the foundational references for acne endpoints. Together they document the scientific basis of the scoring and what remains to be measured before it is used as a clinical trial endpoint.
Validation status and reproducibility
The ALADIN methodology was published in Skin Health and Disease in 2026. The study is retrospective: it compares ALADIN and individual dermatologists, image by image, with the consensus of expert dermatologists on two datasets. Its results are descriptive, with no acceptance criterion set in advance, and the formula was fitted on the larger dataset, so the independent evidence rests on the smaller one. Agreement with the treating investigator in a prospective study is the subject of the post-market clinical follow-up.
Severity scoring has no objective gold standard, and the IGA is no exception: it rests on broad descriptors that each rater applies differently. ALADIN is therefore measured against the consensus of several dermatologists, which is the realistic performance ceiling for any rater, human or AI.
Two properties are decisive for endpoint use in a trial:
- Reproducibility: scoring is deterministic. The same image yields the same count, density and score at every site and every visit, with no calibration drift and no inter-reader variability in the automated read.
- Traceability: every lesion that enters the count is marked on the photograph, so each score can be checked against the image it came from.
Legit.Health is a Class IIb medical device certified by BSI, Notified Body 2797, under the EU Medical Device Regulation (MDR 2017/745), meaning a Notified Body has independently assessed it against the regulation's safety and performance requirements before it could carry the CE mark. Beyond the EU, it is registered with the MHRA in the United Kingdom and with ANVISA in Brazil. Real-world performance is monitored continuously through the manufacturer's post-market surveillance and post-market clinical follow-up (PMCF) programme under MDR Annex XIV.
Legit.Health is a clinical decision support device: the automated scores provide diagnostic support and do not replace the healthcare professional's assessment.
ALADIN (Acne Lesion And Density INdex)
ALADIN (Acne Lesion And Density INdex): a novel tool for automatic acne severity assessment. Medela A, Sabater A, Hernández Montilla I, et al. Skin Health and Disease. 2026;6(4):503–510. DOI: 10.1093/skinhd/vzag062 | PMID: 42539975
The foundational publication for ALADIN, describing the inflammatory lesion detector, the spatial density measurement and the formula that combines them into a score on the IGA scale. The study evaluated ALADIN on two datasets of facial photographs, each graded on the IGA by expert dermatologists working independently. The reference for each image is the consensus of their grades, and each dermatologist's own grade is compared with the same consensus, so the two rows of each table below measure the same thing. The study reports the score on the 0 to 4 scale of the IGA; in the platform, ALADIN reports the same score multiplied by 2.5, on a scale from 0 to 10.
| Dataset | Images | Source | Readers | Role in the study |
|---|---|---|---|---|
| AcneSeverity-V1 | 331 | Clínica Dermatológica Internacional, Spain; predominantly mild to moderate | 3 expert dermatologists | Fitting the formula, and in-sample evaluation |
| AcneSeverity-V2 | 27 | Public dermatology atlases; a larger share of moderate and severe acne | 2 expert dermatologists | Independent, exploratory evaluation |
Agreement with the expert consensus
| On the IGA scale | AcneSeverity-V1 (in-sample) | AcneSeverity-V2 (independent) |
|---|---|---|
| Cohen's κ, ALADIN | 0.53 | 0.61 |
| Cohen's κ, dermatologists | 0.46 | 0.62 |
| Mean absolute error, ALADIN | 0.63 | 0.64 |
| Mean absolute error, dermatologists | 0.65 | 0.44 |
Cohen's κ (with quadratic weights) measures agreement with the consensus beyond chance, giving more weight to larger disagreements, so a higher value is better. Mean absolute error is the average distance from the consensus in IGA grades, so a lower value is better. The dermatologists' rows are the individual dermatologists against the same consensus.
How to read these results:
- On AcneSeverity-V1, ALADIN's agreement with the consensus is at least as high as the dermatologists', with a similar average error. This is the dataset the formula was fitted on, so it is an in-sample result.
- On AcneSeverity-V2, the independent dataset, the agreement is similar to the dermatologists', while the average error is higher for ALADIN than for the dermatologists. The dataset is small, and no confidence intervals were reported.
- What was measured is agreement on single photographs, not the change in severity over a course of treatment, and most images came from patients with lighter skin.
Manual acne assessment requires an investigator to count every inflammatory lesion on the face and to judge the IGA from broad descriptors at every visit. Automated scoring gives every investigator the same objective read of the photographs, with three properties manual scoring cannot offer:
- A reproducible read: the same image always produces the same count, density and score, at every site and every visit.
- No investigator training drift: performance does not depend on how recently a site was trained or on which rater is on shift that day.
- A continuous score: ALADIN keeps the intermediate values that a whole IGA grade rounds away, so it can show smaller changes between visits.
In a trial, ALADIN provides objective, supportive data alongside the investigator's IGA; see Limitations.
The ALADIN formula used in the platform is the one published. The lesion detector behind it is revalidated under the Quality Management System, and its current results are the ones reported in Lesion detector performance below. New models that tell lesion types apart are in development; any change follows the same validation before it is deployed.
Lesion detector performance
The model that detects the inflammatory lesions is validated against its reference under the Quality Management System (IEC 62304 and ISO 14971). It was evaluated on validation images held out from training, annotated with a box around each inflammatory lesion, overall and separately by skin phototype and by the number of lesions.
| Model or subgroup | Metric | Result (95% CI) | Acceptance criterion |
|---|---|---|---|
| Inflammatory lesion detector, all validation images | mAP@50 | 0.45 (0.43 to 0.47) | mAP@50 ≥ 0.21 |
| Fitzpatrick I and II | mAP@50 | 0.42 (0.37 to 0.47) | mAP@50 ≥ 0.21 |
| Fitzpatrick III and IV | mAP@50 | 0.46 (0.44 to 0.49) | mAP@50 ≥ 0.21 |
| Fitzpatrick V and VI (8 validation images) | mAP@50 | 0.45 (0.04 to 0.73) | mAP@50 ≥ 0.21 |
| Mild, 0 to 5 lesions | mAP@50 | 0.40 (0.32 to 0.48) | mAP@50 ≥ 0.21 |
| Moderate, 6 to 20 lesions | mAP@50 | 0.48 (0.44 to 0.52) | mAP@50 ≥ 0.21 |
| Severe, 21 to 50 lesions | mAP@50 | 0.48 (0.44 to 0.51) | mAP@50 ≥ 0.21 |
| Very severe, more than 50 lesions | mAP@50 | 0.43 (0.38 to 0.47) | mAP@50 ≥ 0.21 |
mAP@50 (mean average precision at 50% overlap) measures how accurately the model locates individual lesions, so a higher value is better. The acceptance criterion is non-inferiority to published acne lesion detection studies.
Every group meets the criterion on its point estimate. The darkest phototypes (V and VI) are represented by very few validation images, so their confidence interval is wide and reaches below the criterion, and the model's development report records the need for more data from darker skin. Detection quality is consistent from mild to very severe acne.
Post-market clinical follow-up
ALADIN's agreement with the treating investigator is being measured prospectively in a post-market clinical follow-up study of patients with acne, under the PMCF programme of the certified device. At each visit the investigator records the patient's IGA before the photograph is analysed, and the study compares that grade with ALADIN against a threshold set in advance; photographs taken by the patients at home are also graded. Recruitment starts once the ethics committee has issued a favourable opinion.
Image quality for acne scoring (DIQA)
Dermatology Image Quality Assessment (DIQA): Artificial intelligence to ensure the clinical utility of images for remote consultations and clinical trials. Hernández Montilla I, Mac Carthy T, Aguilar A, et al. Journal of the American Academy of Dermatology. 2023;88(4):927–928. DOI: 10.1016/j.jaad.2022.11.002 | PMID: 36526082
Reliable acne scoring depends on the quality of the input photographs. DIQA is the image quality assessment algorithm that acts as a quality gate in the acne imaging workflow, checking every image against consistent quality criteria across investigator sites before it reaches the detection model. The dependency is sharp in acne, where lesions are small: a slightly blurred or poorly lit photograph can hide them.
Image quality is the hidden failure point of multi-site imaging. DIQA screens every photograph for clinical utility before it reaches the scoring algorithms, so unusable images are caught at capture rather than surfacing as missing data at database lock.
Conference presentations
EADV Congress (European Academy of Dermatology and Venereology) 2025, Paris
Poster: ALADIN: Acne Lesion And Density INdex. A Novel Tool for Automatic Acne Severity Assessment
The ALADIN method: inflammatory lesion detection, spatial density and the formula that combines them into a score on the IGA scale.
Foundational acne references
ALADIN is a new score, but it is built from, and read on, the measures acne trials already use: the inflammatory lesion count and the IGA, the two measures the FDA guidance bases efficacy on. The lesion count can be reported on its own, so the endpoint your protocol specifies is an endpoint the algorithm reports.
Hayashi N, et al. “Establishment of grading criteria for acne severity” J Dermatol. 2008. doi:10.1111/j.1346-8138.2008.00462.x
Grades acne severity by the number of inflammatory eruptions (papules and pustules) on one half of the face.
U.S. Food and Drug Administration “Acne Vulgaris: Establishing Effectiveness of Drugs Intended for Treatment. Guidance for Industry.” FDA Guidance Document. 2018.
Recommends that efficacy be based on both lesion counts and success on an investigator's global assessment (IGA), with inflammatory and noninflammatory lesions counted separately.
Bae IH, et al. “A Comprehensive Review of the Acne Grading Scale in 2023” Ann Dermatol. 2024. doi:10.5021/ad.23.094 PMC10995619
Documents the variability across existing acne grading scales, including the many IGA variants in use.
Traini DO, et al. “Artificial Intelligence in the Assessment and Grading of Acne Vulgaris: A Systematic Review” J Pers Med. 2025. doi:10.3390/jpm15060238 PMC12194645
Reviews AI applications in acne diagnosis and severity grading from 2017 to 2025.
For the summary of the ALADIN study, see ALADIN in the clinical validation section; for the full list of clinical evidence across all indications, see the clinical validation section.
Accélérez vos essais
Rejoignez les sociétés pharmaceutiques, biotechs et CROs qui atteignent le marché 4 à 6 mois plus rapidement avec une qualité de données supérieure. Remplissez le formulaire pour discuter des exigences de votre étude.
