Scoring Methodology

The platform from Legit.Health scores hidradenitis suppurativa from lesion counts: the software detects every lesion, classifies it by type and scores the visit, and the investigator then reviews the result after examining the patient. The reported IHS4 and HiSCR are calculated from the confirmed counts, and the counts are reported in their own right, so any other count-based measure a protocol needs comes from the same capture.

Lesion detection and counting

1PhotographEach affected region

2DetectA bounding box per lesion

3ClassifyNodule, abscess or tunnel

4CountThree counts per visit

5ScorePreliminary IHS4 and HiSCR

6 · InvestigatorReviewInvestigator confirms, edits or adds

7RecalculateFinal IHS4 and HiSCR

A deep learning object detection model, trained on images annotated by a board of specialists and unified into a consensus reference, finds each lesion and assigns it to one of three types.

Inflammatory nodules

A firm, rounded, tender inflammatory lump, without pus. The earliest active lesion.

What the AI looks for

Solid, raised, well-defined borders, uniform colour

Abscesses

An inflamed lesion filled with pus, fluctuant to the touch. More active disease.

What the AI looks for

Fluctuant appearance, surrounding erythema, irregular surface

Draining tunnels

A tract under the skin that opens to the surface and can discharge pus or fluid.

What the AI looks for

Linear tracts, surface openings, discharge

Investigator review

Hidradenitis suppurativa is the one indication where the software's analysis is not the final word, because much of the disease sits beneath the skin and cannot be confirmed from a photograph alone. Once the software has returned its detections and scores, the investigator examines and palpates the patient and, in the annotation interface, confirms each detected lesion, corrects its type, removes it, or adds lesions the photograph does not show, before confirming the assessment.

1. Original photograph
The affected region, captured at the visit

2. Software analysis
Each lesion detected and classified by the AI

3. Investigator review
Detections confirmed, and a lesion added after palpation

Inflammatory nodulesAbscessesDraining tunnels

The three counts for the visit (inflammatory nodules, abscesses and draining tunnels) are then recalculated from the confirmed lesions, together with the IHS4 and HiSCR. These confirmed values are the ones reported and exported.

Why draining and non-draining tunnels are not separated

Hidradenitis suppurativa develops mostly beneath the skin. A photograph shows a tunnel's opening and the tract it outlines, but whether it is actively draining depends on the patient's history and the clinical examination, which the image does not carry. The AI therefore counts every visible tunnel. See Limitations.

IHS4

The International Hidradenitis Suppurativa Severity Score System (IHS4) was developed by the European Hidradenitis Suppurativa Foundation and validated in 2017 (Zouboulis et al., British Journal of Dermatology). It is a dynamic score that follows the lesions present at each visit.

It weights each lesion type by how advanced the disease it reflects:

$$\text{IHS4} = (1 \times \text{inflammatory nodules}) + (2 \times \text{abscesses}) + (4 \times \text{draining tunnels})$$

where each term is the number of lesions of that type counted at the visit.

IHS4 is purely a count: no affected area and no calibration markers, only every lesion visible and in focus. For the same reason the scale has no maximum, unlike PASI or EASI (72).

Severity thresholds

The bands are those defined in the original IHS4 validation study.

Mild

Moderate

Severe

0–3

4–10

11 and above →

IHS4 severity bands. The severe band is open-ended because every additional lesion adds to the score.

Why automated counting matters

Investigators disagree on which lesions are abscesses and which are nodules, and on whether a tract is a tunnel, and each disagreement moves the score by 1 to 4 points. The AI applies one classification boundary to every image, so every investigator starts from the same first read instead of counting from scratch.

HiSCR

The Hidradenitis Suppurativa Clinical Response (HiSCR) is the response endpoint most used in hidradenitis suppurativa trials. Where IHS4 scores one visit, HiSCR compares a follow-up visit with the patient's baseline visit, using the same lesion counts. It is built on the AN count:

$$\text{AN count} = \text{abscesses} + \text{inflammatory nodules}$$

A patient achieves HiSCR at a visit when all three conditions below are true. If any one fails, HiSCR is not achieved.

1

AN count falls by at least half

The AN count at the visit is 50% or less of the AN count at baseline

2

Abscesses do not increase

The number of abscesses is the same as at baseline or lower

3

Draining tunnels do not increase

The number of draining tunnels is the same as at baseline or lower

The platform stores the baseline counts and checks the three conditions automatically at every follow-up visit. Stricter versions used in some protocols, such as HiSCR 75 or HiSCR 90, raise the first condition to a 75% or 90% reduction.

A worked example

The same patient followed across a 48-week trial, with an assessment every 16 weeks. The IHS4 tracks how much disease remains at each visit; the AN count, measured against the baseline, shows when the patient reaches HiSCR 50 and HiSCR 75.

IHS4 by visit

Against the IHS4 severity bands

AN count by visit

Against the HiSCR thresholds set by the baseline

Visit	Baseline	Week 16	Week 32	Week 48
Inflammatory nodules	8	4	2	1
Abscesses	4	1	1	0
Draining tunnels	2	2	2	1
IHS4	24 Severe	14 Severe	12 Severe	5 Moderate
AN count	12	5 (−58%)	3 (−75%)	1 (−92%)
HiSCR 50	Reference	Achieved	Achieved	Achieved
HiSCR 75	Reference	Not yet	Achieved	Achieved

The two measures move differently. The patient reaches HiSCR 50 at week 16 and HiSCR 75 at week 32, while the IHS4 stays in the severe band until the draining tunnels, which weigh most and respond slowest, start to close.

IHS4 and HiSCR

IHS4

Severity at one visit

Output

A score from 0 upwards, with severity bands

Needs baseline

No

Typical use

Severity at inclusion, stratification, secondary endpoint

HiSCR

Response to treatment

Output

Achieved or not achieved

Needs baseline

Yes

Typical use

Primary efficacy endpoint

IHS4 and HiSCR are calculated from the same confirmed lesion counts. The choice of primary and secondary endpoints is made during protocol design.

Manual vs. automated scoring

Manual scoring

IHS4 counted by a clinician

• Lesion identification

  Visual inspection and palpation

• Type classification

  Clinician judgement, abscess vs. nodule often ambiguous

• Reproducibility

  Low inter-rater agreement across sites

• Training

  Calibration exercises to align raters

• Time per assessment

  Grows with lesion burden: every lesion found, examined and classified

Automated scoring

AI detection with investigator review

• Lesion identification

  AI detection on every photograph, confirmed by the investigator

• Type classification

  One consistent boundary, learned from specialist consensus

• Reproducibility

  Same first read for the same image, every time

• Training

  Reviewing detections instead of counting from scratch

• Time per assessment

  Seconds of AI analysis, plus capture and the investigator's review

The model version is locked at study initiation, so every patient is scored by the same model from first visit to last; see model version locking. The peer-reviewed validation of the model is on Clinical evidence.

Image-based scoring

The AI reads the lesions from visual cues. What only touch can tell, such as the fluctuance of an abscess or involvement beneath the skin, is established by the investigator's examination during review; see Limitations.