First steps into AI‐supported hidradenitis suppurativa severity assessment

Hidradenitis suppurativa (HS) is a debilitating, chronic, inflammatory skin disease with clinical heterogeneity and comorbidities, like psoriasis.1 Its prevalence is estimated to be around 1%, and it occurs more frequently in females than males and usually presents after puberty. Lesions arise mainly in the axillae, groin, and gluteal area and include inflammatory nodules, painful abscesses, and draining tunnels. Patients often experience long delays before being correctly diagnosed. Furthermore, HS, unlike psoriasis, can progress from an inflammatory component to a non-reversible scarring component, if not treated early enough in an evidence-based manner. Correct and in time treatment has been shown to be quite crucial for HS progression. The 'window of opportunity' refers to the period during which efforts to control inflammatory activity may be most effective.2 Immunomodulatory therapy during this window may alter the natural progression of the disease by reducing the accumulation of tissue damage. A critical step to capture this window is the correct and in time severity assessment of HS, both when it comes to identifying the patients with non-mild disease and to correctly classify the disease severity in a dynamic manner. In the last years we have experienced a vast number of efforts to develop AI-supported decision-making tools in medicine. Many of them are trying to provide an automated AI-assisted disease severity assessment. In order for such a tool to be of importance, validity, and clinical relevance specific points should be taken into consideration under development: (1) The model used and the analyses employed should be appropriate with a well-defined possible clinical significance; (2) the model proposed should be developed and validated with reference to well-validated already existing scores that have been shown to have clinical relevance and significance; and (3) the limitations are well acknowledged. I have read with interest the effort of Wiala et al.3 to develop an automated AI-supported disease classification that could be used not only by HS experts, but most importantly by nonexperts, general practitioners, and other dermatology specialists as well. They employed a correct methodology and a clear rationale to develop and initially validate a tool that could correctly classify severity of an HS clinical picture and also correctly differentiate a mild from moderate/severe case. They used as a reference for the development and validation of a well-accepted, validated, and with clinical importance score, International Hidradenitis Suppurativa Severity Score System (IHS4), which has been shown to serve the correct capture of window of opportunity and is a more holistic measure of HS disease severity compared with Hurley staging.4, 5 Furthermore, the authors clearly report the limitations of the current effort, like limitations in assessing tattooed and hairy skin, skin of colour, and the low number of follow-up images of the affected areas in order to assess the disease dynamics assessment. This proof-of-concept study demonstrates that AI-based scoring of HS is a feasible and possibly important tool in the hands of clinicians. The current effort is of interest and, on the right path, future steps should include variability assessment, assessment of the ability to capture the disease dynamics using validated scores as reference (IHS4-55)6 and including appropriate follow-up and include HS experts during development to increase consensus and facilitate improvement under development. None. Data sharing is not applicable to this article as no new data were created or analysed in this study.