Search for a command to run...
Standardized scoring tools are crucial for assessing severity, following disease activity and evaluating treatment efficacy both in randomized controlled trials and in real life. The rapid development of efficient biotherapies in various chronic cutaneous conditions prompts dermatologists to know about and use those scores, sometimes on a daily basis: Psoriasis Area and Severity Index (PASI, psoriasis), EASI (atopic eczema), VASI (vitiligo), SALT (alopecia areata), CLASI (cutaneous lupus) and so forth. Scoring suffers from several limitations: challenges to learn complex scoring systems, subjectivity, limited sensitivity to smaller lesions and significant intra- and interrater variability may impact the outcome. Even though practice leads to mastery, their use can be time-consuming within the limited duration of a consultation, which is already demanding for patients with chronic inflammatory dermatoses due to history taking, clinical examination, assessment of quality-of-life impact through questionnaires and discussion of treatment strategies. Some tools are of course at the disposal of dermatologists such as hand-held calculator (Figure 1) or free online score calculator on the internet (Figure 2), however they barely save time much time nor solve the limitations stressed above. In an article of this issue [1], Taig MacCarthy et al. reported on the development and validation of APASI. This new AI-driven system automatically evaluates the severity of psoriasis using clinical images. Psoriasis severity is traditionally measured by the PASI, but PASI scoring suffers from subjectivity, limited sensitivity to smaller lesions and significant variability between clinicians. The authors assembled a large data set of clinical images annotated by dermatologists and trained deep learning models for two tasks: segmenting psoriatic lesions and classifying the intensity of key visual signs (erythema, induration, desquamation). They tested various neural network architectures and found one (MiT_b2) that achieved performance comparable to human experts for classifying visual signs, and another (Xception) that outperformed experts in lesion segmentation. APASI can provide rapid, objective and standardized assessments of psoriasis severity. In addition, tools as APASI could also be valuable for potential integration in teledermatology, where only a correct image-taking would be necessary to do the evaluation, without the need for a real on-site specialist. APASI undeniably offers a considerable amount of benefits, but so far, it also lacks the accuracy of the human eye because of the 2D surface of the image it analyses. This represents the inability to evaluate as accurately as possible the induration and scaling of the lesions in comparison to an expert dermatologist. The data set used to train the AI must also include a comprehensive photography of all Fitzpatrick skin types; moreover, each data set of skin type should include a relatively similar number of images to ensure accuracy of the analysis. Needless to mention is that there is also a need for standardization of the way images are taken, to the light explosion and background, to allow maximum objectivity when the AI software is used. This is not the first time AI tools have been developed for assessing scores or severity in patients with chronic dermatoses. Previous studies included hidradenitis suppurativa [2, 3], vitiligo [4] or atopic dermatitis [5]. Acknowledging the need for diversity of the populations included in the data sets, standardization of imaging techniques, mobile phone integration for automated and reproducible assessment of the disease, integration of AI tools into clinical practice and research settings is likely to improve disease monitoring, enhance treatment evaluation and reduce time associated with manual scoring in clinical practice. Both authors contributed equally to this manuscript. The authors received no specific funding for this work. Nicolas Kluger: AbbVie Finland (lecture), Novartis Finland (Advisory board), Novartis Belgium (lecture), UCB Finland (Advisory board). Lidiya Todorova: AbbVie Bulgaria (consulting, research and events), Novartis Bulgaria (lectures). Data sharing is not applicable to this article as no data sets were generated or analysed during the current study.