skinzar

Psoriasis is a chronic, relapsing skin disease that presents with inflammatory red and scaly plaques on the skin surface (Langan et al., 2012), and affects 2-3% of the world’s population according to the World Psoriasis Day consortium (Psoriasis,org, 2021). Accurate assessment of the extent and severity of psoriasis is essential for quantifying the improvement of the disease and the effectiveness of treatment regimens during clinical trials. The Psoriasis Area and Severity Index (PASI) score is a quantitative score that measures the severity of psoriatic lesions based on subjective parameters. The PASI score measures severity of psoriatic lesions based on area of psoriasis coverage and the appearance of the plaques; Erythema, induration and scaling. One main component of the PASI score is the area percentage, which is divided into seven grades as shown in Table 1. The PASI score has been considered the gold standard for decades, and has been used in clinical trials to demonstrate percentage improvement. Currently, drug approval relies on this score changing by 75, also known as a Delta PASI 75. Additionally, between patient visits, the dermatologist will calculate a PASI score to help them remember how the illness is evolving (Gomez et al., 2002), and adapt treatment accordingly. 

Many authors have argued that the PASI has significant flaws. One drawback is the fact that it is calculated by dermatologists and takes more than 30 minutes per patient (Berth-Jones et al., 2006), which is a problem considering the shortage of dermatologists in less developed and developing countries (Li et al., 2020). Despite this, the PASI score is still used in dermatological practice and in the scientific literature. More worryingly though, there is a distinct level of subjectivity and low intra- and inter-reproducibility, highlighting the need for an automated system (Fink et al.,, 2018). A study looking at reproducibility of PASI scores between three dermatologists found that there was only a 51% agreement in face-to-face scores and an average of 63% agreement when the dermatologists were shown digital images (Singh et al., 2011). Additionally, the area scoring system of seven categories allows for little variation between area percentage grades, causing the accurate estimation of this by eye to be more difficult. As shown in Table 1, the final PASI score is heavily dependent on a correct area score, which is multiplied twice in the calculation. Moreover, one group of authors suggested that the impact of the area scoring component should be increased (Jacobson and Kimball, 2004), as it correlates better to the patient’s self-assessment of their condition. However, increasing the weighting of the area score only magnifies the subjectivity and problems of reproducibility discussed above. With this in mind, it is important that there is a shift towards a more automated system, using deep learning (DL) models, in order to minimise subjectivity and maximise accuracy of the area score and subsequent PASI score.

Attempts have been made to devise a new scoring system such as the Copenhagen Psoriasis Severity Index (CoPSI), which omits an area score entirely, and produced an interuser reliability of 83%, further highlighting the subjectivity of an estimation-based system (Berth-Jones et al., 2008). Maybe it’s time we collaborated with AI to produce better methods that are not so unreliable and based on area guessing!