Automated clinical scoring in dermatology: developing a machine/deep learning algorithm for image guided analysis of psoriatic lesions

Afbeelding_project AI voucher

  • Radboud University/Radboudumc: Marieke Seyger and Mirjam Schaap (Department of Dermatology), Ajay Patel (Radboudumc technology center Deep Learning, AI for Health)
  • External partners: University of Twente (TechMed Center)


Psoriasis is a common chronic inflammatory skin disease, with a prevalence of 2-3% of the Western population. In the Netherlands, over 500.000 people suffer from psoriasis. The skin disease is characterized by raised, red, scaling lesions, which can be present all over the body. These lesions are often a burden to patients since they can be visible, itchy, or painful. Due to its chronic nature, psoriasis can cause a remarkable decrease in the quality of life of patients, with negative effects on self-esteem, relations, family planning and occupations. It is therefore important to adequately treat patients with psoriasis. This is possible with topical treatments, phototherapy, conventional systemic treatments, biologics, and small molecule inhibitors. In order to determine which treatment should be commenced, the assessment of severity of psoriasis is crucial.

Clinical severity scoring


In daily clinical practice and clinical research, psoriasis severity is often measured by the PASI (Psoriasis Area and Severity Index) score. In this score, the extend of the affected area in each body region is estimated by the physician. In addition, redness (erythema), scaling (desquamation), and thickness (induration) of the lesions are scored on a 5-points scale. Consequently, the PASI score is a subjective (intra-and interobserver variability) and time-consuming severity score. Nevertheless, it is globally used by physicians and pharmaceutical industries to determine treatment efficacy.

Innovation and Aim

It would highly benefit patients, physicians, and the pharmaceutical industry if the PASI score would be objectively and automatically performed because it would improve treatment evaluation and stimulate shared decision making by patients and physicians. In this project, we aim to develop a deep learning algorithm which is able to objectively determine the severity of psoriasis.

Collaboration and Future

With this project, a new AI research line has been established, in which the Dermatology Department of the Radboudumc, the Radboudumc Technology Center Deep Learning (AI for Health), and the TechMed Centre, Robotics and Mechatronics group of the University of Twente collaborate. We are exploring the field of machine/deep learning to develop an algorithm for image guided analysis and classification of psoriatic lesions. The outcomes of this project are a proof of concept for the preparation of a grant application and further collaboration. Our group finally aims to develop a handheld automated PASI scoring device which can be used by psoriasis patients at home and will improve their quality of care.