Background: Psoriasis requires long-term monitoring, yet conventional tools like the Psoriasis Area and Severity Index (PASI) are time-consuming and prone to inter-rater variability. Artificial intelligence (AI) may overcome these limitations by providing remote and objective assessments.
Objectives: To evaluate the feasibility and performance of AI-based smartphone imaging for remote monitoring of psoriatic lesion response to treatment.
Methods: In a 16-week, double-blinded, placebo-controlled trial, 26 patients with mild-to-moderate psoriasis received guselkumab or placebo. Using a color-calibrated smartphone medical platform, patients captured daily home images of target lesions. Physicians assessed PASI, Target Lesion Score (TLS) and clinical images. Machine learning quantified and predicted erythema, scaling, and to generate an AI-derived TLS (aiTLS).
Results: aiTLS showed strong agreement with physician-assessed TLS (r=0.82, p<0.001). From week 8, AI...
Show moreBackground: Psoriasis requires long-term monitoring, yet conventional tools like the Psoriasis Area and Severity Index (PASI) are time-consuming and prone to inter-rater variability. Artificial intelligence (AI) may overcome these limitations by providing remote and objective assessments.
Objectives: To evaluate the feasibility and performance of AI-based smartphone imaging for remote monitoring of psoriatic lesion response to treatment.
Methods: In a 16-week, double-blinded, placebo-controlled trial, 26 patients with mild-to-moderate psoriasis received guselkumab or placebo. Using a color-calibrated smartphone medical platform, patients captured daily home images of target lesions. Physicians assessed PASI, Target Lesion Score (TLS) and clinical images. Machine learning quantified and predicted erythema, scaling, and to generate an AI-derived TLS (aiTLS).
Results: aiTLS showed strong agreement with physician-assessed TLS (r=0.82, p<0.001). From week 8, AI detected reductions in erythema and scaling in the guselkumab group (p < 0.05), across both at-home and in-clinic images. Baseline comparisons showed no significant differences between clinic and home image quality.
Limitations: Limited by a small cohort, single-lesion focus, and lack of diverse skin tones or severe phenotypes.
Conclusions: AI-based smartphone imaging reliably quantifies psoriasis severity, matching in-clinic, physician scores. This approach enables scalable remote monitoring in clinical trials and routine practice.
Show less
