AI-Driven Psoriasis Care Evolves Toward Multimodal Diagnosis and Personalized Treatment, Bibliometric Analysis Reveals

A comprehensive bibliometric analysis spanning two decades of intelligent psoriasis (search)Search disease research reveals a field undergoing rapid transformation—from rudimentary image-based classification to sophisticated, multimodal AI systems capable of guiding both diagnosis and personalized treatment. Published in Clinical, Cosmetic and Investigational Dermatology, the study systematically maps the knowledge structure, research hotspots, and evolutionary trajectory of AI-assisted psoriasis care between 2005 and 2025.

Drawing from the Web of Science Core Collection, researchers analyzed 122 papers on intelligent diagnosis and 81 papers on intelligent treatment using VOSviewer, CiteSpace, and Pajek software. The findings paint a picture of a discipline maturing in distinct phases, with deep learning and machine learning now firmly established as core technologies.

Three Phases of Intelligent Diagnosis

The volume of intelligent diagnosis literature progressed through three identifiable stages. Between 2005 and 2017, annual publications remained sparse, never exceeding five papers per year. From 2018 to 2023, output accelerated markedly as machine-based diagnostic research emerged as a hotspot. The field reached its zenith between 2024 and 2025, with Chinese publications alone accounting for approximately one-third of total global output in 2024.

Co-citation analysis identified Andre Esteva and colleagues' landmark 2017 paper on deep neural network-based skin cancer classification as the most cited work, with 16 citations. That study demonstrated that large-scale convolutional neural network models could achieve diagnostic accuracy comparable to dermatologists.

Keyword emergence analysis tracked the field's shifting priorities: "images" first appeared in 2015 and peaked between 2020–2021; "machine learning" emerged in 2017 and dominated from 2022–2025; "deep learning" surfaced in 2020 and reached peak frequency between 2023–2025. The analysis indicates that machine learning-based intelligent diagnosis using ultrasound images has become a prominent recent focus.

Five keyword clusters define the research landscape: differential diagnosis based on image and computer processing (green cluster), pathological mechanisms of psoriasis (search)Search disease (blue cluster), AI applications for disease management and risk assessment (red cluster), AI for dermatological diagnosis (pink cluster), and epidemiological analysis. Density mapping revealed that AI-based psoriasis subtyping and classification represent the most mature research areas, while clinical management and diagnostic risk prediction retain substantial room for further exploration.

Intelligent Treatment: From Drug Discovery to AI-Optimized Protocols

Intelligent treatment research followed a different trajectory. After a prolonged period of low output from 2005 to 2020, publications surged, peaking at 25 papers in 2025—nearly double the 2024 figure. The most highly cited works included April W. Armstrong's 2020 review on psoriasis (search)Search disease pathophysiology and treatment, Christopher E.M. Griffiths' 2021 Lancet paper on the IL-17 (search)Search target/IL-23 (search)Search target immune axis, and T. Fredriksson's pioneering 1978 study establishing oral retinoid therapy and the PASI scoring system—each cited eight times.

These three works collectively trace the field's evolution from effective treatment and targeted intervention to intelligent management. The Griffiths review clarified the central role of the IL-17 (search)Search target/IL-23 (search)Search target pathway, driving a therapeutic revolution in targeted biologics. Armstrong's commentary further proposed stratified treatment strategies and outlined AI's potential for personalized precision therapy.

Keyword analysis identified four clusters: fundamental pathological mechanisms (blue), intelligent treatment approaches and population-specific research (red), therapeutic drug formulation development (green), and drug delivery technologies (yellow). Nanocarrier-mediated targeted delivery of methotrexateView drug details and AI-enabled clinical tools for psoriasis (search)Search disease typing emerged as current research hotspots.

Density mapping showed that R&D for psoriasis (search)Search disease biotherapies based on formulations and drug delivery systems has reached maturity, while AI-driven research—reflected in keywords like "artificial intelligence" and "digital health"—has yet to peak and is poised for rapid growth.

The Closed-Loop Vision

The study articulates an integrated vision where intelligent diagnosis and treatment form a closed-loop system. Multimodal diagnostic data—including images and genomic information—feed into AI classification algorithms that generate personalized treatment plans. AI-optimized therapeutic interventions, whether through refined drug delivery or adjusted phototherapy parameters, are then monitored continuously. If deterioration occurs, the system adjusts the treatment plan immediately.

"By leveraging AI, this approach creates a closed-loop system integrating intelligent diagnosis and treatment, significantly enhancing the efficiency of intelligent psoriasis (search)Search disease diagnosis and therapy," the authors write.

Global Collaboration Patterns

Among 40 countries contributing to intelligent diagnosis research, China, the USA, India, and England lead in publication volume. However, direct collaboration between China and the USA remains limited. For intelligent treatment, India, China, the USA, and Germany top the rankings, with India, China, and the USA maintaining relatively close direct connections. Countries such as Italy, Germany, and England lack robust collaboration in this area.

Persistent Research Gaps

Despite progress, significant gaps remain. Intelligent diagnosis lacks standardized imaging protocols and model validation. Data models underrepresent dark-skinned populations, including Africans and Latinos. Most studies stop at diagnostic severity scores without predicting biologic response, and samples of rare subtypes—pustular and erythrodermic psoriasis (search)Search disease—are scarce. Existing treatment data derive from static, single-time-point images, lacking longitudinal tracking of relapses or remissions. Current accuracy rates come from retrospective data without validation in randomized controlled trials.

The study notes that psoriasis (search)Search disease affects an estimated 1–3% of the global population, with 70–90% of patients experiencing moderate to severe pruritus, 50% experiencing nail involvement, and up to 30% developing associated arthropathy. Psychological burden is substantial, with 59.1% of patients affected by social discrimination.

The authors conclude that future research should prioritize strengthening international collaboration and addressing these research gaps to fully realize the potential of intelligent systems in enhancing psoriasis (search)Search disease diagnosis and treatment.