AI Demonstrates Diagnostic Accuracy for Conditions Such as Psoriasis, Eczema

Credit: Pexels

Artificial intelligence (AI) and machine learning show potential in diagnostic accuracy, patient outcomes, and reductions in cost for aesthetic medicine, new findings suggest, though limitations related to algorithm bias and data quality may require further research.¹

These findings resulted from a new review of research into the implementation of AI in aesthetic medicine settings. Some of the studies included in this review highlight its applicability for conditions such as psoriasis, seborrheic dermatitis, and atopic dermatitis.

This analysis was authored by Alvin Kar Wai Lee, from EverKeen Medical Centre in Hong Kong, alongside a team of other investigators. Lee et al. noted the potential that artificial intelligence -powered systems have shown in aesthetic medicine.

“This review aims to provide a comprehensive overview of the current state of AI-powered systems in aesthetic medicine, highlighting their benefits and limitations,” Lee and colleagues wrote. “The review will focus on studies that have been graded as Level I or II evidence by the Oxford Centre for Evidence-Based Medicine (CEBM), providing a robust foundation for understanding the role of AI-powered systems in this field.”^1,2

Trial Design

The investigators first highlighted the improvements in diagnostic accuracy that they believe AI has shown in prior data by reducing the likelihood of human error and bias. AI and machine learning, they noted, can help to facilitate personalized treatment plans for patients with various dermatologic conditions that are tailored to individual needs.

Lee and the other researchers noted that AI technologies may be able to promote lowered healthcare costs, given the ability to potentially minimize unnecessary procedures and therapies. They further suggested that improvements can result from this technology’s use in terms of patient engagement, as real-time information can be delivered and patients can then be encouraged to take a more active role in their own healthcare.

In their review, the investigators sought to provide an in-depth analysis of machine learning and AI’s current use in the dermatology space, touching on advantages as well as limitations in aesthetic medicine. The review involved a systematic search conducted across several different databases, with PubMed, MEDLINE, and Ovid used to identify relevant research on AI in cosmetic dermatology and, especially, for photoaging therapy uses.

For their search, the research team implemented specific keywords such as “Deep Learning,” “Artificial Intelligence,” “Machine Learning,” and “Cosmetic Medicine.” The team assessed studies based upon a set of stringent inclusion criteria.

Examples of their criteria for inclusion included utilization of controls, double-blind reviews, sample size, randomization, and objective endpoints. Additionally, the investigators included research that was classified based on the Oxford Center for Evidence-Based Medicine’s evidence hierarchy.

Major Findings

The research team’s findings suggested that use of high levels of diagnostic accuracy in detecting and managing conditions such as psoriasis, skin cancer, seborrheic dermatitis, and acne were observed with machine learning and AI. One study they highlighted was a study looking at a decision support system built on the C5.0 machine learning algorithm, designed with the aim of assisting dermatologists in skin disease diagnoses.

The model covered in this study had been trained using a dataset of 1,000 patients diagnosed with diseases such as psoriasis, eczema, and vitiligo. Following a test on an independent dataset of 200 individuals, it had been concluded that the tool’s diagnostic accuracy was 92.5%, with correct diagnoses in 87% of psoriasis-related cases, 90% of eczema-related cases, and 85% of vitiligo-related cases.

The review’s research team concluded that, overall, AI-driven platforms had been shown to have strong abilities to personalize treatment strategies, diminish diagnostic errors, improve outcomes among patients, and streamline diagnostic workflows.

While these improvements were noted, the team emphasized that there remain significant challenges in the technology’s use. Such challenges include the presence of possible biases and the necessity of larger and more diverse datasets for effective algorithms training.

Despite AI’s potential for immense promise in the cosmetic dermatology and aesthetic medicine spaces, addressing the aforementioned limitations was noted by the investigators as essential to fully realizing such potential.

“The integration of AI and [machine learning] in aesthetic medicine is a promising area of research that has significant potential benefits for patients and clinicians alike,” they wrote. “However, further research is needed to address the limitations and challenges associated with AI adoption in aesthetic medicine.”¹

References

1. ^{Lee, A.K.W., Chan, L.K.W., Lee, C.H., Bohórquez, J.M.C., Haykal, D., Wan, J. and Yi, K.-H. (2025), Artificial Intelligence Application in Diagnosing, Classifying, Localizing, Detecting and Estimation the Severity of Skin Condition in Aesthetic Medicine: A Review. Dermatological Reviews, 6: e70015. https://doi.org/10.1002/der2.70015.}
2. ^{Centre for Evidence-Based Medicine. Oxford centre for evidence-based medicine: levels of evidence (March 2009) [Internet]. Oxford: University of Oxfod; 2009 [cited 2023 Feb 5]. Available from: https://www.cebm.ox.ac.uk/resources/levels-of-evidence/oxford-centre-for-evidence-based-medicine-levels-of-evidence-march-2009.}