A group of dermatologists has created an artificial intelligence (AI) model that enables people with atopic dermatitis (AD) to recognize the difference between eczema and skin lesions caused by a specific form of blood cancer, as well as complications from bacterial or viral infections.
The AI model is described in depth in a study published on January 11th, 2023, in the Journal of Dermatological Science.
Almost 12% of people have AD, a chronic condition that frequently originates in childhood. Individuals with AD frequently have skin immune barriers that are suppressed, which lowers their defenses against microbial pathogens and increases the risk of eczema complications from bacterial or viral infections.
These could include Kaposi varicelliform eruption, impetigo, and herpes simplex (eczema herpeticum).
As the symptoms’ appearance on the skin closely resembles AD itself, it can be difficult for patients to determine if AD has caused any of these consequences. Besides that, mycosis fungoides, a type of blood cancer that results in skin lesions, can display symptoms that are similar to AD and may co-exist with AD. Certain AD drugs could make infections or mycosis fungoides worse.
Complications and malignant diseases must be correctly and promptly diagnosed to receive effective therapy and achieve better results. Due to the similarity of symptoms, people may not usually detect any odd symptoms and consult a doctor as soon as possible.
To solve this problem, the scientists used non-standard images of AD, impetigo, mycosis fungoides, herpes simplex, and Kaposi varicelliform eruption to train their convolutional neural network (CNN) model.
The accuracy of the AI’s diagnosis was then compared with a group of non-standard images that dermatologists manually cropped and annotated with diagnostic information. They discovered that the diagnostic accuracy of their algorithm was practically on par with the manually evaluated image set.
The team is currently working on a smartphone app powered by AI that will interpret their approach and allow patients to remotely control their skin conditions using only their phone’s camera. To enhance the functionality of the software, they are also conducting experiments with AD patients.
A dermatologist would of course be able to spot the difference, but it is incredibly impractical for an AD patient to visit a dermatologist every day. If only there were some handy, low-cost mechanism that replicated that dermatologist’s knowledge and could be used during a patient's daily regimen of checking their skin.
Yuta Yanagisawa, Study Co-Author and Researcher, School of Medicine, Tohoku University
The research team is confident that this technology will enable patients with skin problems to effectively and efficiently manage their symptoms, leading to improved health outcomes.
Yanagisawa, Y., et al. (2023) Convolutional neural network-based skin image segmentation model to improve classification of skin diseases in conventional and non-standardized picture images. Journal of Dermatological Science. doi:10.1016/j.jdermsci.2023.01.005