At Tohoku University, a system for extracting lesions using images taken with smartphones has been developed.Furthermore, using this system, an AI software for assessing atopic dermatitis complications (AD-AI) was also developed.Yuta Yanagisawa, a fifth-year medical student who can be described as a “Generation Z” who grew up in a digital environment, was the main developer.
There has long been a desire for a system in which patients themselves can take images of their skin eruptions and analyze them using AI. was there.
This lesion extraction system was developed using deep learning, and automatically focuses on the lesion and trims the image even from images that are easily taken with a smartphone or the like.By using this system, it is possible to buffer shooting biases such as information other than rashes and reflections of the background, enabling stable image analysis even for images shot at various shooting distances.
In fact, AD-AI, which was developed using this lesion extraction system, performs preprocessing with AI that extracts lesions, and then performs two-step image analysis with AI that determines disease. Improved detection of diseases (infectious diseases and malignancies) that may be associated with dermatitis.As a result of verifying the task of judging infections and malignant transformation of atopic dermatitis from images, the judgment accuracy of images trimmed by the system and images trimmed by dermatologists was equally high.
In the future, we aim to implement AD-AI into an application that can be used by general patients.If the app makes it possible to take a picture of a skin rash with a smartphone when the patient is concerned and make an AI judgment of whether complications have occurred, it will lead to the realization of self-managed medical care and early detection and diagnosis of diseases. It is hoped that early intervention will become possible.