(AI) Skin in the Game
An artificial intelligence-based system can assist with tricky dermatopathology diagnoses
Michael Schubert | | Quick Read
When devastating diseases manifest themselves, patients want answers – fast. Unfortunately, those are not always forthcoming, especially with conditions such as epidermolysis bullosa acquisita (EBA), an autoimmune blistering disease in which antibodies attack type VII collagen in the skin’s epidermal basement membrane.
Currently, the disease can take months or years to diagnose, partly because diagnostic antibody patterns can be difficult to identify on a skin biopsy slide. To address the problem, dermatologist Joost Meijer and colleagues designed an AI-based system to spot the serrated, U-shaped immunodeposition patterns characteristic of the disease (1).
The system yielded sensitivity and specificity of 89.3 percent – outperforming many doctors – but the work is still ongoing. “It will take a year to collect the data of new skin biopsies, which will be able to validate the system,” said Meijer (2). “Hopefully, we will then have a quicker and easier way to diagnose EBA and prevent the sometimes debilitating scarring.”

- C Shi et al., Am J Pathol, [Online ahead of print] (2021). PMID: 34197776.
- University of Groningen (2021). Available at: https://bit.ly/3feYUho.
