If you would like to submit a blog post for consideration, please email [email protected]
[4 min read] How will artificial intelligence benefit GPs in skin cancer medicine?
Yes, the robots are coming! This month I share an article published in PLOS One in March this year. There is now a steady stream of research papers showing how Artificial Intelligence (AI) is effective in accurately diagnosing skin cancers.
This technology – it is all based on what are called “convolutional neural networks (CNNs)” – is developing very rapidly. It is important to appreciate that AI is not simply ‘very powerful computers’ doing heaps of calculations very quickly. Yes, AI requires these computers, but the CNNs inside them actually do “learn”. They do, literally, learn. The CNNs are taught by feeding them information – for example (in this case) labelled images of acral naevi and melanomas.
This paper shows that the AI was as good as an expert in diagnosing acral lesions – hitting around 80 per cent. And, the AI is much better than the non-expert (around 65 per cent).
Clearly the implication is that we will have artificial intelligence that is as good as an expert dermoscopist, available to us as GPs. For me, the place in the clinical workflow is obvious. As GPs, we will have immediate and easy access to a simple technology that gives us very robust advice on what a lesion actually is – and most importantly whether we need to biopsy the lesion, how to biopsy it, and so on.
Professor David Wilkinson
Read more from Professor David Wilkinson on recent research:
- The problem of dysplastic naevi
- Managing Spitz lesions in children
- Managing patient anxiety while undergoing skin cancer excision
- Managing subungual melanoma in situ in general practice
- Diagnostic accuracy for skin cancer among non-doctors