Artificial intelligence (AI) can be defined as ‘the science of making machines do things that would require intelligence if done by ‘people’1 or simply as the ability of computers to simulate intelligent human behaviour. Machine learning is a branch of AI that uses algorithms and statistical models that can learn from data. Deep learning is a specific form of machine learning. AI can refer to either machine learning or deep learning2.
The current evidence base for AI interventions in Dermatology
The COVID-19 pandemic has accelerated adoption of digital technologies to reduce the need for face-to-face contact with patients and created a sense of urgency to implement AI in dermatology. Despite this rapidly advancing field and the commercial drive to integrate AI algorithms into clinical practice as soon as possible, the current evidence-base for effectiveness of AI interventions in dermatology is weak3.
Most AI interventions in dermatology focus on differentiating between benign and malignant skin lesions with a particular emphasis on melanoma diagnosis2. However, other applications exist including monitoring of inflammatory skin disorders e.g. psoriasis, atopic eczema, acne vulgaris, leg ulcer assessment, and nail disease1although their development is still in the early stages.
Many skin smartphone apps claim that they can diagnose, monitor and treat a range of skin disorders but most have not been appropriately investigated in an intervention study setting, and a systematic review of algorithm-based smartphone apps used for skin cancer diagnosis identified serious flaws4.
The accuracy of AI algorithms for skin cancer diagnosis may also be overestimated as studies are frequently conducted in artificial settings which are not reflective of standard clinical practice, for example, using a retrospective database of images, sometimes without supplementary clinical information, excluding atypical presentations, using highly selected patient groups (in particular excluding skin of colour images), or limiting study cases to those already selected for excision5.
Algorithms are often evaluated as stand-alone systems, rather than as diagnostic aids to inform clinician decision–making.
Patient pathways in dermatology are complex and to robustly evaluate AI interventions in current pathways requires clinically-led research studies. An AI intervention to triage skin cancer referrals at the interface between primary and secondary care requires both robust evaluation of diagnostic accuracy at this interface, but also a thorough evaluation of effectiveness i.e. whether it is possible to safely reduce the volume of suspected skin cancer referrals freeing resources in secondary care without causing a significant increased risk to patients through missing a malignancy.
Developing AI in Dermatology
The BAD encourages the development of AI technologies in areas which clearly address an unmet need; by improving disease management and the quality of care and by enhancing patient experience without compromising safety. When considering the use of AI it is important to consider the problem you are trying to solve, and whether artificial intelligence is the right solution. You should be able to explain why you are choosing AI, what additional ‘intelligence’ do you need, and why AI is the solution 6.
Clinicians / developers should consider:
The problem you're trying to solve is associated with a large quantity of data which an AI model could learn from.
Analysis of that data would be on a scale so large and repetitive that humans would struggle to carry it out effectively
You could test the outputs of a model for accuracy against empirical evidence i.e standard care pathways.
Model outputs would lead to problem-solving (or provide a solution) in the real world
The data in question is available - even if disguised or buried - and can be used ethically and safely.
The BAD has set up an AI Working Party Group. For further information click here
For any other queries please contact:
1 Proposal for the Dartmouth Summer Research Project on Artificial Intelligence” (1955).
2 Du-Harpur X, Watt FM, Luscombe NM, et al. What is AI? Applications of artificial intelligence to dermatology. Br J Dermatol 2020 doi: 10.1111/bjd.18880
3 Liu X FL, Kale AU, et al. A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis. The Lancet Digital Health 2019;1(6):E271-E97.
4 Freeman K, Dinnes J, Chuchu N, et al. Algorithm based smartphone apps to assess risk of skin cancer in adults: systematic review of diagnostic accuracy studies. BMJ 2020;368:m127. doi: 10.1136/bmj.m127 [published Online First: 2020/02/12]
5 Ferrante di Ruffano L, Takwoingi Y, Dinnes J, et al. Computer-assisted diagnosis techniques (dermoscopy and spectroscopy-based) for diagnosing skin cancer in adults. Cochrane Database Syst Rev 2018;12:CD013186. doi: 10 1002/14651858.CD013186 [published Online First: 2018/12/07]
6 Artificial Intelligence: How to get it right https://www.nhsx.nhs.uk/media/documents/NHSX_AI_report.pdf