Artificial Intelligence systems which are used to diagnose skin cancer have been found to not be as accurate on people with dark skin according to research.
The new systems, which are said to recognise images by learning algorithms to determine different skin cancers just as successfully as human experts, could falter when trying to detect skin cancer on patients with dark skin tones.
Researchers have said that more needs to be done to train these systems to ensure all patient care is the same, after finding very few image databases that contained information on ethnicity or skin type to train AI systems for skin cancer diagnosis.
Dr David Wen, first author of the study from the University of Oxford, said: “You could have a situation where the regulatory authorities say that because this algorithm has only been trained on images in fair-skinned people, you’re only allowed to use it for fair-skinned individuals, and therefore that could lead to certain populations being excluded from algorithms that are approved for clinical use.
“Alternatively, if the regulators are a bit more relaxed and say: ‘OK, you can use it [on all patients]’, the algorithms may not perform as accurately on populations who don’t have that many images involved in training.”
Problems including missing treatable cancers, giving incorrect diagnosis, or risking avoidable surgery could arise from the lack of data around darker skin tones for the AI system.
Dr Wen and colleagues were able to identify 21 open-access datasets for skin cancer images. From 106,950 images, only 2,436 recorded their skin type and only 10 of those images were from people who described themselves as having ‘brown skin’ and 1 person recorded themselves as having ’dark brown or black skin’.
Dr David Wen and his team said: “No images were from individuals with an African, African-Caribbean or South Asian background.
“Coupled with the geographical origins of datasets, there was massive under-representation of skin lesion images from darker-skinned populations,”
