The college’s consensus statement points out that doctors and patients need to know when and how to use AI safely and appropriately.
The TGA must create and regulate specific labelling requirements for dermatology AI-based software as medical devices, says a new consensus statement from the Australasian College of Dermatologists.
While the TGA has established labelling requirements applicable to all medical devices, including Software as a Medical Device (SaMD), to date there are no separate specifications which highlight medical device labelling requirements in the context of dermatology Al software.
In dermatology, only a few Al-based software solutions have been approved by the major regulatory bodies such as the FDA in the US, the UK’s MHRA, European Notified Bodies or Australia’s TGA.
However, since capturing an image of a patient’s skin can be as easy as taking a photo on a smartphone, many of the available Al-based dermatology software products are general public-facing, such as smartphone apps, and have eluded regulatory scrutiny.
“End users such as health professionals and the general public need relevant and transparent information about the software,” said ACD President, Dr Adriene Lee on releasing the consensus statement.
Lead author of the ACD consensus statement, Professor Victoria Mar, a member of the college’s Digital Health Committee, said there was “substantial evidence” on the vulnerabilities of AI, particularly when it was used in patients or contexts different from the dataset in which the AI had been trained.
“In dermatology, this includes differences in skin colour, sex and age distribution, in image quality, and in the algorithm design,” said Professor Mar.
“The availability of smartphone apps to self-diagnose skin cancer, when the AI is not yet at the point where it is sufficiently accurate for clinical, let alone direct-to-consumer use, and the lack of readily available information about their risks and limitations is concerning.
“So that we can use the AI appropriately and interpret the results correctly, it is critical that we can understand how the AI was developed and the data used to train and test it.
“Like any product we use, the operating manual or product information should clearly explain when and how to use it safely and appropriately.”
The college’s Digital Health Committee evaluated which labelling items should be considered as a minimum requirement for AI-based SaMD. Consensus was achieved across 10 specific labelling domains.
The panel agreed that the labelling should be the same for all TGA categories and risk classes of AI-based SaMD.
“Acknowledgement of the intended user is critical to ensure AI outputs provide appropriate and interpretable data to the end-user and to ensure the product falls within the correct risk classification,” Mar et al wrote in the consensus paper.
“Although our [expert panel] voted for the inclusion of number and sources of images, use of synthetic images, breakdown on skin phototype/race, breakdown on skin diagnosis categories and description of gold standard determination in the minimum labelling requirements of AI-based SaMD, some aspects remained contentious, namely breakdown on sex, age and the image capturing device.
“After discussions, the EPM agreed that the former two should generally be required to assure fair and trustworthy AI output.
“Finally, as we have learnt from pharmaceutical product development, the proper recording of updates and adverse events is essential in ensuring trustworthy and transparent AI.
“This study provides critical evidence for setting labelling standards by the Therapeutic Goods Administration to safeguard patients, health professionals, consumers, industry, and regulatory bodies from AI-based dermatology SaMDs that do not currently provide adequate information about how they were developed and tested.”
