The Ministry of Science and Technology on 28th August issued ethical guidelines for artificial intelligence (AI) medical imaging research to standardise the development of such algorithms, as well as the verification of trials and the applications of the technology in clinical practice.

AI medical imaging is a technology that leverages AI to analyse CT, Magnetic Resonance Imaging and ultrasound scans to help doctors detect lesions and enhance the efficiency of diagnosis. 

But it comes with sensitive health information and potential hazards, including privacy issues, algorithmic bias and accountability problems.

The six main principles, detailed in the guidelines, are improving human welfare, fostering fairness, respecting human autonomy, keeping data and privacy secure, keeping things safe and controllable, and making things transparent. Patients’ health interests are prioritised, the guidelines noted.

Guidelines call for informed consent of participants, and participants can withdraw their consent at any time.

The accountability of humans is the only one that will not change. According to the guidelines, doctors are still the ones who have the final say in diagnosis, while AI can generate outputs only as a suggestion.

The guidelines also prevent exaggerated model performance and emphasise the need for algorithmic fairness to minimise diagnostic bias resulting from an unequal distribution of training samples.