RANZCR draft guidelines calls for “correct use” of AI and machine learning

The Royal Australian and New Zealand College of Radiologists (RANZCR) is on a mission to improve decision-making transparency, data privacy and ethics in the industry’s use of AI and machine learning. 

• Safety: Patient safety and quality of care should be the first and foremost consideration in the development, deployment or utilisation of AI or machine learning, with an evidence base to support it.
   
• Avoidance of bias: As AI and machine learning systems are limited by their algorithmic design and the data they have access to, they are prone to bias. To minimise bias, RANZCR suggests that the same standard of evidence used for other clinical interventions be applied when regulating machine learning systems and AI tools, with their limitations transparently stated.
   
• Transparency and explainability: As machine learning and AI can produce results which are difficult to interpret or replicate, RANZCR suggests that a doctor must be capable of interpreting how a decision was made and weighing up the potential for bias.
   
• Privacy and protection of data: Storing a patient’s data must be done securely and in line with relevant laws and best practice. RANZCR suggests that patient data isn’t transferred from a clinical environment of care without the patient’s consent or approval from an ethics board. Where data is transferred or otherwise used for AI research, it should be de-identified in a way that the patient’s identity cannot be reconstructed.
   
• Decision making on diagnosis and treatment: While machine learning and AI can enhance decision-making capacity, RANZCR suggests that final decisions on patient care are recommended by a doctor with due consideration given to the patient’s current state, history and preferences. 
   
• Liability for decisions made: The liability for decisions made about patient care rests with the responsible medical practitioner, while the potential for shared liability needs to be identified and recorded upfront when researching or implementing machine learning and AI.
   
• Application of human values: As machine learning and AI tools are programmed to operate in line with a “specific world view”, RANZCR says it is the role of the doctor to apply humanitarian values (from their training and the ethical framework in which they operate) and consideration of that patient’s personal values to any circumstances in which these technologies are used in medicine.
   
• Governance: RANZCR says that machine learning and AI are fast-moving technologies with the potential to add great value but also do harm. It suggests that a hospital or practice using these technologies have accountable governance committees in place to oversee implementation and ensure compliance with ethical principles and standards. 

RANZCR President Dr Lance Lawler said these principles are “the first of their kind” devised by a healthcare body and that they aim to ensure the protection of patient data, balanced with the application of humanitarian values. 

“New technologies such as AI are having a huge impact on healthcare, with enormous implications for both health professionals and patients. They have the ability to help doctors work in a more time-efficient and effective manner and – ultimately – provide even greater treatment for patients,” he said. 

Lawler said these guiding principles are necessary as the way radiology adapts to AI has a flow-on effect for patients and other healthcare professionals. 

“The agreed principles will, when established, complement existing medicinal ethical frameworks, but will also provide doctors and healthcare organisations with guidelines regarding the research and deployment of machine learning systems and AL tools in medicine,” he said.  

“There are lots of hype and misinformation around AI; it is important to look beyond that and concentrate on… how we can best use it for the maximum benefit of patients.”  

The principles are out for public consultation, with submissions due before April 26.