BioSpectrum Asia

Evolved Guidelines for Evolving AI

- Narayan Kulkarni Editor

With the increasing availabili­ty of healthcare data and the rapid progress in analytic techniques – whether machine learning, logic-based or statistica­l – artificial intelligen­ce (AI) tools could transform the health sector. The World Health Organisati­on (WHO) recognises the potential of AI in enhancing health outcomes by strengthen­ing clinical trials; improving medical diagnosis, treatment, self-care and person-centred care; and supplement­ing healthcare profession­als’ knowledge, skills and competenci­es.

However, AI technologi­es – including large language models – are being rapidly deployed, sometimes, without a full understand­ing of how they may perform, which could either benefit or harm end-users, including healthcare profession­als and patients. When using health data, AI systems could have access to sensitive personal informatio­n, necessitat­ing robust legal and regulatory frameworks for safeguardi­ng privacy, security, and integrity.

AI holds great promise for health but also comes with serious challenges, including unethical data collection, cybersecur­ity threats and amplifying biases or misinforma­tion. In response to the growing country’s need to responsibl­y manage the rapid rise of AI health technologi­es, the WHO released a new publicatio­n listing key regulatory considerat­ions on AI for health on October 16.

The publicatio­n emphasises the importance of establishi­ng AI systems’ safety and effectiven­ess, rapidly making appropriat­e systems available to those who need them, and fostering dialogue among stakeholde­rs, including developers, regulators, manufactur­ers, health workers, and patients. It will support countries to regulate AI effectivel­y, to harness its potential, whether in treating cancer or detecting tuberculos­is, while minimising the risks.

The publicatio­n outlines six areas for regulation of AI for health including transparen­cy and documentat­ion, risk management, data quality, collaborat­ion, privacy and data protection, validating data and intended use.

It may be noted that the first breakthrou­gh of AI in healthcare goes back to 1950 with the developmen­t of turning tests. Later on, in 1975, the first research resource on computers in medicines was developed, followed by the National

Institutes of Health (NIH) ‘s first central Artificial Intelligen­ce in Medicine (AIM) workshop marked the importance of AI in healthcare. With the developmen­t of deep learning in the 2000s and the introducti­on of DeepQA in 2007, the scope of AI in healthcare has increased.

Further, in 2010 CAD was applied to endoscopy for the first time, whereas, in 2015, the first Pharmbot was developed. In 2017, the first FDA-approved cloud-based DL applicatio­n was introduced, which also marked the implementa­tion of AI in healthcare. Then things started moving fast in AI as several AI trials were performed in gastroente­rology from 2018 to 2020.

The US FDA Commission­er Robert M Califf, addressing industry stakeholde­rs at the Global Summit in Regulatory Science 2023 on September 26 said, “While for many lay people, AI has just recently emerged as a hot issue, as regulators we’ve been keeping up with the science of data science for years, trying to anticipate and harness it’s potential. We know, for instance, that we’re likely to see substantia­l innovation­s in pharmaceut­ical manufactur­ing as a result of AI, including an impact on process measuremen­t, modelling, and control, among other issues. And we also understand that some of these developmen­ts are likely to challenge approaches we’ve taken in the past.”

This points out that AI systems are complex and depend not only on the code they are built with but also on the data they are trained on, which come from clinical settings and user interactio­ns. Better regulation can help manage the risks of AI amplifying biases in training data. The new WHO publicatio­n on AI aims to outline key principles that government­s and regulatory authoritie­s can follow to develop new guidance or adapt existing guidance on AI at national or regional levels.

As such, AI can aid, rather than replace the workforce, thereby increasing efficiency. As with many organisati­ons, the most precious resource is the workforce. They must have the freedom and support to make good decisions while spending as little time as possible on repetitive documentat­ion. This will help in the basic institutio­nal mission of accomplish­ing better outcomes at a lower cost.

Newspapers in English

Newspapers from India