Prof. Syed Munir Khasru
South China Morning Post (Hong Kong)
October 15, 2023
_____________________
Artificial intelligence in healthcare is increasingly an important topic, and one that will be discussed at the World Health Summit next week in Berlin. AI has the potential to improve patient outcomes but it also poses risks – from data collection and use to biases that can skew patient outcomes.
By 2025, more than US$30 billion is expected to be invested into AI for healthcare, a reflection of the growing trust in AI-driven healthcare solutions.
A major driver of this investment is the imperative to make better use of the vast amount of healthcare data available. Through data analysis and tools such as virtual health assistants, AI can help to significantly cut healthcare costs.
Also becoming popular are wearable health devices and AI-backed treatment plans that focus more on individual needs and early healthcare measures. In China for instance, such devices from companies including Huawei Technologies monitor vital health statistics, while platforms like Ping An Good Doctor offer AI-driven preliminary diagnoses. China’s AI healthcare evolution also includes innovations such as Infervision’s AI algorithms for early disease detection through medical imaging, and Tianji Robot’s precision in robot-assisted orthopaedic surgery.
And it’s not just China. The integration of AI into healthcare has also seen significant advances in Europe and the United States. Countries including Britain, France and Germany are spearheading public-private initiatives, such as digital innovation hubs in healthcare.
The US, home to pioneers like DeepMind, which trained its AI algorithms to detect more than 50 eye diseases from medical scans, emphasises cutting-edge research with regulatory vigilance. The Food and Drug Administration’s (FDA’s) guidelines for AI-based medical devices and the Health Insurance Portability and Accountability Act’s (HIPAA’s) data privacy mandates demonstrate this commitment.
A study in the journal Nature found that AI might be as good as, or better than, doctors at detecting breast cancer from mammograms. AI is also game-changing in predicting the spread of disease. During the Covid-19 outbreak, for instance, a Canadian company, BlueDot, used AI to quickly track the spread of the virus, using news and flight data.
Surgery also has seen big changes with AI and robot-aided procedures. For example, the Da Vinci system from California-based Intuitive Surgical, used in more than 10 million surgeries, gives better results and quicker patient recoveries.
Meanwhile, Japan is addressing the needs of its ageing population by integrating AI in elderly care with innovations like the Robear robot.
The integration of AI into healthcare offers both transformative potential and challenges that demand robust regulatory oversight. Ensuring patient safety, data security and addressing ethical dilemmas remain paramount. Misinterpretations by AI, for instance, can lead to misdiagnoses with grave implications.
A classic example is IBM’s now-defunct company Watson Health, which promised to transform medical care through AI, particularly in cancer treatment. By diving into massive amounts of health data, Watson would suggest the most suitable treatments to doctors, acting as a super-smart assistant.
But as time went on, Watson’s advice was found to have failed to hit the mark, even incorrect, which risked faulty medical decisions. The Watson experience served as a lesson: while technology can be a powerful ally in healthcare, blind trust without human oversight can be hazardous.
Most countries are grappling with crafting regulations that balance innovation with safety. America’s FDA and HIPAA are the guardians of medical AI regulation; the EU’s General Data Protection Regulation (GDPR) sets tight controls over data processing, with rigorous provisions for health data.
While nations such as Canada and Japan are incorporating AI guidelines into existing medical device regulatory structures, others are at the nascent stage and still evaluating the best approaches. The rapid evolution of AI technology means it often outpaces regulation. Transparency in AI algorithms, essential for building trust, also remains elusive, making regulation even more challenging.
One significant struggle is in addressing AI’s potential bias, which can lead to skewed healthcare outcomes. Back in 2019, a Science journal article found that a widely used algorithm was showing racial bias – it recommended more white patients than black ones for further health treatments.
Such instances show that AI is susceptible to inheriting biases in its training data. By identifying such biases or inaccuracies in AI predictions, humans can improve the system’s accuracy, fairness and reliability.
Moreover, healthcare is not just about diagnosing and treating diseases; it also encompasses vital human elements such as trust, understanding and the ability to address the cultural, socio-economic and psychological nuances of individual patients. In mental health services, for example, while AI can play a role in the initial diagnosis or monitoring, a robot cannot replicate the deep understanding, empathy and trust established in a therapist-patient relationship.
The challenges that AI presents do not negate its potential. Instead, they underscore the necessity of a synergistic approach. AI’s strengths lie in its data-processing capabilities, pattern recognition and efficiency. Humans bring experiential wisdom, ethical discernment and empathetic understanding. Combining AI with human skills is vital, not only for accuracy, but to maintain the essence of personal care in healthcare.
The future of healthcare requires the integration of AI tools without losing sight of the human-centric essence. This delicate balance, while challenging, is vital. It is the path to a healthcare future where technology and humanity coalesce, redefining the contours of quality healthcare in the 21st century.
