What is narrow AI and how can it improve healthcare?

You’ve almost certainly heard of artificial intelligence (AI) and perhaps even generative AI (GenAI), but what about narrow AI? Unlike GenAI, narrow AI doesn't aim to replicate human intelligence in its entirety. Instead, it functions within predefined rules, demonstrating proficiency in specific tasks. Alongside the recent surge of interest in AI, narrow AI has been quietly reshaping several industries for many years.

In healthcare, for example, narrow AI’s application has already led to notable improvements in patient care by assisting healthcare professionals, augmenting their skills and expertise. This includes advancements in diagnostic accuracy, telemedicine and drug research.

Surgery is another area in which narrow AI is making a difference. Operations save millions of lives every year, but they can also lead to severe complications for patients. Narrow AI can not only help medical staff better prepare for a surgery (for example through automation of administrative processes or analysis of medical history and risk factors), it can also provide real-time guidance during the surgery and insights after surgery. This can help with patient recovery, but is also a source of video-based learning for surgeons.

So, narrow AI could significantly improve patient outcomes and reduce complication rates, while also potentially increasing efficiency and cutting costs for hospitals. However, surveys also show skepticism and distrust of AI among patients and healthcare providers, who worry about safety and reliability.

To unlock the full potential of narrow AI, then, the public and private sectors must help foster trust about using AI among healthcare professionals and patients. There are three pillars that could support AI use in almost all sectors, but these pillars are even more important when it comes to dealing with sensitive healthcare information and decision-making.

The cornerstone of any AI system lies in the quality of the data it is trained with. Diverse datasets help prevent bias and maintain the precision of narrow AI, which builds trust in these systems. An unwavering commitment to data quality will ensure outcomes mirror real-world medical scenarios, while also upholding the integrity of the AI model.

Transparent communication by organizations is paramount in conveying this commitment. For example, at Caresyntax, ongoing efforts to incorporate a variety of medical scenarios in datasets has helped with the use of AI algorithms for real-time intraoperative decision support and to guide surgical teams. This has reduced surgical variability and promoted patient safety.

Establishing robust standards, certification bodies and frameworks that guarantee safety, transparency and accountability is another pivotal aspect of building trust in narrow AI among professionals and patients. There are a number of examples of this being developed across both the public and private sectors.

The EU’s proposed AI Act, the world’s first legal framework on AI, outlines a certification framework for high-risk AI and suggests voluntary measures for lower risk AI systems. A number of independent bodies have focused on translating AI standards to healthcare-specific contexts.

Strict enforcement of regulations and robust measures for patient data protection, including data anonymization, is essential for establishing trust in the development and deployment of AI. Existing regulations like HIPAA in the US and GDPR in the EU form a solid foundation, although regional disparities persist.

The private sector also plays a crucial role in driving privacy and protection initiatives. For instance, becoming a Patient Safety Organization (PSO) involves collecting, de-identifying (anonymizing) and aggregating patient safety event data. Working with a PSO ensures confidentiality protection and legal privilege, enabling vital data comparison with peers through the Network of Patient Safety Databases (NPSD), for example.

If healthcare systems can build sufficient trust in narrow AI, it could have a significant effect on patient outcomes. But the benefits go far beyond the patient. Narrow AI could bring many advantages for our planet too, such as reduced energy consumption and resource waste.

Globally, the healthcare sector was responsible for 4.6% of global greenhouse gas emissions in 2017. Hospitals are typically very energy-intensive buildings, while operating rooms (ORs) generate 50–70% of total hospital clinical waste. The carbon footprint of surgery in the US, UK and Canada alone is estimated at 9.7 million tons of CO2 per year.

However, our internal data shows using narrow AI in the OR via the Caresyntax platform can significantly reduce energy consumption and decrease waste through increased surgical efficiency. The contribution of narrow AI to a greener, more environmentally conscious healthcare industry could be substantial.

Narrow AI holds immense potential to transform surgical practices, offering precision, efficiency and improved patient outcomes. However, the successful adoption of narrow AI in healthcare depends on building trust among professionals and patients. Reliable data, strong standards and data protection form the cornerstones of this trust.

By establishing robust frameworks that prioritize patient well-being over pure profit, stakeholders can collectively contribute to the effective integration of narrow AI. This would not only enhance the quality of healthcare delivery, but it could also have a positive impact on our planet.