Artificial intelligence is becoming increasingly apparent in modern society as advanced algorithms and deep learning technologies make incredible progress. The medical sciences are no exception, as researchers at Stanford University, California, have helped develop a framework to assist engineers looking to construct new generalist medical artificial intelligence (GMAI) models.
Image Credit: Iurii Motov/Shutterstock.com
Published in the journal Nature, the team, which includes researchers at Harvard University, Scripps Research Translational Institute, and the University of Toronto, claims that the exponential progress observed in artificial intelligence systems could radically alter the technological landscape and capabilities of what is possible in medicine.
We see a paradigm shift coming in the field of medical AI.
Professor Jure Leskovec, Professor of Computer Science at Stanford Engineering
Medical GMAI applications could help improve various areas of medicine, including diagnosis and screening, personalized medicine, surgical procedures, drug development and many more.
The researchers anticipate that GMAI-enabled applications will challenge existing strategies for regulating and validating AI in medicine and require an adjustment in certain practices associated with mass data collection.
GMAI: A New Era
In the very near future, artificial intelligence will grant doctors and medical professionals the ability to access a patient’s personal healthcare record and compare it with all available literature and data published online in an instant.
Previously, medical AI models could only address very small, narrow pieces of the health care puzzle. Now we are entering a new era, where it’s much more about larger pieces of the puzzle in this high stakes field.
Professor Jure Leskovec, Professor of Computer Science at Stanford Engineering
The team understands how GMAI will outperform current advanced concurrent AI models such as ChatGPT.
The researchers claim that GMAI can also annotate medical records and images and offer feedback and treatment pathways. The flow of information between individual patient records and the mass data available online tends to be inefficient, which can cause oversights or delays in certain areas.
A lot of inefficiencies and errors that happen in medicine today occur because of the hyper-specialization of human doctors and the slow and spotty flow of information… The potential impact of generalist medical AI models could be profound because they wouldn’t be just an expert in their own narrow area, but would have more abilities across specialties.
Michael Moor, an MD and Postdoctoral Scholar at Stanford Engineering
Performance, Safeguards, and Concerns
As it stands, there are around 500 AI models that the FDA has approved for use in clinical conditions. However, the capabilities and tasks these models can perform are extremely limited.
This is because existing models have been trained on specific datasets for specific tasks, such as scanning X-Ray images for a particular condition or ailment, which means signs of other conditions or illnesses can go undetected.
The exciting and the groundbreaking part is that generalist medical AI models will be able to ingest different types of medical information – for example, imaging studies, lab results, and genomics data – to then perform tasks that we instruct them to do on the fly.
Professor Jure Leskovec, Professor of Computer Science at Stanford Engineering
GMAI could offer unprecedented potential to perform a handful of tasks simultaneously and perhaps carry out thousands of tasks beyond initial development capabilities. Furthermore, GMAI could potentially converse with patients directly and take notes that could be cross-referenced against all available data and even suggest treatment plans or make the recommended referrals.
However, one of the biggest concerns is how accurate the GMAI models are when it comes to offering advice and making recommendations for treatment plans. This is based on current AI capabilities such as ChatGPT, which can offer eloquent responses to seemingly complex questions but has shown that there are inaccuracies or just erroneous claims in the information being communicated.
We think the biggest problem for generalist models in medicine is verification. How do we know that the model is correct – and not just making things up?
Professor Jure Leskovec, Professor of Computer Science at Stanford Engineering
Therefore, the appropriate safeguards must also be developed when it comes to overseeing patient health or calculating risk in potential life or death scenarios; there is no room for inaccuracies given what is at stake.
It is the responsibility of the owners and developers of such models and vendors, especially if they’re deploying them in hospitals, to really make sure that those biases are identified and addressed early on.
Michael Moor, an MD and Postdoctoral Scholar at Stanford Engineering
Ultimately, GMAI, if implemented correctly, could shift the paradigm and provide medical professionals with unprecedented possibilities for healthcare.
Moreover, clinicians could feel the benefit of extra support in essential tasks, which could deliver high-quality, accessible healthcare while reducing administrative loads and freeing up doctors, allowing them to spend more one-on-one time with their patients.
References and Further Reading
Binns, C. (2023) Advances in generalizable medical ai, Stanford News. Available at: https://news.stanford.edu/2023/04/12/advances-generalizable-medical-ai/
Moor, M. et al. (2023) “Foundation models for generalist medical artificial intelligence,” Nature, 616(7956), pp. 259–265. Available at: https://doi.org/10.1038/s41586-023-05881-4.
Disclaimer: The views expressed here are those of the author expressed in their private capacity and do not necessarily represent the views of AZoM.com Limited T/A AZoNetwork the owner and operator of this website. This disclaimer forms part of the Terms and conditions of use of this website.