To help you understand the trends surrounding new technologies and what we expect to happen in the future, our highly experienced Kiplinger Letter team will keep you abreast of the latest developments and forecasts. (Get a free issue of The Kiplinger Letter or subscribe). You’ll get all the latest news first by subscribing, but we will publish many (but not all) of the forecasts a few days afterward online. Here’s the latest…
As generative artificial intelligence gets better at interpreting images, the tech industry is setting its sights on health care. Cue the AI radiologist.
The futuristic vision includes AI providing an accurate analysis of multiple medical scans, combining it with an understanding of patient history, and delivering a personalized diagnosis and course of treatment. When paired with a trained clinician, AI tools have the potential to improve the quality of care, save time and expand access to specialist expertise, among other benefits, according to a new paper, “Multimodal generative AI for medical image interpretation,” in the medical journal Nature.
The authors say early research suggests that AI “could one day match human expert performance in generating reports across disciplines, such as radiology, pathology and dermatology.”
Subscribe to Kiplinger’s Personal Finance
Be a smarter, better informed investor.
Save up to 74%
Sign up for Kiplinger’s Free E-Newsletters
Profit and prosper with the best of expert advice on investing, taxes, retirement, personal finance and more – straight to your e-mail.
Profit and prosper with the best of expert advice – straight to your e-mail.
Interpreting medical images and writing reports is a time-consuming challenge for human specialists. Risks include delays in getting results and human errors. Many current AI medical tools have narrow uses to find specific issues in a certain type of scan. A future AI model could have an expansive knowledge of multiple types of scans, all sorts of medical conditions and a range of treatments to recommend.
Opportunity for profit
Tech giants and start-ups alike see huge moneymaking potential. Microsoft, Google and OpenAI all have AI models or research in medical imaging. Start-up Harrison.ai recently raised $112 million in funding to speed up diagnoses for radiologists, calling its widely used tool “a second set of eyes for clinicians.” Microsoft is working with major hospitals on AI tools to interpret thousands of conditions, trying to tap into the tens of billions of dollars health systems spend annually on imaging and uncover cost savings.
“Generative AI has transformative potential to overcome traditional barriers in AI product development and to accelerate the impact of these technologies on clinical care,” said Keith J. Dreyer, D.O., Ph.D., chief data science officer and chief imaging officer at Mass General Brigham, in a news release last summer about the Microsoft collaboration.
Risk of inaccuracy and tampering
But there are “formidable obstacles” to finding a truly helpful AI assistant in radiology, notes the paper. AI models have been plagued by biases, inaccuracies and so-called hallucinations, the industry term for made-up answers, including false or misleading text that sounds authoritative. Those types of flaws are nonstarters in a high-stakes medical setting unless there are strict guardrails.
There are concerns that AI could overlook rare diseases for different populations, plus cyber fears of the tools being tampered with to deliver certain results, “leading to overprescription, insurance fraud and falsifying clinical trials,” says the paper. However, AI models are getting better and multiple models can be used together for better results.
Human governance
One thing the researchers make clear: “Human evaluation is critical.” AI tools for medical imaging need doctors to fine-tune them, and better benchmarks so they can be tested and improved. There need to be agreed-upon metrics to decide when and how to rely on AI.
In the near term, AI could become a reliable assistant to clinicians, taking initial readings, drafting preliminary reports and even answering questions via a chatbot. Increased use of generative AI will seriously pick up in coming years, aiming to save time for radiologists and help train the next generation of students.
The future for patients includes interacting with an AI doctor about test results and getting answers, and even treatment suggestions, stripped of medical jargon. Also likely: getting a second opinion from one or multiple AI models, rather than a real doctor. The hope is patients will have higher compliance with treatment and better outcomes through being able to ask a lot more questions and not feeling rushed during a short visit with a human doctor.
In the meantime, be wary of using free public AI tools for medical advice with your own medical scans. Uploading personal images and medical info pose big privacy risks, on top of concerns about accuracy.
So are human radiologists on the road to extinction? It’s unlikely, but it’s not hard to find such predictions. It’s likely human judgment remains solidly in the loop, even as AI greatly improves. Plus, future regulations may prevent AI from completely taking charge.