AI Moves Beyond Diagnostic Support and Into Treatment Decisions

Artificial intelligence (AI) is making itself at home in many areas of the clinical workflow. Most recently, it’s been tapped for treatment decisions – an area where it’s proving quite effective based on the outcomes of recent research. One example comes by way of a study  published in Nature Medicine, where researchers from Imperial College London used a “reinforcement learning agent,” dubbed the AI Clinician, to identify the best treatment strategy for a sepsis patient by analyzing physician decisions across approximately 100,000 patients in intensive care units.

By looking at US patient records from 130 intensive care units over a 15-year period, researchers were able to explore whether the AI Clinician’s recommendations would improve patient outcomes when compared to standard care. This information was then used by the system to predict the best treatment strategy for each patient. In 98 percent of the cases, the AI Clinician’s decision either matched or was better than the physician’s decision.

In an article by Imperial College on the study, Dr. Aldo Faisal, senior author from the Department of Bioengineering and the Department of Computing at Imperial, noted that AI is capable of far more than the diagnostics for which it is currently used.

“…Doctors do more than just diagnose, they treat people. Our AI Clinician system focuses on capturing this cognitive capacity of doctors: Imagine having a doctor watching over you every second of every day, administering a course of treatment, observing how you respond to the treatment, and then adjusting the treatment as your condition evolves,” he said. “The AI Clinician technology we developed can have many applications in medicine, whenever we need to choose, observe and adjust treatment. Whenever there are large amounts of patient data the AI Clinician can assess and learn from, the system can be used. We have applied this technology previously to treatment in diabetes and in anaesthesia during surgery and can use this to optimize the delivery of expensive treatments, e.g. in cancer therapy.”

Elsewhere, AI is being used to predict an individual patient’s outcome under a new method for determining the effectiveness of medical treatments, a process normally reserved for randomized trials. In a study published in Healthcare Informatics Research, Finnish researchers used modeling to compare different treatment alternatives and to identify patients who will benefit from treatment. Built around AI, the model is based on causal Bayesian networks and opens significant avenues for medical research.

“We can now predict the treatment outcome in individual patients and evaluate existing and new treatment methods,” notes Professor Emeritus Olli-Pekka Ryynänen from the University of Eastern Finland. “With this method, it is also possible to replace some randomised trials with modelling.”

Researchers used the method to evaluate treatment effectiveness in obstructive sleep apnea. However, the model can also be applied to other treatments, such as heart conditions and hypertension.

For all its early successes, however, there is still room for improvement. In a study published in Lancet Diabetes & Endocrinology, researchers looking at the application of precision medicine – medical care that is designed to optimize diagnostic efficiency or therapeutic benefit by leveraging genetic, biomarker, phenotypic or psychosocial characteristics – to the treatment of obesity found that a more multifaceted approach was required than could be provided by AI.

“…Interventions to tackle obesity should take a more multifactorial approach, addressing the array of factors that have resulted in the current epidemic of obesity, and pursuing a more socially conscious medicine,” they wrote. “To be applicable to obesity, the rapidly expanding field of precision medicine needs to acknowledge these complexities and realise that genetic analysis is not the sole path to personalised treatment.”

Sign up for News and Updates.