Artificial Intelligence and Machine Learning

A Healthy Dose of AI

The boundaries of algorithmic medicine continue to expand.

Diagnosing and treating medical problems is at the heart of effective healthcare. However, too often, symptoms are vague, errors occur, or people simply do not have access to adequate healthcare at the moment they need it most. As a result, many conditions—from minor annoyances to catastrophic illnesses—go undiagnosed or misdiagnosed.

The answer? There’s a growing focus on plugging artificial intelligence into devices, smartphone apps, and medical-grade diagnostics tools. These include AI-powered heart monitors, stethoscopes, body scanners, blood sugar monitors and even a cough analysis app that can help determine whether a person has tuberculosis, COVID-19, or asthma.

“AI is transforming medical devices by enhancing diagnostic precision, enabling predictive analytics, and facilitating personalized medicine,” says Manuja Sharma, a University of Washington post-doctoral researcher. “AI introduces capabilities that streamline healthcare delivery, optimize patient outcomes, and pave the way for cost-effective diagnostics and treatments.”

Of course, whether AI qualitatively improves outcomes remains to be seen. Says Leo Anthony Celi, director and senior research scientist at the Laboratory for Computational Physiology at the Massachusetts Institute of Technology (MIT), “Simply adding AI-powered devices to an already flawed system won’t help. We must also address the structural inequities that exist in healthcare.”

AI Doesn’t Skip a Beat

For better or for worse, artificial intelligence increasingly is reshaping healthcare. It already is widely used to parse patient data, analyze medical images, predict diseases, and aid in drug discovery and development. At the same time, health trackers like the Apple Watch have put heart monitors, sleep recorders, fall detectors, and other diagnostics on people’s wrists.

AI plays an expanding role in collecting, managing, and analyzing this data. For example, healthcare wellness technology company Withings has developed a $250 handheld contactless device called BeamO, which conducts a 1-minute home check-up using a body temperature sensor, digital stethoscope, medical-grade electrocardiogram (ECG), and blood oxygen monitor.

The device might not yet be the famed Tricorder from Star Trek, but it’s inching closer. “AI empowers patients by delivering important insights into their health,” says Livia Robic, Withings product manager for BeamO. But broader research is also a goal. “It also can help doctors and the broader healthcare system obtain data that can improve outcomes. It can identify important patterns and trends.”

Withings is hardly alone. A non-invasive handheld Cardiac Performance System (CPS) from Sensydia relies on biosensors and algorithms to gauge heart performance, including cardiac output, pulmonary artery pressure, and pulmonary capillary wedge pressure. Medical device company Medtronic has introduced a smartphone connected system called Guardian Connect, which monitors blood sugar and automates insulin delivery for people with Type I ore Type II diabetes.

At Glasgow University in Scotland, researchers have developed a laser stethoscope that captures precise readings through quantum technology. It uses high-speed cameras to snap pictures of a person’s throat at a rate of 2,000 frames per second. AI filters out movements and vibrations. The research group has launched a startup that hopes to get the devices into homes, medical offices, and other places.

“There are intriguing possibilities when data from different sensors is combined and run through AI. It’s possible to see things, and identify trends, that could otherwise go undetected,” Robic says.

Medicine Gets Personal

The boundaries of algorithmic medicine continue to expand. For instance, researchers at the University of Washington have developed an AI tool that records and diagnoses a cough using a smartphone. The algorithm can differentiate coughs of subjects with pulmonary tuberculosis from coughs due to other respiratory health issues.

The team developed the system using audio recordings of more than 33,000 coughs from 197 patients with various types of respiratory illnesses. “We transformed the 1-D audio signal to 2-D using wavelet transformation,” Sharma explains. “This let us visualize the change in energy over frequency and time.” Finally, using machine learning on GPU clusters, they compared images of the transformed signal along with known recordings of coughs to fine-tune and verify the algorithmic model.

“AI can be a game-changer,” Sharma says. “This tool could play an important role in identifying areas where there’s a high frequency of TB or other cough related illness. It could help prevent the spread of disease early.”

In fact, Withings, which also sells smart scales and other devices that tap AI, hopes to develop a broad framework that allows it to securely share anonymous data with doctors and public health officials. This could aid individuals and their medical providers, but it could also help epidemiologists and others understand and respond to broader events, such as influenza patterns or a future pandemic.

Researchers and engineers say their aim is not to replace doctors and other healthcare practitioners with AI apps and bots, but to promote early and effective diagnoses, simplify data collection for physicians and nurses, and help personalize medicine. As Sharma puts it, “These advancements increase the accuracy of radiological tests, improve remote health monitoring, and broaden the analysis of bio-signals through innovative wearable technology.”

Not everyone shares her enthusiasm. MIT’s Celi, for instance, believes that AI-powered medical devices can deliver specific benefits. He is deeply concerned about intrinsic biases that frequently become entangled in AI models, and he openly questions whether AI-powered medical devices will be available to all, or simply lead to greater inequities in healthcare.

“It’s critical to avoid structural inequities that result from unequal access to technology and care,” Celi says. “At present, many populations are underserved by technology and healthcare. Simply adding more technology doesn’t fix the problem.”

Beyond the Algorithm

Whether AI will usher in better healthcare or magnify existing problems and disparities remains to be seen. Yet it’s clear that in the years ahead, AI will appear in more devices and, at the very least, allow consumers and doctors to better track patient vitals and other health factors. AI could also boost early detection for certain types of diseases, including cancer, while driving down costs.

Concludes Celi, “AI holds promise, but only if we can rethink how we use these devices and systems, and how we make decisions and deliver healthcare.”

Samuel Greengard is an author and journalist based in West Linn, OR, USA.

Join the Discussion (0)

Become a Member or Sign In to Post a Comment

The Latest from CACM

Shape the Future of Computing

ACM encourages its members to take a direct hand in shaping the future of the association. There are more ways than ever to get involved.

Get Involved

Communications of the ACM (CACM) is now a fully Open Access publication.

By opening CACM to the world, we hope to increase engagement among the broader computer science community and encourage non-members to discover the rich resources ACM has to offer.

Learn More