The AI Evolution in Hospital Health Care
Using AI to increase efficiency and decrease time to diagnose serious conditions.
Written by John Ferrari
Every day, Torrance Memorial physicians make diagnoses, recommend treatments and perform procedures. As the volume of medical data available to physicians increases, doctors are adopting a new set of tools to aid in managing and analyzing information. These automated tools—artificial intelligence (AI) systems—are helping physicians provide care more quickly and efficiently, leading to improved outcomes—that is, healthier patients.
Data drives much of modern medicine. Sorting and analyzing data are the back-end functions necessary to make diagnoses and treat patients, and they’re perfectly suited for AI. “We’re a front-runner in using AI in California and the nation,” says Torrance Memorial interventional radiologist George So, MD. AI is used in almost every department, from emergency medicine to family practice, but nowhere more than in radiology.
Providing and interpreting medical imagery—from X-rays and ultrasounds to MRIs—radiology supports departments throughout the hospital. Its practices are a prime example of AI integration.
“For radiologists,” explains Dr. So, “everything consists of data. Pixels make up the images we use, and those can be processed by AI.”
By scanning databases of previously analyzed medical imagery, AI radiology programs learn to improve images and recognize potential trouble signs they reveal, making images easier for radiologists to read and interpret.
“For example, with AI assisting an MRI scan, we can speed up the scan and obtain high-resolution images,” Dr. So says. “AI can increase the resolution of the scan, making it sharper, or maximize the speed of the scan, or a combination of both.”
With CT scans, essentially a series of X-rays, AI’s ability to improve images means patients’ exposure to radiation can be reduced. Specialized AI programs also learn what tumors, blood clots, fractures and other features look like and can highlight suspicious features for radiologists.
“We’re trained to look for these pathologies,” explains Torrance Memorial interventional radiologist Jonathan Park, MD. “Having AI accurately detect pathologies adds a level of confidence. AI can take on some of the busywork involved in processing the data, and it alerts radiologists to positive findings so we can prioritize that.”
“AI can make lesions stand out more, and that makes the radiologist more efficient in terms of interpreting a scan that can contain thousands of images,” adds Dr. So. “We use AI as a backseat driver. It screens images to prioritize them. That’s very important, especially in emergency cases. We use AI for emergency screening of patients suffering from a stroke or pulmonary embolism. If the AI detects a major occlusion of the brain, the stroke team will be alerted via cell phone alerts. Even before the patient gets off the CT scanner table, the team is alerted to start preparing. Having that rapid AI combined with a stroke team and the workflow to get them in place and ready can save hours. AI is only a small part of that, but it’s an important part.”
The use of AI in emergency medicine doesn’t end with radiology. “In the ED, an AI system starts monitoring for sepsis symptoms as soon as a patient arrives and has vitals taken,” says Torrance Memorial’s ED co-medical director and executive director of medical informatics, Gina Sulmeyer, MD. Across the hospital’s departments, AI works in the background, aiding physicians in piecing together medical information.
“Anytime I order or prescribe a medication, an AI system scans the patient’s records for allergies, any other medications that might interact with the new medication and duplicate prescriptions,” Dr. Sulmeyer says.
Then there’s the ubiquitous task of recording patient information. Torrance Memorial uses an AI program to transcribe physician notes into the patient records system. That’s a game changer, Dr. Sulmeyer says. It allows physicians to both input and review patient data in real time. AI can also highlight potential medical problems, for example by linking changes in a patient’s health with a family history of stomach cancer.
Dr. Sulmeyer notes that most cancers are asymptomatic early in their development. In the past, that led to late-stage diagnoses, but AI is helping radiologists and oncologists spot potential tumors that would’ve gone undetected previously. “They’re called incidental findings—somebody comes in for one thing, and the system notices something else as well,” she says.
For example, a patient undergoes an MRI or CT scan for a heart issue. AI detects a faint shadow in the lung and flags it as a feature that needs to be investigated. Or an AI review of lab work reveals bone abnormalities indicative of a potential autoimmune disease.
Still, adopting AI technology doesn’t change what physicians do, Dr. Park notes. “It’s like being a pilot in a cockpit. Advanced aids have automated a lot of the busywork, but the pilot is still flying. In a hospital setting, AI is not making decisions. It’s just one of the tools we have at our disposal. It’s been an amazing addition to our practice."
