AI: EMPOWERING INNOVATION
Geisinger uses AI to improve cancer outcomes, reduce costs and boost patient care
Cancer rates are rising, including cancers among young people and rare cancers that can be difficult to diagnose and treat. At the same time, the U.S. is facing a shortage of radiologists and other key health-care professionals.
So clinicians must find ways to detect and treat the disease that are more effective, faster and less labor-intensive than conventional methods.
The good news: providers have a powerful new tool at their disposal – artificial intelligence. Hospitals across the country are harnessing AI to assist radiologists, physicians and other medical professionals in their war on cancer.
At Geisinger Medical Center, in Danville, Pennsylvania, world-class oncologists and radiologists use AI to analyze and interpret medical images, including from CT scans and MRIs, to create and carry out cancer treatment plans.
Geisinger is accessing the technology through a partnership with Siemens Healthineers, whose AI-Rad Companion software – running on Intel chips optimized for AI – is proving to be a game-changer at the Danville hospital.
“Tools like artificial intelligence can increase confidence in terms of diagnoses for different abnormalities that the patient has and can also speed up the treatment that patients require,” says Peter Shen, head of digital and automation, North America, for Siemens Healthineers.
AI-Rad Companion has algorithms for analyzing images from the chest, brain, prostate and organs.
Efficiency gains from AI-assisted care can reduce the cost of treatment at a time when U.S. health-care costs are at record highs in the trillions.
“You want to give people quality care, but you want to do it in a way that’s cost efficient,” says Heath Mackley, MD, Geisinger’s chair of radiation oncology. “I really look at AI as a way to treat more people with the same amount of staff. You’re getting the same outcome, but you’re doing it with less cost.”
Geisinger’s technology partnership with Siemens Healthineers also helps to improve health-care access, as residents of the rural communities where Geisinger typically operates can get the same advanced care as they would in urban centers like New York City or Philadelphia.
Contours of care
Among other things, Geisinger radiologists use AI-Rad Companion to assist with a process known as contouring – in which the boundaries of a tumor and organs at risk are outlined on patients’ medical images so practitioners can direct radiation beams to just the right spot.
Turnaround times for AI-assisted organ contouring are reduced in some cases from hours to minutes, compared to manual tracing. Given the number of cases that Geisinger treats, this can add up to hundreds of hours saved per month.
Speeding up treatment also helps to alleviate patient anxiety. “Once they’ve heard they have cancer, they want to start yesterday,” says Dr. Mackley. With some cancers, there’s no time to lose. “There are cancers that literally grow by the day. If someone comes in with leukemia, they need to be treated within a day of diagnosis.”
Using AI-Rad Companion for contouring in “extremely difficult” areas like the head and neck region is particularly useful, says Rick Holly, PhD, director of physics residency at Geisinger. The AI-assisted contouring is subject to human review.
To date, about 95 percent of the AI-generated contours are deemed clinically acceptable, Dr. Holly says.
AI-Rad Companion also has diagnostic modules, which use pattern recognition to spot signs of disease in medical images. Geisinger officials say the technology in some cases outperforms humans. “There’s the possibility of exceeding a human for diagnosis,” says Dr. Mackley. This is especially likely in cases where a cancer has metastasized to an unexpected location.
Faster AI with Intel inside
Although it can run locally on workstations, Geisinger is accessing AI-Rad Companion through the cloud to ensure seamless, campus-wide access for medical staff.
The software runs on Intel hardware that is purpose-built for AI. This includes the latest Intel® Xeon® processors, which feature Intel Advanced Matrix Extension (AMX) hardware accelerators for AI and which support Brain Float 16 data for deep learning.
Software optimizations are made through Intel’s distribution of OpenVINO™, an open-source toolkit that can be used to accelerate AI inferencing without sacrificing accuracy. Intel engineers worked directly with Siemens Healthineers to optimize AI-Rad Companion for Intel-based systems.
Geisinger’s Dr. Mackley expects the software to be used in even more advanced scenarios as it evolves. “I’m very optimistic it will be able to help us with every case at some point.”
Intel can help your organization use AI to get to the next level.