Skip to content


Health |
AI may be helping your doctor. Here’s how the technology is being used by health networks in the Lehigh Valley

Dr. Michael Hooper, of Sentara Norfolk General Hospital in Virginia, discusses the use of artificial intelligence in health care in this August 2019 file photo. Hospitals around the country, and in the Lehigh Valley, are using AI in some cases to assist doctors and better patient outcomes. (Kaitlin McKeown / The Virginian-Pilot)
Dr. Michael Hooper, of Sentara Norfolk General Hospital in Virginia, discusses the use of artificial intelligence in health care in this August 2019 file photo. Hospitals around the country, and in the Lehigh Valley, are using AI in some cases to assist doctors and better patient outcomes. (Kaitlin McKeown / The Virginian-Pilot)

If you had an imaging study done at a Lehigh Valley Health Network-owned hospital in the last few months, there is a chance that an artificial intelligence program was helping the radiologist check for serious conditions.

Last year, LVHN began implementing a series of AI tools for the radiology departments at all 13 network hospitals. Dr. Devang M. Gor, chair of Radiology & Diagnostic Medical Imaging for LVHN, said these tools are changing the way radiologists do their jobs.

“I’m a super user of this technology. I use it every day and I help others in my department adopt the technology and keep using it,” Gor said.

AI has been on the minds of many people thanks to viral news reports and social media discourse surrounding AI-generated voice technology or AI-generated art, as well as existential dread that AI will replace the need for humans in many work fields. But AI-powered medical technology has been in use within the Lehigh Valley for years at both LVHN and St. Luke’s University Health Network.

LVHN began using AI technology in 2018, spokesperson Jamie Stover said. St. Luke’s also started in 2018, and these efforts have ramped up over the last several years, Charles Sonday, associate chief medical information officer for St. Luke’s, said in an emailed statement.

But rather than replacing doctors or health care workers Dr. Maulik Purohit, former chief health information officer for LVHN, said AI-powered tools are helping health care professionals be more effective at their jobs and create better outcomes for patients.

One of the AI tools LVHN implemented in its radiology departments, Aidoc, immediately reads all imaging studies to help radiologists diagnose patients and help identify which ones need immediate care. Purohit said Aidoc has been trained using a huge dataset of diagnostic medical images so it can identify patterns in the images consistent with serious medical conditions.

The Aidoc package LVHN is using is capable of identifying blood clots in the lungs, collapsed lungs and fractures in the neck area, all of which can be imminently life-threatening. The program was about 93% successful at spotting cases of pulmonary embolism, the medical term for a blood clot in the lungs, and accurate about 95% of the time in identifying that no blood clot was present, . Gor said the AI has even caught cases of pulmonary embolism when imaging studies were being done for entirely different reasons.

If any of these three serious conditions are identified, Aidoc’s technology notifies the radiology team and prioritizes putting the imaging study toward the top of the queue of studies radiologists need to examine. If a pulmonary embolism is detected, Aidoc also will notify the hospital’s pulmonary embolism response team so the patient may receive care immediately.

“Until you see the study you have no way of knowing whether it is urgent or not,” Purohit said. “This allows us to automate that process of identifying what’s urgent and less urgent so that patients that need the most urgent care receive it quickly.”

Gor said even though Aidoc was only implemented networkwide in September, the technology already has made a difference. It saves radiologists time, makes communication between different subspecialty care teams easier and leads to better care outcomes for patients, he said.

However, Gor said, the AI does not diagnose patients — a radiologist on staff needs to do that. It only provides suggestions and alerts so it is easier for radiologists to make the final call on an imaging study.

St. Luke’s has also used AI-powered technology in its radiology departments. Last May, . The technology can produce faster scans and sharper images, reduce radiation patients are exposed to, detect lesions or tissue abnormalities and map vascular structures. It can also capture fine detail in the head and neck, which is critical when diagnosing stroke, according to the network.

The network , an AI embedded into X-ray machines that is capable of helping clinicians identify collapsed lungs. Critical Care Suite received FDA clearance in 2019. St. Luke’s is no longer using that technology though, said Sam Kennedy, a St. Luke’s spokesperson, and the network is currently evaluating new AI technology from GE Healthcare for similar purposes.

LVHN has adopted or will adopt other AI software to make radiologists’ jobs easier.

One tool, Rad AI Omni, helps radiologists when there are findings that don’t require immediate care, such as nodules in the lung or thyroid gland, adrenal lesions, kidney cysts or enlarged lymph nodes. The software automatically generates as summary of the findings from a radiologist’s dictation, and then it compiles and inserts follow-up guidelines from national medical organizations into reports.

This standardizes the recommendations given to patients and allows radiologists to focus on other responsibilities. Gor said one of the most valuable aspects of this technology is it allows him to take a second look at the studies and make sure nothing is missing, something he often didn’t have time for before he started using Rad AI Omni.

There is another AI tool that LVHN hopes to fully implement in mid-March, Rad AI Continuity, which will help with the management of follow-up care for incidental findings. When incidental findings such as lung nodules or adrenal lesions are identified, the software will automatically send follow-up recommendations to the patient and referring clinician, whether it is more tests, scans or some other follow-up care. Rad AI Continuity will keep checking in with both patient and clinician until follow-up appointments or tests are scheduled.

“Those are hard to manage from a human perspective,” Purohit said. “That system is automatically aware if that person actually followed through on that [appointment] so that we don’t lose track of the patient.”

Beyond better outcomes for patients, Gor said these AI tools can also create better outcomes for clinicians.

There is a global shortage of radiologists due to factors such as burnout among existing radiologists and not enough new people entering the field. Gor said tools like this are helping to augment radiologists so they can keep up with the demands placed on them.

“We do a very large volume of acute studies. It is humanly not possible for anyone to immediately start reading those studies as soon as the studies are done,” Gor said.

He added that beyond increasing the efficiency of radiologists, the tools decrease stress and burnout.

“We are facing increased volumes so people are running short,” Gor said. “They’re not able to read studies fast enough and they don’t have staffing. Where AI fits in is it makes you more efficient. It gives you a second set of eyes where you can rapidly process and triage patients.”

That isn’t to say there hasn’t been a learning curve for radiologists in the network when using these tools. As with any technology, the AI isn’t perfect and does occasionally add disruptions to the workflows of radiologists. Gor said one of these disruptions is receiving notifications when the AI returns a false positive result. But he added these false positives are a minority.

“Humans are creatures of habit. Change is not necessarily accepted everywhere, no matter what it is,” Gor said. “Many of us now feel after using it for three months, how did we do this before? It helps you really be more efficient.”

What other AI technology is being used?

Radiology isn’t the only area of care where LVHN and St. Luke’s are using AI to augment the capabilities of health care providers.

In 2020, LVHN adopted Viz.ai Neuro, AI software that uses an algorithm to determine the probability a patient experienced a stroke. If it suspects a stroke occurred, the stroke team is alerted and the patient’s CT images are sent directly to a stroke specialist.

The network also runs a predictive algorithm in its intensive care units that helps predict sepsis, a condition that can occur during an infection where the body’s immune system begins targeting the body itself causing tissue and organ damage. Sepsis can progress further and its side effects may ultimately lead to death.

“You want to minimize the spread of the infection and you also want to minimize the negative effects of fighting the infection such as tissue damage. Sepsis also has a high mortality rate, particularly as it progresses to severe sepsis and beyond. We want to catch it early before it becomes full-blown severe sepsis,” Purohit said.

The algorithm LVHN uses helps avoid that outcome by analyzing key data points about patients and their conditions and alerting clinicians when it suspects there is a potential case of sepsis so that they can intervene if the algorithm’s prediction is correct.

LVHN has introduced other AI to help clinicians with their workflow. Nuance is an AI software that helps generate notes for clinicians to review and sign off on, said Stover, the health network spokesperson.

St. Luke’s also has adopted a variety of AI-powered tools and software. Sonday said the network uses algorithms developed by Epic Systems to help address risks such as sepsis, unexpected ICU transfer and readmissions by analyzing clinical data.

In 2022, St. Luke’s announced it installed the Varian Ethos therapy system at St. Luke’s Cancer Center in Upper Macungie Township. This AI-driven system allows physicians to adjust cancer treatments to precisely match the patient’s specific anatomy and the position of the tumor.

About 11 months ago, St. Luke’s began using the GI Genius endoscopy module, a polyp detection system that uses an AI algorithm to help clinicians catch colon polyps before they develop into colon cancer. During a colonoscopy, the AI searches for polyps, which are unusual tissue growths that can become cancerous, other lesions as well as other points of interest and marks them to help the clinician determine if further assessment or treatment is needed.

The adoption of AI in medicine is unlikely to slow down. The capabilities of AI are growing at an alarming rate, meaning new technologies and uses for AI are not too far over the horizon.

However, Gor said one of the limitations to adopting new technology is cost. AI is not cheap to create nor securely host by software companies, and that means the costs of AI software subscriptions for health care providers are quite pricey.

He added that actual implementation is rarely seamless. It requires a lot of information technology time and resources to get AI software working within existing networks and systems. And as technology improves, that means there are more updates and more software that is adopted.

“It keeps on getting better and better. Tomorrow there’ll be more and more algorithms to detect other situations,” Gor said. “Whatever you have keeps evolving so there is a good pace at which you need to keep updating.”

More in Health