Artificial intelligence could personalize disease care

Madabhushi’s research found that combining AI with medical images to find non-invasive and economic treatments for major cancers and diseases could detect such issues earlier and help clinicians better manage treatment decisions.

More clinical testing is needed before doctors can use AI to change their treatment strategies. But early results show its potential for reducing overtreatment.

“AI is on everyone’s lips right now,” he said. Researchers used AI computer programming to gather detailed information about disease treatment in the past, but they lacked medical data doctors could understand to prove AI’s veracity. “The skepticism is real. It’s not misplaced, but a large part of it is the lack of real understanding about AI.” If doctors don’t understand how AI works, they can’t trust its predictions, he said.

Madabhushi led one study, published July 9 in The Lancet Digital Health, that showed how AI could be used to better manage treatment for patients with early-stage breast cancer known as ductal carcinoma in situ (DCIS), also considered stage 0 breast cancer. It accounts for about 20% of all new cancer diagnoses each year affecting more than 55,000 women.

The data for the study was pulled from a clinical trial conducted in the UK, Australia and New Zealand from 1990 to 1998 to determine the effect of tamoxifen, a breast cancer drug, and radiotherapy on 755 patients with DCIS.

About 30 years later, in 2020, AI was used to analyze the data and identified with 95% accuracy which patients had a higher risk of disease progression and could benefit most from radiation therapy. Medical records of the patients from the ‘90s confirmed the AI predictions of how their cancer would progress.

Though the analysis was based on 30-year-old data, Madabhushi explained that the management of women with DCIS hasn’t changed much in that time despite the arrival of newer cancer treatments. Clinicians still have to decide between surgery or surgery plus another treatment such as a drug or other therapy, said Madabhushi, who is also a cancer immunology researcher with Emory’s Winship Cancer Institute.

Using the old data to confirm the new technology’s predictions is “like the Lazarus Effect,” Madabhushi said, referring to the medical phenomenon where someone declared dead comes back to life. “We can learn and discern new information from [data gathered] so long ago in a different country that could benefit patients today.”

More discoveries to come

In the next year, Madabhushi expects to seek approval from the U.S. Food and Drug Administration to use AI as a medical device and conduct clinical testing that will monitor AI’s impact on patient health. He predicts that within two years doctors will be able to use AI to take some of the guesswork out of their decisions by helping them pinpoint the most effective treatment for their patients.

Madabhushi and a team of researchers also used AI to detect eye inflammation that is a serious side effect caused by drugs to treat a common eye disease.

AMD is the leading cause of vision loss for older adults affecting 11 million Americans. A specific type of AMD, neovascular age-related macular degeneration (nAMD), is associated with abnormal blood vessel growth under the retina. The treatment involves injecting medication into the eye, but inflammation can be a serious side effect.

Using a machine learning model developed by Emory AI.Health, researchers identified patterns seen in eye scans that signaled inflammation even before it was visible to doctors. The study was published in June in the Cell Press journal Heliyon.

Another Madabhushi-led study used AI to analyze the lung damage caused by COVID. Using CT scans from more than 3,400 patients and three continents. The study showed that patients with severe COVID experienced significant deformities to the surfaces of the lungs, according to the research published in May in the Journal of Computers in Medicine and Biology.

“As we are thinking about long COVID, we still try to understand its impact long-term,” Madabhushi said. “We can quantify the impact the disease has to the extent of lung injury, the quantitative impacts on lung function.”

‘Revolutionizing’ medical care

Madabhushi is considered one of the global pioneers in combining AI with high-resolution medical images for diagnosing diseases and predicting the results of patient treatment, according to Dr. Jacob “Jake” Scott, a radiation oncologist who conducts AI research through the Cleveland Clinic and Case Western Reserve University. They supervised students at the latter when Madabhushi was a professor of biomedical engineering and director of the Center for Computational Imaging and Personalized Diagnostics before joining Emory.

Several Cleveland researchers co-authored recent AI studies with Madabhushi.

“Anant is leading the charge in redefining the medicine patients need,” said Scott, co-director of Case Western’s Center for AI Enabling Discovery in Disease Biology. “It’s a newer field that allows a deeper level of intuition.”

Scott added that Madabhushi and his team are “pushing the limits … taking a new and powerful tool and revolutionizing how we understand disease.”

Madabhushi and his team are now trying to apply what they learned from their AI research to other cancers, including those impacting the prostate and lungs. “Cancer is not necessarily a death sentence,” he said.

In September, Madabhushi plans to present new research on hormonal therapy for DCIS and breast cancer patients at the European Society of Medical Oncology. Building off research on the use of radiation, researchers were able to use AI to identify which stage 0 breast cancer patients would benefit from hormonal therapy and which should avoid it, he said.

“It’s fulfilling to see my work come out,” he said of the recently published research. “The breast cancer work has been near and dear to me for over 20 years. … This particular journey I have been on to find a cure for the disease that killed my aunt.”