Health Rounds: AI speeds MRI exams

(Health Rounds is published on Tuesdays and Thursdays. Think your friend or colleague should know about us? Forward this newsletter to them. They can also subscribe here.)

Jan 19 - Hello Health Rounds Readers! Today we report on two studies of potential applications for artificial intelligence in medicine: MRI scanning, and smart watch identification of heart rhythm problems. We also share early data to suggest that plant-based therapy may be helpful for some patients with ulcerative colitis, a debilitating gastrointestinal disease.

Artificial intelligence gets MRI done faster

Artificial intelligence (AI) allows magnetic resonance imaging (MRI) scans to be done faster with no less accuracy, according to researchers who hope to reduce examination costs and allow more patients to get the tests.

MRI exams can take a long time to perform because the machines take hundreds of images of body parts in “slices” and then compile the slices into three-dimensional representations for doctors to analyze.

At NYU Langone Health in New York City, doctors and scientists collaborated with Meta AI Research to make the machines scan body parts faster, but at the cost of collecting only a fourth of the usual amount of data. They then used AI to “fill in” the missing data, similar to the way the brain fills in missing information using local context and previous experiences.

For a study reported on Tuesday in Radiology, 170 patients underwent MRI of the knee in the usual way, which took an average of roughly 10 minutes, and in an accelerated AI protocol that took slightly over five minutes. More complex MRI exams can take 30 minutes or longer.

The radiologists judged the AI-reconstructed images to be just as good as conventional images for detecting problems in patients’ knees and found the overall image quality of the faster scans to be significantly better than the conventional images.

Important next steps, the researchers said, include confirming that radiologists’ interpretation of the faster scans correlates with what surgeons later find in the body, and testing the software on different manufacturers’ MRI machines.

Smart watch ECGs still no match for doctors

Smart devices hold promise for identifying a common heart rhythm abnormality but their algorithms still need improvement, a new study found.

Researchers assessed the accuracy of five wearable devices at detecting atrial fibrillation (AF) - an irregular beating of the heart that can lead to blood clots and strokes - in 201 patients, including 62 with AF.

If inconclusive tracings were excluded from the analysis, accuracy at identifying people with and without AF were comparable among the devices, at 85% and 75%, respectively, for the Apple Watch 6, 85% and 75% for the Samsung Galaxy Watch 3, 58% and 75% for the Withings Scanwatch, 66% and 79% for the Fitbit Sense, and 79% and 69% for the AliveCor KardiaMobile.

There were significant differences in proportions of inconclusive readings from the five devices. With the Apple Watch 6, 18% of readings were uninterpretable by the device’s algorithm, the researchers reported on Wednesday in JACC Clinical Electrophysiology. For the Samsung Galaxy Watch 3 that figure was 17%, for Withings Scan Watch 24%, 21% for Fitbit Sense, and for AliveCor KardiaMobile, 26% of the readings were inconclusive.

Cardiologists, however, were able to interpret 99% of the readings that were labeled as inconclusive by the device algorithms, the researchers found.

“Signal quality for manual review is good, but the algorithms’ ability to automatically classify the rhythm is in need of further improvement to be of medical value in daily clinical practice,” they concluded. At this point, a smart device result can only be considered a “pretest,” they said, adding that “manual verification by a trained professional is mandatory.”

Plant-based therapy may ease ulcerative colitis

Curcumin, a component of turmeric, a common spice often found in curry, may have some benefit for patients with ulcerative colitis, a small study suggests.

Forty-two patients with active ulcerative colitis were enrolled in the randomized, controlled trial, with two-thirds of them receiving treatment with a combination of curcumin and the herbal medicine QingDai for eight weeks and one-third receiving a placebo. Patients whose disease improved continued with curcumin alone or placebo for another eight weeks.

Symptoms had improved by eight weeks in 85.7% of the treatment group and 30.7% of the placebo group. In 75% of the curcumin and QingDai group, doctors also saw evidence of healing inside the intestines versus 20% for placebo, according to data presented this week at the Crohn’s and Colitis Congress in Denver.

Among patients receiving just curcumin after benefiting by week 8, 93% still had symptomatic improvement, and 80% were in remission, at week 16, researchers found.

The treatment was associated with changes in gene and protein activity in patients’ intestinal lining that might explain its benefits, which suggests that induction of those changes might be a way to treat ulcerative colitis, the researchers said.

Ulcerative colitis affects roughly a million people in the United States alone, with high prevalence in Europe as well. U.S. Food and Drug Administration-approved drugs for treating the disease are costly and include Bristol Myers Squibb’s Zeposia, Johnson & Johnson’s Remicade and Stelara, Abbvie’s Humira, and Takeda Pharmaceuticals’ Entyvio. (Reporting by Nancy Lapid; Editing by Bill Berkrot)