A round heart, or left ventricle sphericity, predicted cardiomyopathy and atrial fibrillation (AFib) in a deep learning analysis of MRI images from close to 39,000 participants in the UK Biobank, a new study shows.
An increase of 1 standard deviation in the sphericity index (short axis length/long axis length) was associated with a 47% increased incidence of cardiomyopathy and a 20% increased incidence of AFib, independent of clinical factors and traditional MRI measures.
Furthermore, a genetic analysis suggested a shared architecture between sphericity and nonischemic cardiomyopathy, pointing to NICM as a possible causal factor for left ventricle sphericity among individuals with normal LV size and function.
“Physicians have known the heart gets rounder after heart attacks and as we get older,” David Ouyang, MD, a cardiologist in the Smidt Heart Institute at Cedars-Sinai Medical Center, Los Angeles, and a researcher in the division of artificial intelligence in medicine, said in an interview. “We wanted to see if this sphericity is prognostic of future disease even in healthy individuals.”
Although it is too early to recommend heart shape assessment in healthy asymptomatic people, he said, “physicians should be extra careful and think about treatments when they notice a patient’s heart is particularly round.”
The study was published online March 29 in the journal Med.
Sphericity index key
The investigators hypothesized that there is variation in LV sphericity within the spectrum of normal LV chamber size and systolic function, and that such variation might be a marker of cardiac risk with genetic influences.
To test this hypothesis, they used automated deep-learning segmentation of cardiac MRI data to estimate and analyze the sphericity index in a cohort of 38,897 individuals participating in the UK Biobank.
After adjustment for age at MRI and sex, an increased sphericity index was associated with an increased risk for cardiomyopathy (hazard ratio, 1.57), AFib (HR, 1.35), and heart failure (HR, 1.37).
No significant association was seen with cardiac arrest.
The team then stratified the cohort into quintiles and compared the top 20%, middle 60%, and bottom 20%. The relationship between the sphericity index and risk extended across the distribution; individuals with higher than median sphericity had increased disease incidence, and those with lower than median sphericity had decreased incidence.
Overall, a single standard deviation in the sphericity index was associated with increased risk of cardiomyopathy (HR, 1.47) and of AFib (HR, 1.20), independent of clinical factors and usual MRI measurements.
In a minimally adjusted model, the sphericity index was a predictor of incident cardiomyopathy, AFib, and heart failure.
Adjustment for clinical factors partially attenuated the heart failure association; additional adjustment for MRI measurements fully attenuated that association and partially attenuated the association with AFib.
However, in all adjusted models, the association with cardiomyopathy showed little attenuation.
Furthermore, the team identified four loci associated with sphericity at genomewide significance – PLN, ANGPT1, PDZRN3, and HLA DR/DQ – and Mendelian randomization supported NICM as a cause of LV sphericity.
Looking ahead
“While conventional imaging metrics have significant diagnostic and prognostic value, some of these measurements have been adopted out of convenience or tradition,” the authors noted. “By representing a specific multidimensional remodeling phenotype, sphericity has emerged as a distinct morphologic trait with features not adequately captured by conventional measurements.
“We expect that the search space of potential imaging measurements is vast, and we have only begun to scratch at the surface of disease associations.”
Indeed, Dr. Ouyang said his group is “trying to evaluate the sphericity in echocardiograms or heart ultrasounds, which are more common and cheaper than MRI.”
“The main caveat is translating the information directly to patient care,” Richard C. Becker, MD, director and physician-in-chief of the University of Cincinnati Heart, Lung, and Vascular Institute, said in an interview. “Near-term yield could include using the spherical calculation in routine MRI of the heart, and based on the findings, following patients more closely if there is an abnormal shape. Or performing an MRI and targeted gene testing if there is a family history of cardiomyopathy or [of] an abnormal shape of the heart.”
“Validation of the findings and large-scale evaluation of the genes identified, and how they interact with patient and environmental factors, will be very important,” he added.
Nevertheless, “the study was well done and may serve as a foundation for future research,” Dr. Becker said. “The investigators used several powerful tools, including MRI, genomics, and [artificial intelligence] to draw their conclusions. This is precisely the way that ‘big data’ should be used – in a complementary fashion.”
The study authors and Dr. Becker reported no relevant financial relationships.
A version of this article first appeared on Medscape.com.