Can cardiac MRI replace echocardiography for diagnosing pulmonary hypertension?

Cardiac MRI can accurately diagnose patients with pulmonary hypertension (PH), according to an Oct. 23 study published in Radiology. Results may help improve PH detection rates and enhance care for patients with severe cases.

Right-sided heart catherization (RHC) is the current reference standard for PH, but its invasive nature and “serious” complication rate make echocardiography the recommended test to classify suspected PH patients, wrote lead author Christopher S. Johns, with Royal Hallamshire Hospital in England, and colleagues. Even so, the techniques’ measurements aren’t as accurate as RHC, and popular metrics, such as black blood scoring, have not been widely adopted in MRI methods.

“Thus, developing models with and without this parameter would be advantageous,” Johns et al. wrote. “Previously identified cardiac MRI metrics and markers have been tested in relatively small numbers and often have modest diagnostic accuracy.”

The researchers analyzed 2,437 MR images from consecutive patients with suspected PH who also received a cardiac MRI and RHC between 2012 and 2016. The group was separated into a 300 patient derivation cohort (used to create a regression algorithm) and 303-patient validation cohort.

The cardiac model, comprised of right ventricle and black blood metrics, had good agreement with RHC-measured mean pulmonary arterial pressure (mPAP) and high diagnostic accuracy (area under the ROC curve of .95). When a threshold of at least 25 mm Hg was used in the validation cohort, that model had a sensitivity of 96 percent, achieved 79 percent specificity, and positive and negative predictive value of 96 and 67 percent, respectively.

The second model, which did not include black blood flow score, achieved the same diagnostic accuracy as model one, the authors reported.

“In conclusion, cardiac MRI has high diagnostic accuracy in a cohort of incident patients referred to a tertiary referral center with suspected PH,” Johns and colleagues concluded. “A reproducible model comprising simple and easy-to-obtain metrics (interventricular septal angle, ventricular mass index, and black blood score) can enable the identification of patients with PH with high accuracy.”

In a corresponding editorial, Patrick M. Colletti, with the USC Keck School of Medicine in Los Angeles, acknowledged that MRI could displace echocardiology when evaluating hypertension.

However, he wrote, the three measurements central to Johns and colleagues’ study require time-consuming analysis and leave potential for varying results. But machine learning may be an answer, Colletti argued.

“Potential solutions to these limitations may be achieved by the future application of deep learning approaches to automate the key cardiac MRI measurements and their analyses so that right-sided heart pressure estimates may accompany all cardiac MRI reports,” Colletti concluded.