Diastolic function assessment with four-dimensional flow cardiovascular magnetic resonance using automatic deep

Federica Viola1, Mariana Bustamante2, Ann Bolger3

  • 1Division of Diagnostics and Specialist Medicine, Department of Health, Medicine and Caring Sciences, Linköping University, Linköping, Sweden; Center for Medical Image Science and Visualization (CMIV), Linköping University, Linköping, Sweden.

Insights

This study compared automated and semi-automated 4D Flow CMR methods for assessing the E/A ratio in diastolic dysfunction. The deep learning-based automated method showed strong association with echocardiography and the fastest results.

Area of Science:

  • Cardiovascular Imaging
  • Medical Technology
  • Diastolic Function Assessment

Background:

  • Diastolic left ventricular (LV) dysfunction significantly impacts heart failure symptoms and prognosis.
  • The E/A ratio, derived from transmitral flow velocities, is crucial for grading diastolic dysfunction, especially in patients with reduced LV systolic function.
  • Doppler echocardiography is standard for diastolic function assessment, but cardiovascular magnetic resonance (CMR) is less established.

Purpose of the Study:

  • To compare automated and semi-automated 4D Flow CMR techniques for assessing the E/A ratio against conventional echocardiography.
  • To evaluate the performance of a deep learning-based automated method versus semi-automated approaches for 4D Flow CMR E/A ratio calculation.
  • To determine the feasibility of using 4D Flow CMR for diastolic function assessment in patients with cardiovascular disease.

Main Methods:

  • Ninety-seven subjects with chronic ischemic heart disease underwent both echocardiography and CMR.
  • Three 4D Flow CMR methods were used: two semi-automated (MVvel, MVflow) measuring inflow at the mitral valve plane, and one fully automated (LVvel) using deep learning for LV segmentation.
  • E/A ratios were computed using all three 4D Flow methods and compared to echocardiography-derived E/A ratios.

Main Results:

  • All three 4D Flow CMR methods (MVvel, MVflow, LVvel) demonstrated strong associations with the echocardiography-derived E/A ratio (R² = 0.60, 0.58, 0.72).
  • The automated LVvel method showed the strongest correlation (R² = 0.72) and fastest runtime (∼40 seconds).
  • Semi-automated methods (MVvel, MVflow) showed very strong association with the LVvel method, suggesting their utility as alternatives.

Conclusions:

  • The E/A ratio assessed by 4D Flow CMR, particularly the automated deep learning method, strongly correlates with Doppler echocardiography.
  • These 4D Flow CMR methods can expand the utility of CMR in assessing diastolic dysfunction in cardiovascular disease patients.
  • While absolute velocities may be underestimated, the E/A ratio derived from 4D Flow CMR is reliable, with the automated approach offering the best performance.
Abstract