A preliminary study of deep learning-based compressed sensing accelerated mDIXON for segmented coronary adipose tissue evaluation in patients with suspected coronary artery disease

Background

The secretion of dysfunctional PCAT is positively correlated with coronary artery stenosis, degree of calcification, and plaque progression. It is important to develop novel clinical diagnostic tools for coronary heart disease based on PCAT assessment. Homsi et al. introduced and validated coronary magnetic resonance angiography (MRA), based on the three-dimensional (3D)-modified Dixon (mDIXON) technique, for epicardial adipose tissue quantification. Therefore, the present study was to use non-contrast-enhanced compressed sensing artificial intelligence framework 3D mDIXON coronary MRA for PCAT quantification in patients with suspected CAD. It also evaluated segmented PCAT's relationship with coronary plaque characteristics and stenosis severity.

Methods

The study protocol was approved by the institutional ethics committee of the hospital. We included 35 symptomatic patients with CAD (111 arteries with plaque, 169 without plaque) (Figure 1). All the subjects underwent CMR on a 3T clinical MR scanner to evaluate segmented PCAT volume and fat-fraction of 8 coronary segments. We manually traced the segmented PCAT volume, and calculated the fat fraction of the segmented PCAT by formula: only fat images (F)/F + only water images (W). We compared the segmented PCAT volume and fat-fraction across 8 coronary segments with different plaque types and degrees of stenosis defined with CCTA and explored the relationship between them.

Results

The coronary segments with plaques had a higher segmented PCAT volume and fat-fraction than those without plaques. Meanwhile, segmented PCAT volume around mixed plaques was larger than non-calcified or calcified plaques (p = 0.014 and p < 0.001) (Figure 3). There was a moderate correlation between the segmented PCAT volume and plaque type (r = 0.493, p < 0.001). The fat-fraction had similar results (r = 0.480, p < 0.001).

Conclusion

The non-contrast-enhanced, whole-heart coronary MRA framework with CSAI is able to measure segmented PCAT volume and fat-fraction. The segmented PCAT volume is more significantly associated with the coronary plaque characters than fat-fraction.