Abstract
Quantitative first-pass myocardial perfusion imaging has demonstrated great diagnostic and prognostic utility in coronary artery disease (CAD), which affects 18.2 million patients in the US and is responsible for 1 in 7 deaths. A retrospective analysis of patients without known heart disease who underwent elective x-ray coronary angiography, the invasive gold standard, determined that only 38% of patients had obstructive CAD, which only increased to 41% with a positive non-invasive test. Furthermore, ~40% of the patients in this retrospective study who had a coronary angiography did not have significant obstructive CAD. The low diagnostic yield of coronary angiography highlights a clinical need for improved non-invasive detection of obstructive CAD, where patients are unnecessarily undergoing this invasive procedure.
Although comprehensive cardiovascular magnetic resonance (CMR) exams could be the ideal cardiovascular imaging modality, Cartesian CMR quantitative perfusion techniques suffer from limited ventricular coverage, low spatial resolution, and sensitivity to motion artifacts. Our group has developed high resolution, highly accelerated acquisition techniques for quantitative perfusion, leveraging spiral k-space trajectories to improve scanning efficiency for more robust whole-heart coverage. The image processing and quantitative perfusion modeling remain a challenge at 3T for spiral perfusion imaging.
In particular, our spiral k-space acquisition for whole-heart quantitative perfusion imaging at 3T may be sensitive to saturation efficiency and B1-inhomogeneity, which may lead to biases in the estimation of T1 values. Additionally, the advanced reconstruction techniques used for non-Cartesian imaging may also bias the calculated quantitative perfusion values.
Furthermore, motion correction remains a challenge for quantitative perfusion imaging due to the changing signal intensity during the first pass of gadolinium. Neural networks may be used to automatically segment the myocardium, where the output segmentation contours then can be used for motion correction. Correcting for motion may improve the output quantitative perfusion maps, which coupled with CMR’s high transmural-spatial resolution, may improve the diagnostic accuracy for the detection of CAD and microvascular disease (MVD).
I hypothesize that our high resolution, whole heart quantitative perfusion acquisition scheme, coupled with an improved, novel pipeline (Aim 1) will measure voxel-wise quantitative perfusion in porcine subjects with CAD and MVD (Aim 2) with excellent agreement to our manual processing pipeline. Aim 3 will probe the diagnostic accuracy of spiral whole-heart quantitative perfusion imaging at 3T with the novel pipeline from Aim 1.
Specific Aim 1: to (A) develop and optimize a robust Bloch simulation scheme for our 3T acquisitions which accounts for saturation efficiency, slice profile effects, and B1-inhomogeneity that may cause bias in the estimated T1 for both the tissue function and arterial input function images. (B) Train a neural network for automatic segmentation of the myocardium for motion correction. (C) Finally, to validate an automatic processing pipeline with reconstruction, machine learning segmentation for improved registration, and model fitting for our high resolution, whole heart quantitative perfusion sequence.
Specific Aim 2: to evaluate the temporal evolution of perfusion defects in porcine models of (A) coronary artery disease and (B) microvascular disease with CMR. I hypothesize that our novel quantitative perfusion sequence, coupled with the advancements in Aim 1, will detect changes in the quantification of ischemic, infarcted, and remote myocardial regions in MVD and CAD models of cardiovascular disease. This aim serves two purposes: to validate our quantitative perfusion technique while also characterizing the perfusion changes associated with these disease models.
Specific Aim 3: to clinically evaluate the diagnostic accuracy of the spiral whole heart quantitative perfusion imaging at 3T with the automatic pipeline from Aim 1. Patients with suspected coronary artery disease will undergo comprehensive CMR exams, coupled with coronary angiographies as the ground truth. I hypothesize that quantitative perfusion imaging will outperform visual analysis for diagnostic accuracy.
