Image Reconstruction (including machine learning)

1356037 - Self-supervised learning with self-supervised regularization reconstruction method for high spatiotemporal fidelity of accelerated myocardial perfusion MRI

Friday, January 27, 2023

11:40 AM – 11:55 AM

Location: Silver Pearl 1

Presenting Author(s)

CS

Changyu Sun, PhD

Assistant Professor
University of Missouri-Columbia
Columbia, Missouri, United States

Primary Author(s)

CS

Changyu Sun, PhD

Assistant Professor
University of Missouri-Columbia
Columbia, Missouri, United States

Co-Author(s)

Kenneth Bilchick, MD

Professor
University of Virginia Health System
Charlottesville, Virginia, United States
Michael Salerno, MD, PhD, MSc

Professor
Stanford University, California, United States
TA

Talissa A. Altes, MD

Professor
University of Missouri-Columbia
Columbia, Missouri, United States

Disclosure(s):

Michael Salerno, MD, PhD, MSc: No relevant disclosure to display

Background: Highly accelerated first-pass myocardial perfusion MRI is useful for improving assessment of coronary artery disease; however, a tradeoff exists between SNR and spatial and temporal fidelity^[1] for parallel imaging and compressed sensing. Self-supervised learning (SSL) of MRI reconstruction^[2,3,4] has high potential to improve spatiotemporal fidelity without fully-sampled k-space as reference^[2,3,4]. However, SSL of physics-guided network (PG-SSL)^[5] is challenged by noise amplification^[4] using one unrolled network and collapsing^[3] using a Siamese structure. We sought to develop a PG-SSL method with self-supervised regularization (SSL-SSR) using a Siamese structure based on stop gradient^[3], re-undersampling block and consistency of two PG networks. Here the new SSL-SSR method is proposed and compared to GRAPPA, ESPIRiT and L+S^[1]methods.

Methods:

Image datasets were acquired on a 1.5T system (Aera, Siemens) using 20-34 RF receiver channels. Twenty-two datasets (2640 images) were acquired using routine saturation-recovery gradient-echo sequence with rate-2 undersampling and reconstructed using GRAPPA. Synthetic rate-8 undersampling data were generated with a 24 lines calibration region. SSL-SSR (Figure 1) is designed with a PG Siamese network, where one with stop-gradient and the other one with an additional Unet. No fully sampled datasets and temporal constraints are used in the loss function. The SSR term is formulated by a re-undersampling block (Figure 1C) and the consistency of two PG unrolled networks(Figure 1A and B). Ten datasets, two datasets and ten datasets were used as training, validation, and testing datasets. Synthetic rate-8 k-t undersampled datasets were generated and reconstructed using ESPIRiT and L+S methods^[1]. Rate-8 ESPIRiT, L+S and SSL-SSR were compared to rate-2 GRAPPA reconstructed images using normalized RMSE, SSIM of the images, and nRMSE of time curves to quantify spatiotemporal fidelity.

Results: Figure 2 illustrates example images at different time points and signal x-t plots from one patient, where rate-8 ESPIRiT, L+S, SSL-SSR and reference rate-2 GRAPPA images at a mid-ventricular location are shown. SSL-SSR shows higher image quality and is more similar temporal fidelity to reference images than ESPIRiT and L+S (Figure 2 and 3C). Quantification of RMSE and SSIM are shown in Figure 3 for all 10 testing datasets. RMSE values were 0.086±0.026, 0.042±0.011 and 0.041±0.009. SSIM values were 0.64±0.09, 0.82±0.06 and 0.84±0.04. RMSE of intensity-time signal were 0.79±0.26, 0.55±0.18 and 0.32±0.13 (*1e-3). Compared with rate-2 GRAPPA images, SSL-SSR shows significantly higher temporal fidelity with the lowest RMSE of time signal (*P < 0.05, ANONA), and slightly higher image quality with highest SSIM and lowest RMSE (*P < 0.05, ANONA).

Conclusion: SSL-SSR outperformed ESPIRiT and L+S methods, and is a promising method for improving spatiotemporal fidelity of the reconstruction of undersampled first-pass perfusion MRI without fully-sampled k-space.