1342938 - Real-time Cine imaging with Convolutional Neural Network Transformer (CNNT) for Denoising

Real-time cine imaging does not require breath-holding and is a robust cine imaging technique in the presence of irregular heartbeats. It is a good alternative to the conventional breath-hold retro-gated cine for simplified acquisition and improved patient comfort. To maintain good temporal and spatial resolution, higher acceleration (e.g. rate 4 or higher parallel imaging) is required. As a result, the real-time cine images experience reduced signal-to-noise ratio (SNR), which limits its clinical acceptance. We developed a novel deep learning model architecture, the Convolutional Neural Network Transformer (CNNT), to improve the quality of real-time cine, under high acceleration.

Methods:

Convolutional Neural Networks (CNN) are widely used in CMR research to process cardiac images. Cardiac images are often acquired as a time series with strong temporal correlation. We combined the CNN with the more recent transformer model to develop a novel CNNT architecture. It takes in the entire 2D+T time series as input and has advantages of CNN for efficient computation and spatial invariance. It further inherits the advantages of an attention layer in the transformer and is able to efficiently exploit temporal correlation within a time series of images.

A CNNT model was developed to improve the SNR of cine imaging. N=10 patients were scanned at the Barts Heart Center, with 4x, 5x and 6x acceleration. Typical imaging parameters are: FOV 360×270mm2, flip angle 50°, acquired matrix size 160×90 for R=4 acceleration, 192×108 for R=5 and 6, with temporal resolution 40ms for R=4, 42ms for R=5 and 35ms for R=6. The real-time images went through a TGRAPPA reconstruction [1] and the proposed CNNT model. The SNR of TGRAPPA was measured by means of SNR units reconstruction [2]. A Monte-Carlo pseudo-replica test was used to measure SNR for the CNNT model. For every cine series, two phases were picked for the end-systole and end-diastole. For every image picked, regions-of-interests were drawn in the myocardium and in the LV blood pool. The CNNT model was deployed inline on the MR scanner using the Gadgetron InlineAI [3].

Results:

Figure 1 gives real-time cine images for three accelerations, reconstructed with and without CNNT denoising. The parallel imaging TGRAPPA reconstruction suffers significant SNR loss from elevated g-factor and less acquired data. By applying the deep learning CNNT model SNR is recovered even at the very high 6x acceleration, without observed loss of boundary sharpness.

Table 1 lists the SNR measurement results. Without CNNT SNR decreased 4-fold from R=4 to R=6 in both the blood and myocardium. With CNNT the SNR in the blood increased by 170%, 335%, 371% at R=4, 5 and 6. For the myocardium, the SNR increases were 335%, 634% and 828%.

Conclusion:

We developed a convolutional neural network transformer model to significantly improve the SNR for real-time cine imaging at higher acceleration.