Deblurring for spiral real‐time MRI using convolutional neural networks

Abstract

Purpose To develop and evaluate a fast and effective method for deblurring spiral real‐time MRI (RT‐MRI) using convolutional neural networks. Methods We demonstrate a 3‐layer residual convolutional neural networks to correct image domain off‐resonance artifacts in speech production spiral RT‐MRI without the knowledge of field maps. The architecture is motivated by the traditional deblurring approaches. Spatially varying off‐resonance blur is synthetically generated by using discrete object approximation and field maps with data augmentation from a large database of 2D human speech production RT‐MRI. The effect of off‐resonance range, shift‐invariance of blur, and readout durations on deblurring performance are investigated. The proposed method is validated using synthetic and real data with longer readouts, quantitatively using image quality metrics and qualitatively via visual inspection, and with a comparison to conventional deblurring methods. Results Deblurring performance was found superior to a current autocalibrated method for in vivo data and only slightly worse than an ideal reconstruction with perfect knowledge of the field map for synthetic test data. Convolutional neural networks deblurring made it possible to visualize articulator boundaries with readouts up to 8 ms at 1.5 T, which is 3‐fold longer than the current standard practice. The computation time was 12.3 ± 2.2 ms per frame, enabling low‐latency processing for RT‐MRI applications. Conclusion Convolutional neural networks deblurring is a practical, efficient, and field map‐free approach for the deblurring of spiral RT‐MRI. In the context of speech production imaging, this can enable 1.7‐fold improvement in scan efficiency and the use of spiral readouts at higher field strengths such as 3 T.

Deep Learning and Its Application to Function Approximation for MR in Medicine: An Overview

“…DNNs were used for deblurring off-resonance blur. 86,87 In the case of dynamic imaging, DNNs approximated functions that received multiple frames as the inputs and produced a single frame as the output, 88 and received multiple frames as the inputs and produced multiple frames as the outputs. 89 As shown in Fig.…”

Real‐Time Magnetic Resonance Imaging

“…Inline flow quantification has been shown to expedite comprehensive cardiac examinations (76). And, inline off-resonance artifact correction (deblurring) has been shown to substantially improves the sharpness of speech articulator depiction (77).…”

“…Additionally, it is popular for rapid post-processing of real-time data; including identification and segmentation of the vocal tract (199)(200)(201)(202), as well as segmentation of the left and right ventricles from real-time cardiac MRI (203). Other applications include rapid needle detection and segmentation in MR-guided interventions (204,205), enabling real-time localization of the Fetal brain (206), spiral off-resonance deblurring in speech imaging (77), and combination of reconstruction and post-processing for real-time MR thermometry (207). Some commercially available software has started using ML in real-time imaging, including HeartVista (208) which uses ML to automate the MRI exam, control the scanner and assist scan planning, and MeVisLab (209) for segmentation and annotation.…”

A multispeaker dataset of raw and reconstructed speech production real-time MRI video and 3D volumetric images

“…The vocal tract contains multiple rapidly moving articulators, which can change position significantly on a millisecond timescale -an imaging challenge necessitating high temporal resolution adequate for observing these dynamic speech process [6]. While under-sampling of MRI measurements on time-efficient trajectories enables the desired resolution, such measurements are hampered by prolonged computation time for advanced image reconstruction, low signal-to-noise-ratio, and artifacts due to undersampling and/or rapid differences of magnetic susceptibility [13,14,15] at the articulator boundaries, which are of utmost interest in characterizing speech production. These limitations often render present day RT-MRI's operating point beneath application demands and can introduce bias and increased variance during data analysis.…”