Advanced Abdominal MRI Techniques and Problem-Solving Strategies

Abstract

MRI plays an important role in abdominal imaging because of its ability to detect and characterize focal lesions. However, MRI examinations have several challenges, such as comparatively long scan times and motion management through breath-holding maneuvers. Techniques for reducing scan time with acceptable image quality, such as parallel imaging, compressed sensing, and cutting-edge deep learning techniques, have been developed to enable problem-solving strategies. Additionally, free-breathing techniques for dynamic contrast-enhanced imaging, such as extra-dimensional-volumetric interpolated breath-hold examination, golden-angle radial sparse parallel, and liver acceleration volume acquisition Star, can help patients with severe dyspnea or those under sedation to undergo abdominal MRI. We aimed to present various advanced abdominal MRI techniques for reducing the scan time while maintaining image quality and free-breathing techniques for dynamic imaging and illustrate cases using the techniques mentioned above. A review of these advanced techniques can assist in the appropriate interpretation of sequences.

Keywords: Deep Learning Reconstruction; Liver; MRI Scan; Magnetic Resonance Imaging.

Fig. 1. Representative diagrams of k-space sampling.

A-C. Fully sampled k-space (A), parallel imaging (B), and compressed sensing (C).

Fig. 2. T2-weighted images using parallel imaging (A-C) and not using parallel imaging (D).

A, B. First image (A) and fat-suppressed images (B) were obtained with T2 HASTE with partial Fourier factor 5/8 and image acquisition time if both are 1 minute (acceleration factor, 2, respectively), and both display comparable image quality and spatial resolution (reconstructed voxel size A: 0.5 × 0.5 × 5.0 mm³, B: 0.6 × 0.6 × 5.0 mm³). C, D. HASTE with parallel imaging (C) (GRAPPA, acceleration factor phase encoding 2, voxel size 0.5 × 0.5 × 5.0 mm³, TR 1000 ms, TE 88 ms, matrix 384 × 312, FOV 360 × 293, partial Fourier factor 5/8) and without parallel imaging (D) (TR 1200 ms, TE 88 ms, matrix 384 × 312, FOV 360 × 293, partial Fourier factor 4/8). Without parallel imaging, the image displays increasing blurring artifacts due to longer echo-train lengths (D). *All figures were obtained using a 3T scanner (MAGNETOM Vida; Siemens Healthcare, Erlangen, Germany). FOV = field of view, GRAPPA = generalized auto-calibrating partially parallel acquisition, HASTE = half Fourier single-shot turbo spin-echo, TE = echo time, TR = repetition time

Fig. 3. Dual arterial phase images in a 67-year-old female.

Images were obtained using CAIPIRINHA with an acceleration factor of 5 (acceleration factor phase encoding 1, acceleration factor 3D 5, Reordering shift 3D 2) and 0.5 × 0.5 × 3.0 mm³ voxel size. A. The first image was obtained in the early arterial phase. B. The next image was obtained in the late arterial phase. Image acquisition time is 14 seconds in total including a preparation scan of 2 seconds, and two arterial phases (each 6 seconds, respectively). C, D. Subtraction arterial phase images were reconstructed using (A, B) images and pre-contrast images in early and late arterial phases. CAIPIRINHA = controlled aliasing in parallel imaging results in higher acceleration

Fig. 4. Hepatobiliary phase using CAIPIRINHA and compressed sensing in gadoxetic acid-enhanced liver MRI in a 63-year-old male with liver cirrhosis and chronic hepatitis B.

A. First hepatobiliary phase image was obtained using CAIPIRINHA with an acceleration factor of 4 (acceleration factor phase encoding 2, acceleration factor 3D 2, reordering shift 3D 1; reconstructed voxel size, 1.3 × 1.3 × 3.0 mm³, image acquisition time, 16 seconds). B. Next hepatobiliary phase image was obtained using compressed sensing with acceleration factor of 8 (reconstructed voxel size, 0.9 × 0.9 × 1.5 mm³, image acquisition time, 15 seconds). Non-respiratory motion artifacts are visible in the image (arrows), which slightly decreases overall image quality. CAIPIRINHA = controlled aliasing in parallel imaging results in higher acceleration

Fig. 5. Non-respiratory motion artifacts using a compressed sensing research application on hepatobiliary phase (reconstructed voxel size, 0.9 × 0.9 × 1.5 mm³; total acceleration factor of 8) of gadoxetic acid-enhanced liver MRI.

A, B. Hepatobiliary phase images in a 61-year-old male with chronic hepatitis C display moderate non-respiratory artifacts related to incoherent sampling (arrows). C. The hepatobiliary phase image in a 61-year-old male with liver cirrhosis displays non-respiratory artifacts including cardiac ghosting artifact (arrow) in the left lateral segment. D, E. The hepatobiliary phase in a 62-year-old male with liver cirrhosis and chronic hepatitis B displays non-respiratory artifacts including ringing artifacts (arrow, magnification image).

Fig. 6. MR cholangiopancreatography images in a 67-year-old female with chronic pancreatitis.

A. The first image was obtained using CAIPIRINHA with an acceleration factor of 4 (acceleration factor phase encoding 2, acceleration factor 3D 2, reordering shift 3D 1). B. The next image was obtained using compressed sensing with an acceleration factor of 25. Although both images display comparable image quality and spatial resolution (reconstructed voxel size A: 0.5 × 0.5 × 1.0 mm³, B: 0.5 × 0.5 × 0.9 mm³), image acquisition time is 3 minutes 46 seconds in (A) and 18 seconds in (B).

Fig. 7. Hepatobiliary phase using CAIPIRINHA and DL in gadoxetic acid-enhanced liver MRI in a 57-year-old man with chronic hepatitis B.

A-C. Images were obtained using CAIPIRINHA with an acceleration factor of 4 (acceleration factor phase encoding 2, acceleration factor 3D 2, reordering shift 3D 1; 16 seconds; matrix 384 × 270). D-F. Images were obtained with a research DL reconstruction (k-space DL reconstruction) and CAIPIRINHA with an acceleration factor of 6 (acceleration factor phase encoding 3, acceleration factor 3D 2, reordering shift 3D 1; 12 seconds; matrix 448 × 276). CAIPIRINHA = controlled aliasing in parallel imaging results in higher acceleration, DL = deep-learning

Fig. 8. T2-weighted images using T2 HASTE and DL in liver MRI in a 45-year-old male with liver cirrhosis and chronic hepatitis C.

A-C. Images were obtained with HASTE (GRAPPA; acceleration factor 2; acquisition time 60 seconds; matrix 320 × 221; partial Fourier factor 7/8; reconstructed voxel size, 1.3 × 1.3 × 5.0 mm³). D-F. DL images were obtained using HASTE (acceleration factor 3; acquisition time 55 seconds; 384 × 218; partial Fourier factor 7/8; 0.5 × 0.5 × 5.0 mm³) with a research DL reconstruction. Although DL HASTE displays high spatial resolution with improved edge sharpness and reduced noise, it was obtained with a short image acquisition time compared to HASTE without DL. DL = deep-learning, HASTE = half-Fourier acquisition single-shot turbo spin-echo

Fig. 9. T2-weighted images using BLADE with DL reconstruction in liver MRI in a 71-year-old male with hepatic metastasis.

A. The image was obtained with BLADE (GRAPPA; acceleration factor 3; 16 minutes; matrix 320 × 320; partial Fourier factor 7/8; reconstructed voxel size, 1.3 × 1.3 × 5.0 mm³). B. DL image was obtained using BLADE (3; 55 seconds; 288 × 288; 7/8; 1.3 × 1.3 × 5.0 mm³) with a research DL reconstruction. Both images display mild focal dilatation of B5 with improved bile duct sharpness, reduced image noise, and decreased image acquisition time dramatically using DL compared to those using conventional T2 BLADE without DL reconstruction (A). DL = deep-learning, GRAPPA = generalized auto-calibrating partially parallel acquisition

Fig. 10. Dynamic phases in gadoxetic acid-enhanced liver MRI and diffusion-weighted imaging, apparent diffusion coefficient map using deep-learning in liver MRI in a 54-year-old male with liver cirrhosis.

Approximately 1 cm nodule in segment six radiologically probable hepatocellular carcinoma displays arterial phase hyperenhancement with washout on portal venous phase (not displayed). A, B. DWI (b = 900 sec/mm²) and ADC map were obtained using a research DL reconstruction (GRAPPA; acceleration factor 3; 119 s; matrix 140 × 140, reconstruction voxel size, 1.4 × 1.4 × 5.0 mm³; k-space based DL reconstruction combined with image-based super-resolution). C, D. DWI and ADC maps were obtained without DL reconstruction (GRAPPA; acceleration factor 3; 209 seconds; matrix 140 × 140, reconstructed voxel size, 2.7 × 2.7 × 5.0 mm³). The DL DWI and ADC map demonstrates improved conspicuity of the focal lesion (arrow in A-D) and clearer edges, with reduced image noise, despite a dramatically decreased image acquisition time, compared to conventional DWI and ADC maps. ADC = apparent diffusion coefficient, DL = deep-learning, DWI = diffusion-weighted imaging, GRAPPA = generalized auto-calibrating partially parallel acquisition

Fig. 11. Free-breathing diffusion-weighted imaging, apparent diffusion coefficient map using motion correction in liver MRI in a 48-year-old male with chronic hepatitis B.

A-D. Free-breathing DWI and ADC map were obtained with motion correction (A, B) DWI and ADC map were obtained without motion correction (C, D). Although both images display comparable overall image quality, (A) the image displays more distinct liver edge sharpness and clarity, even at the left lateral segment (arrows) as well as hepatic vessel clarity compared with a mild blur of liver edge and hepatic vessel on (C) image. ADC = apparent diffusion coefficient, DWI = diffusion-weighted imaging

Fig. 12. Triple arterial phase images in a 45-year-old male with liver cirrhosis and chronic hepatitis C.

All images were from a research application that combines CAIPIRINHA and CS imaging with a total acceleration factor of 14 (CAIPIRINHA Acceleration factor 4; acceleration factor phase encoding 2, acceleration factor 3D 2, reordering shift 3D 1, CS acceleration factor 3.5). A. The first image was obtained in the first arterial phase. B. The next image was from the second arterial phase. C. The third image was obtained in the third arterial phase. Image acquisition time is 19 seconds in total including a preparation scan of 1 second, and three arterial phases (each 6 seconds, respectively). D-F. Subtraction arterial phase images were reconstructed using (A-C) images and pre-contrast images (not displayed) in each arterial phase. Images display excellent image quality and spatial resolution (reconstructed voxel size: 0.5 × 0.5 × 1.2 mm³), without motion artifacts. CAIPIRINHA = controlled aliasing in parallel imaging results in higher acceleration, CS = compressed sensing

Fig. 13. Free-breathing dynamic phases in gadoxetic acid-enhanced liver MRI using XD-VIBE in a 74-year-old male.

Due to the non-cooperative state of the patient, images were obtained with the XD-VIBE research application (compressed sensing; acceleration factor, 5; temporal resolution, 12.5 seconds; the number of measurements, 18; total time, 225 seconds; matrix 352 × 224; reconstruction voxel size, 1.1 × 1.1 × 3.0 mm³) and continuous dynamic MR images in the free-breathing state. A-E. Pre (A), arterial (B), portal (C), and transitional phase (D) MR images display a mass of 4 cm approximately in segment four with gradual centripetal fill-in enhancement and (E) hepatobiliary phase hypointensity, which suggests a hemangioma. XD-VIBE = extra-dimensional-volumetric interpolated breath-hold examination

Fig. 14. Free-breathing dynamic phases in gadoxetic acid-enhanced liver MRI using XD-VIBE in 92-year-old female.

Images were obtained with the XD-VIBE research application (compressed sensing; acceleration factor 5; temporal resolution 12.5 seconds; total 225 seconds; matrix 352 × 224, reconstructed voxel size, 1.1 × 1.1 × 3.0 mm³), a Cartesian acquisition which is relatively sensitive to motion compared with GRASP. Due to its drawback, continuously acquired dynamic MR images in the free-breathing state. A-C. Pre (A), arterial (B), and portal phase (C) display moderate respiratory motion artifacts. GRASP = golden-angle radial sparse paralle, XD-VIBE = extra-dimensional-volumetric interpolated breath-hold examination

Fig. 15. Free-breathing dynamic phases in gadoxetic acid-enhanced liver MRI using SG-GRASP in an 84-year-old female with liver cirrhosis.

As the patient had a short breath-holding capability, the patient underwent continuous dynamic MR images in the free-breathing state during the injection of gadoxetic acid at a rate of 1.6 mL/s. Images were from SG-GRASP (50% of expiratory phase data based on SG motion state, radial spoke, 44; temporal resolution, 13 seconds; matrix 352 × 352). A-C. Pre (A), arterial (B), and portal phase (C) MR images are acceptable without respiratory motion artifacts and display a 3 cm mass in segment seven with arterial phase hyperenhancement (B, arrow) and washout on portal venous phase (C, arrow), suggesting hepatocellular carcinoma. SG-GRASP = self-gated golden-angle radial sparse parallel

Fig. 16. Free-breathing dynamic phases in gadoxetic acid-enhanced liver MRI using SG-GRASP in an 80-year-old female with liver cirrhosis.

Images were obtained with SG-GRASP (50% of expiratory phase data based on SG motion state, radial spoke, 44; temporal resolution, 13 seconds; matrix 352 × 352) and continuous dynamic MR images in the free-breathing state during injection of gadoxetic acid at a rate of 1.6 mL/s. A-C. Pre (A), arterial (B), and portal phase (C) MR images display a mass of approximately 1 cm in segment two with arterial phase hyperenhancement (B, arrow) and washout on portal venous phase (C, arrow) suggesting hepatocellular carcinoma. Images are acceptable without respiratory motion artifacts, but mild streak artifacts. SG-GRASP = self-gated golden-angle radial sparse parallel

Fig. 17. Free-breathing dynamic phases in gadoxetic acid-enhanced liver MRI using SG-GRASP in an 88-year-old female with chronic hepatitis C. Images were obtained with SG-GRASP (50% of expiratory phase data-based on SG motion state, radial spoke, 44; temporal resolution, 13 seconds; matrix 352 × 352) and continuous dynamic MR images in the free-breathing state during injection of gadoxetic acid at a rate of 1.6 mL/s.

A-F. Superior segments of pre (A), arterial (B), portal phase (C) MR images and inferior segments of pre (D), arterial (E), and portal phase (F) MR images display good adequacy of arterial phase timing, (B) and (E) images display mild portal vein opacification with mild parenchymal enhancement, without hepatic venous enhancement, suggestive of good late arterial phase images. SG-GRASP = self-gated golden-angle radial sparse parallel

Fig. 18. Free-breathing dynamic phases in gadoxetic acid-enhanced liver MRI using SG-GRASP in a 91-year-old female.

Images were obtained with SG-GRASP (50% of expiratory phase data based on SG motion state, radial spoke, 44; temporal resolution, 13 seconds; matrix 352 × 352; time axis regularization factor 20) and continuous dynamic MR images in the free-breathing state during injection of gadoxetic acid at a rate of 1.6 mL/s. A-C. Superior segments of pre (A), late arterial (B), and portal phase (C) MR images display too late arterial phase timing. (B) The image displays hepatic venous enhancement, suggestive of hepatic vein contamination in the late arterial phase due to k-space sharing. SG-GRASP = self-gated golden-angle radial sparse parallel