A Comparison of Methods for High-Spatial-Resolution Diffusion-weighted Imaging in Breast MRI

Abstract

Background Diffusion-weighted imaging (DWI) shows promise in detecting and monitoring breast cancer, but standard spin-echo (SE) echo-planar DWI methods often have poor image quality and low spatial resolution. Proposed alternatives include readout-segmented (RS) echo-planar imaging and axially reformatted (AR)-simultaneous multislice (SMS) imaging. Purpose To compare the resolution and image quality of standard SE echo-planar imaging DWI with two high-spatial-resolution alternatives, RS echo-planar and AR-SMS imaging, for breast imaging. Materials and Methods In a prospective study (2016-2018), three 5-minute DWI protocols were acquired at 3.0 T, including standard SE echo-planar imaging, RS echo-planar imaging with five segments, and AR-SMS imaging with four times slice acceleration. Participants were women undergoing breast MRI either as part of a treatment response clinical trial or undergoing breast MRI for screening or suspected cancer. A commercial breast phantom was imaged for resolution comparison. Three breast radiologists reviewed images in random order, including clinical images indicating the lesion, images with b value of 800 sec/mm², and apparent diffusion coefficient (ADC) maps from the three randomly labeled DWI methods. Readers measured the longest dimension and lesion-average ADC on three DWI methods, reported measurement confidence, and rated or ranked the quality of each image. The scores were fit to a linear mixed-effects model with intercepts for reader and subject. Results The smallest feature (1 mm) was only detectible in a phantom on images from AR-SMS DWI. Thirty lesions from 28 women (mean age, 50 years ± 13 [standard deviation]) were evaluated. On the five-point Likert scale for image quality, AR-SMS imaging scored 1.31 points higher than SE echo-planar imaging and 0.74 points higher than RS echo-planar imaging, whereas RS echo-planar imaging scored 0.57 points higher than SE echo-planar imaging (all P < .001). Conclusion The axially reformatted simultaneous multislice protocol was rated highest for image quality, followed by the readout-segmented echo-planar imaging protocol. Both were rated higher than the standard spin-echo echo-planar imaging. © RSNA, 2020 Online supplemental material is available for this article. See also the editorial by Partridge in this issue.

Graphical abstract

Figure 1:

Screenshot of a picture archiving and communication system setup in a 70-year-old participant with large biopsy-proven cancer. Readers were provided with clinical images, including, G, a noncontrast-enhanced image, H, a single contrast-enhanced image, I, a contrast-enhanced subtraction T1-weighted image, and, J, a T2-weighted image, with the lesion indicated (arrow). Diffusion data included axial images with, A–C, b value of 800 sec/mm² and, D–F, apparent diffusion coefficient (ADC) maps for all three methods, randomly ordered as methods A, B, and C, which are labeled here for illustration purposes. The metal needle of the contrast injection port caused the artifact on the right breast. AR-SMS = axially reformatted–simultaneous multislice, RS-EPI = readout-segmented echo-planar imaging.

Figure 2:

Resolution phantom comparing feature detection on b = 0 sec/mm² images. Key indicates feature sizes. The smallest feature (1 mm), indicated by an arrow, was only considered visible on the axially reformatted–simultaneous multislice (AR-SMS)and T2-weighted axial images. The 1 mm and 1.25 mm dots in the resolution grids are differentiable on axially reformatted simultaneous multislice and T2-weighted images. RS-EPI = readout-segmented echo-planar imaging.

Figure 3:

Small lesion example. Shown are, A, a contrast-enhanced subtraction image, E, a T2-weighted image, B–D, images with b value of 800 sec/mm², and, F–H, apparent diffusion coefficient (ADC) maps, focused on an example of a small contrast-enhanced lesion (arrow). Radiologists were asked to measure the longest dimension on images with b value of 800 sec/mm² (B–D) and a lesion-average ADC by drawing a freehand two-dimensional region of interest on each ADC map (F–H). The longest lesion diameter was 4.7 mm, measured on the contrast-enhanced subtraction image (A) and averaged across all readers. Average measurements on images with b value of 800 sec/mm² were as follows: B, standard, 5.8 mm; C, readout-segmented (RS) echo-planar imaging (EPI), 4.6 mm; and D, axially reformatted (AR)–simultaneous multislice (SMS) image, 5.5 mm. Average ADC measurements were as follows: F, standard, 1.45 × 10⁻³ mm²/sec; G, RS echo-planar image, 1.59 ×10⁻³ mm²/sec; and H, R-SMS image, 1.34 ×10⁻³ mm²/sec. The average quality scores on a five-point LIkert scale were as follows: standard, 2.3; RS echo-planar imaging, 2.7; and AR-SMS imaging, 3.7.

Figure 4:

Large lesion example. Shown are, A, contrast-enhanced image, E, T2-weighted image, B–D, images with b value of 800 sec/mm², and, F–H, apparent diffusion coefficient (ADC) maps, which are focused on an example of a large contrast-enhanced lesion (arrow). Radiologists were asked to measure the longest dimension on (B–D) images with b value of 800 sec/mm² and (F–H) a lesion-average ADC by drawing a freehand two-dimensional region of interest on each ADC map. The longest lesion diameter was 28.8 mm, measured on the contrast-enhanced subtraction image and averaged across all readers. Average measurements on images with b value of 800 sec/mm² were as follows: B, standard, 26 mm; C, readout-segmented (RS) echo-planar imaging, 25.6 mm; and D, axially reformatted (AR)–simultaneous multislice (SMS) image, 26 mm. Average ADC measurements were as follows: F, standard, 0.88 ×10⁻³ mm²/sec; G, RS echo-planar imaging, 0.93 ×10⁻³ mm²/sec; and, H, AR-SMS imaging, 0.84 ×10⁻³ mm²/sec. The average quality scores on a five-point Likert scale were as follows: standard, 2.3; RS echo-planar imaging, 3.0; and AR-SMS imaging, 4.0.

Figure 5:

Summary of reader study results. A, Mean overall quality scores on a per-reader basis and, B, method comparison according to linear mixed-effects model accounting for participant and reader. Error bars, A, indicate standard error. Readers consistently scored axially reformatted (AR)–simultaneous multislice (SMS) imaging higher than readout-segmented (RS) echo-planar imaging (EPI) and standard spin-echo (SE) echo-planar imaging. C, Histogram of relative rank across all three readers and all lesions and, D, comparison of ranks according to linear mixed-effects model including P values. AR-SMS imaging was most frequently rated first, followed by RS echo-planar imaging at second. * Indicates statistical significance on the basis of Tukey-adjusted P values. ACRIN = American College of Radiology Imaging Network, CI = confidence interval.

Figure 6:

Confidence of lesion size and apparent diffusion coefficient (ADC) measurements. Histograms of confidence ratings in measurement of, A, lesion size and, C, ADC on diffusion-weighted images (DWIs) across all readers and lesions. B, D, Method comparison represents the linear mixed-effects model accounting for participant and reader. Axially reformatted (AR)–simultaneous multislice (SMS) imaging was rated with the highest confidence in lesion size measurements on images with b values of 800 sec/mm², followed by, B, readout-segmented (RS) echo-planar imaging (EPI), and then standard spin-echo echo-planar imaging, with statistical significance. D, The ADC confidence did not depend on the DWI method. * Statistical significance indicated on the basis of Tukey-adjusted P values.