基于深度学习超分辨率重建的加速高分辨率乳腺T1加权和T2加权磁共振成像

Accelerated High-resolution T1- and T2-weighted Breast MRI with Deep Learning Super-resolution Reconstruction.

作者信息

Mesropyan Narine, Katemann Christoph, Leutner Claudia, Sommer Alexandra, Isaak Alexander, Weber Oliver M, Peeters Johannes M, Dell Tatjana, Bischoff Leon, Kuetting Daniel, Pieper Claus C, Lakghomi Asadeh, Luetkens Julian A

机构信息

Department of Diagnostic and Interventional Radiology, University Hospital Bonn, Venusberg-Campus 1, 53127 Bonn, Germany (N.M., C.L., A.S., A.I., T.D., L.B., D.K., C.C.P., A.L., J.A.L.).

Philips Healthcare, Hamburg, Germany (C.K., O.M.W.).

出版信息

Acad Radiol. 2025 Jun;32(6):3147-3156. doi: 10.1016/j.acra.2024.12.055. Epub 2025 Jan 9.

DOI:10.1016/j.acra.2024.12.055

PMID:39794159

Abstract

RATIONALE AND OBJECTIVES

To assess the performance of an industry-developed deep learning (DL) algorithm to reconstruct low-resolution Cartesian T1-weighted dynamic contrast-enhanced (T1w) and T2-weighted turbo-spin-echo (T2w) sequences and compare them to standard sequences.

MATERIALS AND METHODS

Female patients with indications for breast MRI were included in this prospective study. The study protocol at 1.5 Tesla MRI included T1w and T2w. Both sequences were acquired in standard resolution (T1 and T2) and in low-resolution with following DL reconstructions (T1 and T2). For DL reconstruction, two convolutional networks were used: (1) Adaptive-CS-Net for denoising with compressed sensing, and (2) Precise-Image-Net for resolution upscaling of previously downscaled images. Overall image quality was assessed using 5-point-Likert scale (from 1=non-diagnostic to 5=excellent). Apparent signal-to-noise (aSNR) and contrast-to-noise (aCNR) ratios were calculated. Breast Imaging Reporting and Data System (BI-RADS) agreement between different sequence types was assessed.

RESULTS

A total of 47 patients were included (mean age, 58±11 years). Acquisition time for T1 and T2 were reduced by 51% (44 vs. 90 s per dynamic phase) and 46% (102 vs. 192 s), respectively. T1 and T2 showed higher overall image quality (e.g., 4 [IQR, 4-4] for T1 vs. 5 [IQR, 5-5] for T1, P<0.001). Both, T1 and T2 revealed higher aSNR and aCNR than T1 and T2 (e.g., aSNR: 32.35±10.23 for T2 vs. 27.88±6.86 for T2, P=0.014). Cohen k agreement by BI-RADS assessment was excellent (0.962, P<0.001).

CONCLUSION

DL for denoising and resolution upscaling reduces acquisition time and improves image quality for T1w and T2w breast MRI.

摘要

原理与目的

评估一种由行业开发的深度学习（DL）算法对低分辨率笛卡尔T1加权动态对比增强（T1w）和T2加权涡轮自旋回波（T2w）序列进行重建的性能，并将其与标准序列进行比较。

材料与方法

本前瞻性研究纳入了有乳腺MRI检查指征的女性患者。1.5特斯拉MRI的研究方案包括T1w和T2w。这两个序列均以标准分辨率（T1和T2）以及低分辨率并进行后续DL重建（T1和T2）采集。对于DL重建，使用了两个卷积网络：（1）用于压缩感知去噪的自适应CS网络，以及（2）用于对先前下采样图像进行分辨率提升的精确图像网络。使用5点李克特量表（从1 = 非诊断性到5 = 优秀）评估整体图像质量。计算表观信噪比（aSNR）和对比噪声比（aCNR）。评估不同序列类型之间的乳腺影像报告和数据系统（BI-RADS）一致性。

结果

共纳入47例患者（平均年龄，58±11岁）。T1和T2的采集时间分别减少了51%（每个动态期44秒对90秒）和46%（102秒对192秒）。T1和T2显示出更高的整体图像质量（例如，T1为4 [四分位距，4 - 4]，而T1为5 [四分位距，5 - 5]，P<0.0０1）。T1和T2的aSNR和aCNR均高于T1和T2（例如，aSNR：T2为32.35±10.23，而T2为27.88±6.86，P = 0.014）。通过BI-RADS评估的科恩k一致性极佳（0.962，P<0.0０1）。