南方医科大学学报

南方医科大学学报 ›› 2022, Vol. 42 ›› Issue (7): 1019-1025.doi: 10.12122/j.issn.1673-4254.2022.07.08

• • 上一篇    下一篇

基于多模态MRI脑影像的超分辨率重建

曹泽红，刘高平，张志强，石 峰，张 煜   

  1. 南方医科大学生物医学工程学院，广东 广州 510515；广东省医学图像处理重点实验室，广东 广州 510515；上海联影智能医疗科技有限公司，上海 200030；东部战区总医院（南京大学医学院附属金陵医院）放射诊断科，江苏 南京 210002
  • 出版日期:2022-07-20 发布日期:2022-07-15

Multimodality-based super-resolution reconstruction for routine brain magnetic resonance images

CAO Zehong, LIU Gaoping, ZHANG Zhiqiang, SHI Feng, ZHANG Yu   

  1. School of Biomedical Engineering, Southern Medical University, Guangzhou 510515, China; Guangdong Provincial Key Laboratory of Medical Image Processing, Guangzhou 510515, China; Shanghai United Imaging Intelligence Co., Ltd., Shanghai 200030, China; Department of Diagnostic Radiology, Affiliated Jinling Hospital, Medical School of Nanjing University, Nanjing 210002, China
  • Online:2022-07-20 Published:2022-07-15

RichHTML

51

PDF

522

摘要： 目的 探讨基于多模态脑影像数据的超分辨率合成模型将低分辨率的厚层数据重建成为高分辨率的薄层数据。方法 使用真实成对的多模态低-高分辨率MRI数据（2D-T1，2D-T2 FLAIR和3D-T1）设计结构约束的图像超分辨率重建网络，从不同模态的低分辨率MRI提取重要特征重建更高分辨率的T1图像。将T1作为主要模态使用图像全部信息，T2 FLAIR作为补充模态选取皮层下核团为关键区域进行信息增强。通过比较超分辨率重建图像与真实的高分辨率图像之间的灰度和结构相似性来确定网络的学习方向，同时通过脑分割工具获取重建图像和金标准图像的大脑解剖学结构信息，并将其作为重要约束条件来让重建模型自适应的学习大脑的组织结构特征，从而有效提升模型的重建性能。结果 在整体图像质量评价方面，该模型在149例测试集上的重建图像的平均PSNR值为33.11，SSIM为0.996，质量优于本文的其余对比方法生成的结果。在大脑解剖结构方面，我们的方法可以重建出较为清晰的脑沟、脑回以及皮层下核团，可视化结果显示了根据医学图像特性加入解剖学结构信息的有效性。分别使用单模态T1和多模态T1、T2 FLAIR进行图像重建的结果说明了有效选择第二模态关键区域的可行性。同时，在高分辨率图像作为金标准的情况下，使用本文提出的方法重建得到的超分辨率图像与使用低分辨率图像相比，在大脑灰质、白质和脑脊液上的体积测量平均精度有了较大的提升，灰质体积平均误差从3%降到1%，白质从18%降为了2%，脑脊液从35%降为了8%。结论 基于多模态的MRI脑影像超分辨率模型加入了同一组织的不同模态信息与解剖学信息，相比现有的方法，可以重建出更为接近真实高分辨率的图像，有望将其应用于临床诊疗流程中。

关键词: 图像超分辨率重建；脑影像；MRI；多模态；解剖结构约束

Abstract: Objective To propose a multi-modality-based super- resolution synthesis model for reconstruction of routine brain magnetic resonance images (MRI) with a low resolution and a high thickness into high-resolution images. Methods Based on real paired low-high resolution MRI data (2D T1, 2D T2 FLAIR and 3D T1), a structure-constrained image mapping network was used to extract important features from the images with different modalities including the whole T1 and subcortical regions of T2 FLAIR to reconstruct T1 images with higher resolutions. The gray scale intensity and structural similarities between the super-resolution images and high-resolution images were used to enhance the reconstruction performance. We used the anatomical information acquired from segment maps of the super-resolution T1 image and the ground truth by a segmentation tool as a significant constraint for adaptive learning of the intrinsic tissue structure characteristics of the brain to improve the reconstruction performance of the model. Results Our method showed the performance on the testing dataset than other methods with an average PSNR of 33.11 and SSIM of 0.996. The anatomical structure of the brain including the sulcus, gyrus, and subcortex were all reconstructed clearly using the proposed method, which also greatly enhanced the precision of MSCSR for brain volume measurement. Conclusion The proposed MSCSR model shows excellent performance for reconstructing super-resolution brain MR images based on the information of brain tissue structure and multimodality MR images.

Key words: image super resolution; brain images; magnetic resonance imaging; multimodality; anatomical information