多模态MRI探索宫颈癌侵袭性及同步放化疗疗效评估的应用研究

Diffusion-weighted MR imaging (DWI) and dynamic contrast enhanced magnetic resonance imaging (DCE-MRI) play an important role in the identification of histological characteristics for uterine cervical cancer which include histological subtype, grade of differentiation and lymphatic metastases, and in the evaluation of therapeutic response to concurrent chemoradiotherapy (CCRT) for locally advanced cervical cancer as well. Previously noted conventional approach to these images, however, only placed a discrete region of interest (ROI) encompassing a portion of a lesion in the largest lesion slice. Some of the limitations of such ROI-based methods include interobserver variability in ROI placement, difficulty with reproducibility, and failing to take into account tumor heterogeneity which might be one of the reasons that not all tumors of the same type will respond equally to a specific treatment. Furthermore, one of the greatest challenges in cancer management is to develop a method of rapidly and objectively measuring tumor response to therapy, especially select the optimal time window to make proper and effective observation. A reliable and early marker of response would have immense clinical value to avoid unnecessary toxicities and costs, and to personalize treatment strategy. The overall goal of this work is to investigate the combined role of DWI and DCE-MRI (multimodal MRI) in assessing tumor invasiveness and providing early and reproducible response predictors in cervical cancer patients by using 3D Slicer, which is a free, open source software package for image analysis and scientific visualization. 3D slicer generally uses a volumetric approach, incorporating all voxels within the lesion on all slices. Such an analysis could provide multi- metrics from different functional imaging techniques; this goes beyond what can be achieved by using any single functional technique and will provides a more comprehensive evaluation of the entirety of the lesion.

磁共振扩散加权成像（DWI）和动态增强MRI（DCE-MRI）两项功能成像技术在评价宫颈癌组织学类型、病理级别和淋巴结转移状态，以及评估中晚期宫颈癌患者同步放化疗敏感性方面发挥重要作用。然而常规图像分析方法多选取病灶最大层面设置兴趣区，使得研究结果可重复性较差，更无法全面而准确地反映肿瘤组织的异质性，而后者可能是导致同样病期和同一病理级别宫颈癌患者放化疗疗效产生明显差异的重要原因之一。此外，量化评价患者对放化疗敏感性，特别是确定早期评估疗效的时间窗，以便合理制订和及时调整治疗方案，也是妇科肿瘤领域一直尚未解决的重大问题。本项目在前期相关研究的基础之上，运用多模态MRI技术并借助3D Slicer这一免费、开源的可视化图像分析软件平台，获得可靠的图像分割，从整体上深入挖掘图像信息，获取高通量参数，多维度评估肿瘤异质性，进而获取宫颈癌侵袭性和同步放化疗敏感性的早期影像标志物，以便临床推广应用。

影像组学（Radiomics）产生于大数据的背景下，应用自动化数据特征化算法，将医学影像图像中感兴趣区域的影像数据转化为高通量可分析的定量特征，解析影像与基因和临床信息（分型、疗效和预后等）关联，为临床提供一个全面量化的肿瘤表型，从而提高个性化治疗的选择和监测。本课题通过磁共振扩散加权成像（DWI）及动态增强MRI（DCE-MRI）两项功能成像技术的影像组学分析，对MRI评价宫颈癌肿瘤侵袭性及评估中晚期宫颈癌患者同步放化疗效果进行了初步研究。首先筛选可重复性强即对分割相对不敏感的影像组学特征，分析2D单一层面和3D全肿瘤分割两种肿瘤分割方法及DWI检查中b值对组学特征值影响，结果显示对于宫颈癌磁共振影像组学分析，全肿瘤分割优于单一层面肿瘤分割，b值选取1000时鉴别宫颈癌病理分级的效果较好。然后通过影像病理对照分析，探讨影像组学特征评价宫颈癌肿瘤侵袭性的可行性，研究结果表明影像组学分析可以预测宫颈癌患者肿瘤病理分级、淋巴结转移、淋巴血管间隙侵犯、KI-67、VEGF表达程度、MVD计数及HIF-1α表达水平。此外，通过前瞻性收集接受同步放化疗（CCRT）治疗的宫颈癌病例并进行影像学随访，评价影像组学特征预测最终疗效分组的效能，同时对该组病例进行长期随访，以期今后探讨预测患者生存、复发的可行性。研究结果提示放化疗早期，肿瘤的影像组学特征值变化要早于形态学变化。综合上述结果，本研究为探讨多维度评估肿瘤异质性，进而获取宫颈癌侵袭性和同步放化疗敏感性的早期影像标志物奠定了坚实的工作基础，其临床推广应用将为临床医师制定个体化治疗方案提供有价值的信息。