Development of Motion-Model Ultrasound Localization Microscopy to Support Breast Cancer Diagnosis and Therapy Monitoring in Patients

超声成像通常用于乳腺癌筛查和可疑病变的尊严评估。在DFG的支持下，我们开发了一种称为运动模型超声定位显微镜（mULM）的超分辨率对比增强超声方法，该方法跟踪视频序列中的单个微泡，并以超出衍射的分辨率生成血管图像。mULM方法能够准确评估肿瘤中的血管结构、个体血管速度和流动方向以及相对血容量和灌注的量化。最近，我们在小鼠异种移植肿瘤中表明，这种方法可以提取各种形态学参数，并与放射组学分析相结合，实现了不同肿瘤类型的自动区分。在这个项目中，我们的目标是适应和开发mULM和基于超声的放射组学在患者中的应用，并将该技术推进到3D采集和跟踪。在对临床数据集的初步分析中，我们已经能够提取血管轨迹的超分辨率数据，然而，在临床环境中的稳健应用出现了一些技术挑战：特别是，患者运动和较低的图像分辨率以及增加的切片厚度和非优化的注射协议仅允许使用短图像序列和提取较少数量的轨迹。因此，在初始步骤中，将改进mULM方法并使其适用于临床超声设备上的乳腺成像。随后，2D mULM方法将用于监测乳腺癌对新辅助化疗的反应。其准确性将与传统的对比增强超声数据分析进行比较。此外，我们努力推进的方法从2D到全3D采集与实时体积扫描仪和比较两种方法在一个小的队列中的患者不同的尊严的乳腺肿瘤。为了充分利用mULM数据分析的潜力，将开发和实施允许血管特征的自动提取和聚类分析的算法。将进行相似性测量和聚类分析，以自动对患者进行分组。因此，该项目将是对多参数超分辨率超声临床转化的第一个贡献，并为基于超声的放射组学分析开辟道路。

Ultrasound imaging is routinely applied in breast cancer screening and for the dignity assessment of suspect lesions. Supported by the DFG, we developed a super-resolution contrast-enhanced ultrasound method called motion model ultrasound localization microscopy (mULM), which tracks individual microbubbles in video sequences and generates images of the vasculature at a resolution beyond the diffraction. The mULM method enables an accurate assessment of the vascular architecture in tumors, individual blood vessel velocities and flow directions, and the quantification of relative blood volume and perfusion. Recently, we showed in mouse xenograft tumors that this method allows to extract various morphometric parameters and combined with a radiomics analysis enables the automated discrimination of different tumor types. In this project, we aim to adapt and develop mULM and ultrasound-based radiomics for the application in patients and to advance the technique to 3D acquisition and tracking. In a preliminary analysis of clinical datasets, we were already able to extract super-resolution data of vascular tracks.However, several technical challenges for the robust application in a clinical setting emerged: in particular, patient motion and the lower image resolution together with increased slice thickness and non-optimized injection protocols allowed only to use short image sequences and to extract a reduced number of tracks. Thus, in an initial step, the mULM methodology will be refined and adapted for its application in breast imaging on clinical ultrasound devices. Subsequently, the 2D mULM method will be applied to monitor breast cancers’ responses to neoadjuvant chemotherapy. Its accuracy will be compared to conventional contrast-enhanced ultrasound data analyses. Furthermore, we strive to advance the method from 2D to full 3D acquisition with a real-time volumetric scanner and compare both approaches in a small cohort of patients with breast tumors of different dignity. To exploit the full potential of mULM data analysis, algorithms will be developed and implemented that allow the automated extraction and clustered analysis of the vascular features. Measurements of similarity as well as cluster analysis will be performed to automatically group the patients. Thus, this project will be a first contribution to the clinical translation of multiparametric super-resolution ultrasound and open the way for ultrasound-based radiomics analysis.