Doppler Vascularity/Sonography--Breast Cancer Diagnosis

我们以前的研究表明，灰度超声和多普勒成像是互补的。灰度成像对肿块的表征特异性高，而多普勒成像灵敏度高。本应用提出了两种模式的创新集成，从而可以实现高灵敏度和高特异性。一个密集的5年多学科计划包括三个具体目标提出。在Aim 1中，对300例可疑乳腺肿块患者进行彩色多普勒、功率多普勒和灰度图像采集。所有模式的成像将在同一天对同一患者进行。与最佳成像有关的问题将通过仔细控制实验条件来强调。除其他事项外，这将包括定期图像校准和监测患者体内的黄体酮水平，以解释正常月经周期期间血流量的变化以及激素治疗引起的变化。在目标2中，我们介绍了从多普勒和灰度图像中获得定量特征的新方法。重点将放在医生在评估图像时使用的量化特征上。在Aim 3中，这些特征将补充定性评估，以开发一个多因素模型，以全面诊断乳腺病变。基于先进非线性方法的新方法，包括神经网络，将被开发用于制定癌症诊断的决策树。每种策略都将通过ROC分析进行评估。这种方法将提供客观的措施来证明超声和多普勒特征的重要性，以及如何最佳地使用它们来准确诊断乳腺癌。总的来说，这个项目利用了宾夕法尼亚大学独特的临床经验和基础科学的广度。一种模式代表组织特性，另一种模式代表其功能的综合方法将导致乳腺癌的系统和全面的诊断。

Our previous studies suggest that gray-scale ultrasound and Doppler imaging complement one another. While gray-scale imaging can characterize masses with high specificity, Doppler imaging has high sensitivity. This application proposes an innovative integration of the two modes such that both high sensitivity and specificity can be achieved. An intense 5-year multidisciplinary program encompassing three specific goals is proposed. In Aim 1 color-Doppler, power-Doppler and gray- scale images will be acquired from 300 patients with suspicious breast masses. All modes of imaging will be conducted on the same patients and on the same day. The issues related to optimal imaging will be emphasized by careful control of experimental conditions. Among other things, this will include regular image calibration and monitoring of progesterone levels in the patients to account for variations in blood flow during the normal menstrual cycle as well as for variations due to hormonal therapy. In Aim 2 we introduce new approaches for deriving quantitative features from Doppler and gray-scale images. The emphasis will be on quantifying features that physicians use in evaluating images. In Aim 3 these features will be supplemented with qualitative assessments to develop a multifactorial model for a comprehensive diagnosis of breast lesions. Novel approaches based on advanced nonlinear methods, including neural networks, will be developed to formulate a decision tree for cancer diagnosis. Each strategy will be evaluated by ROC analysis. This approach will provide objective measures to demonstrate the importance of sonographic and Doppler features and how they can be used optimally to accurately diagnose breast cancers. Taken together, this program takes advantage of unique breadth of clinical experience and basic sciences at the University of Pennsylvania. An integrated approach in which one mode represents tissue property and the other its function will result in a systematic and comprehensive diagnosis of breast cancers.