SBIR Phase I: Clinical Assessment of a Computerized Workstation for Characterization of Breast Lesions using Magnetic Resonance Imaging (MRI)

这个信息和通信技术的小企业创新研究第一阶段项目旨在开发一个工作站，用于在MRI上描述乳腺病变的特征，以帮助乳腺癌诊断。这项SBIR建议的具体目标是：1)获取和扩展乳腺病变MRI数据库，2)开发和记录一个工作站，以遵循FDA的提交要求。仅在美国，每8名妇女中就有1人将在一生中被诊断为乳腺癌。正确的诊断是改善患者预后的关键组成部分。特别是，早期诊断可以导致较小的手术切除，并可能消除对辅助放疗或化疗的需要。在过去的十年中，MRI用于乳腺癌筛查和诊断的使用已经迅速扩大，因为它比传统的方法(如乳房X光检查)具有明显更高的灵敏度。由于核磁共振扫描的数据由300多张高分辨率图像组成，这种迁移对放射科医生来说是一个越来越困难和耗时的挑战。只有当考虑到恶性病变在单一图像上可能只呈现为一个小斑点(~0.5 mm)时，MRI数据密集型特性带来的解释挑战才会变得复杂。这项工作的目的是从丰富的数据集中提取相关信息，并提高图像解释的效率、准确性和一致性。所提出的分析和解释技术包括自动病变分割、自动图像信息提取和智能信息融合。

This Small Business Innovation Research Phase I project for Information and Communication Technology aims to develop a workstation for the characterization of breast lesions on MRI to assist in breast cancer diagnosis. The specific aims of this SBIR proposal are: 1) to acquire and expand a breast lesion MRI database and, 2) to develop and document a workstation to follow FDA submission requirements.In the US alone, 1 out of 8 women will be diagnosed with breast cancer during their lifetime. Proper diagnosis is a critical component of improving patient outcomes. In particular, earlier diagnosis can result in smaller surgical resections and may obviate the need for adjuvant radio- or chemotherapy. Over the past decade, the use of MRI for breast cancer screening and diagnosis has been rapidly expanding due to its notably higher sensitivity over traditional methods such as mammography. As the data from MRI scans consist of over 300 high-resolution images, this migration presents an increasingly difficult and time-consuming challenge for radiologists. The interpretive challenge presented by the data-intense nature of MRI is only compounded when considering that a malignant lesion may present as only a small speck (~0.5 mm) on a single image. The effort aims to extract relevant information from the rich dataset and improve the efficiency, accuracy, and consistency of image interpretation. The proposed analysis and interpretation techniques include automatic lesion segmentation, automatic image information extraction, and intelligent information fusion.