Computer Detection of Nodules of the Lung

肺结节的计算机检测

基本信息

批准号：
6785732
负责人：
ANTHONY P REEVES
金额：
$ 17.49万
依托单位：
CORNELL UNIVERSITY
依托单位国家：
美国
项目类别：
财政年份：
2004
资助国家：
美国
起止时间：
2004-09-10 至 2005-08-31
项目状态：
已结题

来源：
https://reporter.nih.gov/project-details/6785732
关键词：
bioimaging /biomedical imaging cancer information system cancer risk clinical research computed axial tomography computer assisted diagnosis computer assisted medical decision making computer program /software computer system design /evaluation early diagnosis human data image processing lung imaging /visualization /scanning lung neoplasms neoplasm /cancer diagnosis patient oriented research three dimensional imaging /topography

项目摘要

DESCRIPTION (provided by applicant): The continuing technologic improvements of CT scanners and their increasing use in screening for lung cancer has exacerbated the radiologist's task of identifying pulmonary nodules from whole lung CT scans. This has resulted in a general realization that there will be a need for computer methods to assist in this task. We have developed such a system. In this application, we seek to improve our current system and extend it to additional categories of types of nodules. We plan to use novel model-based verification methods, and to evaluate the performance of our system on our already existing large clinical database. In addition, we plan to evaluate the effect of CT scan parameters on detection performance. The detection method employs a hypothesis generation stage followed by multiple filters to eliminate false positives. It is based on cur previous methods for characterizing pulmonary nodules and involves three-dimensional image analysis techniques. We plan to extend this system to detect all types of nodules including sub-solid nodules. Furthermore, we will explore the use of model based verification methods to further reduce the number of false positives. The complete detection system will be evaluated with a CT image database that has a ground truth established by multiple radiologist reads.

描述（由申请人提供）： CT 扫描仪技术的不断改进及其在肺癌筛查中的日益广泛使用，加剧了放射科医生从全肺 CT 扫描中识别肺结节的任务。这导致人们普遍认识到需要计算机方法来协助完成这项任务。我们开发了这样一个系统。在此应用中，我们寻求改进当前的系统并将其扩展到其他类别的结节类型。我们计划使用新颖的基于模型的验证方法，并在现有的大型临床数据库上评估我们的系统的性能。此外，我们计划评估 CT 扫描参数对检测性能的影响。该检测方法采用假设生成阶段，然后采用多个过滤器来消除误报。它基于当前用于表征肺结节的方法，并涉及三维图像分析技术。我们计划扩展该系统以检测所有类型的结节，包括亚实性结节。此外，我们将探索使用基于模型的验证方法来进一步减少误报的数量。完整的检测系统将使用 CT 图像数据库进行评估，该数据库具有由多个放射科医生读取建立的基本事实。