COMPUTED 3D SURFACE AND VOLUME ESTIMATION IN CT AND MRI

CT 和 MRI 中的计算 3D 表面和体积估计

• 批准号: 3197546
• 负责人: Charles Raymond Meyer
• 金额: $ 28.33万
• 依托单位: UNIVERSITY OF MICHIGAN AT ANN ARBOR
• 依托单位国家: 美国
• 财政年份: 1991
• 资助国家: 美国
• 起止时间: 1991-08-06 至 1994-07-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/3197546
• 关键词: computed axial tomography; computer data analysis; evaluation /testing; image processing; liver imaging /visualization /scanning; magnetic resonance imaging; mathematical model; model design /development; phantom model

The goal of this proposal is to bring the benefits of routine, accurate 3D volume estimation and display, currently implemented exclusively on high contrast boundaries, to low contrast, soft tissue lesions and organs. We will develop, refine and apply new, robust, automatic surface finding algorithms which require minimal operator intervention to define the spatial extent of low contrast, soft tissue lesions and organs in x- ray computed tomography (CT) and magnetic resonance imaging (MRI). Methods proposed involve usage of segmentation, and robust gradient techniques for general lesion/organ surface detection, and geometric model-guided blackboard techniques for specific organ surface detection. Algorithm development will include adaptation of a new 3D segmentation algorithm originally developed for computer vision under funding from the National Science Foundation for segmentation of 3D medical data sets. Since this segmentation algorithm is the first to perform functional segmentation on 3D data, it potentially represents a revolutionary breakthrough to the problem of disparate edge generation arising from the use of 2D techniques on a slice-by-slice basis. In addition, we will apply new robust edge segment detection and spline linking algorithms to gradient magnitude images computed from the original data set using 3D gradient operators. Bivariate tensor splines will yield 3D lesion surfaces. Since single approach techniques tend to lack the required information to achieve nearly 100% detection accuracy, we will also implement a knowledge-based blackboard system to guide surface detection of the liver using an a priori, geometric shape model. While artificial intelligence techniques have been applied to medical imaging data in the past, to our knowledge this specific geometric model-guided blackboard approach to lever is new, and has a significant chance of achieving a sufficiently high accuracy so as to nearly eliminate operator assistance or editing. We are able to apply this technique to the liver (organ systems in general) because they have a normal shape and location. Evaluation of the algorithms will make major use of clinical data sets from patients undergoing radiation and chemo therapy for focal liver carcinoma, osteogenic and soft tissue sarcomas, as well as physical and computer generated phantoms. The major clinical beneficiaries of substantial success in these algorithm developments are 3D radiation therapy treatment planning and routine and inexpensive quantitative assessment of tumor response to therapy.

本方案的目标是带来常规、准确的好处