Computer-Aided Diagnostic Scheme for Detection and Characterization of Lung Nodules on Low-Dose CT in NLST Utilizing Advanced Modeling Techniques

Principal Investigator

Name
Kavita Garg

Degrees
MD

Institution
University of Colorado Health Science Center

Position Title
Professor of Radiology

Email
kavitagarg@uchsc.edu

About this CDAS Project

Study
NLST (Learn more about this study)

Project ID
2003-90015

Initial CDAS Request Approval
Jul 8, 2003

Title
Computer-Aided Diagnostic Scheme for Detection and Characterization of Lung Nodules on Low-Dose CT in NLST Utilizing Advanced Modeling Techniques

Summary
This study proposes to evaluate a CAD system for the detection of pulmonary nodules. Results of the CAD analyses will be compared to those of radiologists. In addition, the proposal will try to apply advanced image analysis techniques, including artificial neural networks (ANN), to identify nodules likely to grow. Finally, the effectiveness of an ANN model in combination to CAD will be evaluated.

Aims

1. To independently assess the performance of a CAD system in the detection of suspicious pulmonary nodules by comparing detection rates between CAD and radiologists. Hypothesis: There will be a significant difference in suspicious nodule detection rates between the CAD system and the radiologists. 2. To apply advanced image analysis techniques and four separate statistical learning methods to develop and validate a set of methods/algorithms for identifying nodules that are likely to grow (likely malignant) as measured on subsequent scans. Hypothesis: The methods/algorithms that are developed can be used to identify those nodules that are likely to grow, improving the sensitivity and specificity of screening with CT.