|
Abstract:
The long-term goal of this research is to improve breast cancer diagnosis, risk assessment, response assessment, and patient care via the use of large-scale, multi-modality computerized image analysis. The central hypothesis of this research is that large-scale image analysis for breast cancer research will yield improved accuracy and reliability when optimized over multiple features and large multi-modality databases. We designed and executed a pilot study to utilize large scale parallel Grid computing to harness the nationwide cluster infrastructure for optimization of medical image analysis parameters. Additionally, we investigated the use of cutting edge dataanalysis/ mining techniques as applied to Ultrasound, FFDM, and DCE-MRI Breast Image Feature Space Analysis for CADx, specifically, dimension reduction and data representation techniques (t-SNE and Laplacian Eigenmaps) for high dimensional data spaces. These methods allow for an alternative to traditional feature selection methods. Using the256-CPU high-throughput cluster computing capabilities, performance metrics and intensive statistical cross-validation (0.632+ bootstrap and ROC analysis for AUC performance) were performed to gain understanding of the new techniques potential versus previous Breast CADx methodologies. Results indicate the ability to rival or exceed previous state-of-the-art CADx performance.
| Limitations: |
 APPROVED FOR PUBLIC RELEASE |
| Description: |
Annual rept. 1 Oct 2008-30 Sep 2009 |
| Pages: |
51 |
| Report Date: |
Oct 2009 |
| Contract Number: |
W81XWH-08-1-0731 |
| Report Number: |
A139155 |
|
|
|
|
|
Keywords relating to this report:
Email This Abstract
