Improving Detection of Axillary Lymph Nodes by Computer-Aided Kinetic Feature Identification in Positron Emission Tomography
Abstract
The goal of this project is to improve detection of metastatic axillary breast cancer through sophisticated physiological modeling and statistical signal processing techniques. The major focus of Year 2 was to develop statistical hypothesis test criteria for the computer-aided detection of kinetic features in metastases, which was built on modeling, extracting and exploiting physiological features pursued since the last reporting period. Three types of generalized likelihood ratio tests (GLRT) were derived under different assumptions on pixel spatial correlation. We compared these three GLRTs using three different methods of TAC evaluation and computer generated phantom data. The lesions in the phantom data were fully invisible. The results show that the best of the three achieved 97%, 79% and 88% for specificity, sensitivity and accuracy, respectively. Further work on validation and evaluation of physiological feature modeling, kinetic parameter estimation and objective rank-reduction of dynamic PET sinogram for fast image reconstruction has also been continued through this year. The results of the initial clinical dataset studied are promising, but too small to make a conclusive decision. The assessment of all findings in theory and simulations in Year 1 and Year 2 will be continued in the Year 3 when more clinical data will be available.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 2001
- Accession Number
- ADA404943
Entities
People
- Xiaoli Yu
Organizations
- University of Southern California