Research Project
9737939
Project Summary/Abstract Prostate cancer (PCa) is the most-common cancer of men in the United States. The goal of the project is to develop a reliable, imaging tool for characterizing prostate tissue. This tool will improve detection, grading, treatment, and monitoring of PCa. It can be implemented in existing clinical scanners for effective needle-biopsy guidance and for planning and targeting focal therapy. Such a tool would significantly reduce the amount of unnecessary biopsies of noncancerous tissue and the current high rate of false negative diagnoses. It also would allow monitoring diagnosed PCa during watchful waiting or after non-surgical therapies. We propose to investigate the feasibility of fusing acoustic radiation-force impulse (ARFI) imaging and quantitative ultrasound (QUS) for typing prostate tissue. Studies to date have demonstrated encouraging performance of each ultrasonic modality when used alone. However, the ability of the combined modalities has not yet been investigated. The proposed project seeks to extract features from a retrospective data set containing radio-frequency (RF) ultrasound data acquired while performing in vivo prostate ARFI imaging. Images generated from both technologies are intrinsically perfectly registered. We will employ our well-established QUS-processing procedures to extract QUS-parameter maps from the RF-data and will develop and test new algorithms to extend the feature set from our established QUS-processing to incorporate ARFI image features. We will use standard and new features from QUS and ARFI for training linear and non-linear classifiers (e.g. support vector machines) to identify and image PCa. We will compare the performance of classifiers trained with QUS alone, ARFI alone, and MRI alone. Furthermore, we will assess performance improvement if classifiers are trained based on QUS- ARFI, QUS-MRI, ARFI-MRI, and QUS-ARFI-MRI parameter combinations and will determine the best classifier to detect prostate cancerous tissue using cross-validation and receiver-operating-characteristic (ROC) statistics. If this project is successful, subsequent projects will focus on integrating the newly developed tools into clinical scanners.
