Prostate cancer is one of the most frequently detected cancers and a leading cause of death for men in the United States. Fortunately, chances... Show moreProstate cancer is one of the most frequently detected cancers and a leading cause of death for men in the United States. Fortunately, chances of successful treatment significantly improve when prostate cancer is diagnosed at an early stage. An accurate and reliable imaging technique that can localize tumors is very valuable for example in guiding needle biopsy, therapy, as well as surgery. Current imaging techniques such as ultrasound provide unsatisfactory result for tumor localization. Therefore, Multispectral Magnetic Resonance Imaging (MRI) emerged as an alternative method of imaging for prostate cancer. An important part of multispectral MRI is Dynamic contrast enhanced MRI (DCE-MRI). Traditionally, a compartmental model is used with DCE-MRI to estimate parametric images. However these images are usually very noisy. Therefore, we propose to use directly the series of DCE-MRI for tumor localization instead of the pharmacokinetic parameters which are derived from DCE MR images through a compartmental model. As a second contribution, we have developed new features in addition to traditional parameters based on different characteristic of tumor and normal pixels. These new features are standard deviation and the energy of the first half of the spectrum divided by the energy of the second half of the spectrum based on an analysis of TACs for tumor and normal pixels. We use quantitative and qualitative evaluation methods to demonstrate that employing DCE-MRI directly as features and our proposed features result in an improvement in terms of tumor localization performance. M.S. in Electrical Engineering, July 2011 Show less
