Augmented Breast Tumor Classification by Perfusion Analysis

• Main Author: Lin, Bruce Yu-Sun
• Other Authors: Battista, John
• Format: Others
• Language: en
• Published: LSU 2010
• Subjects: Computer Science
• Online Access: http://etd.lsu.edu/docs/available/etd-08252010-154811/

Magnetic resonance and computed tomography imaging aid in the diagnosis and analysis of pathologic conditions. Blood flow, or perfusion, through a region of tissue can be computed from a time series of contrast-enhanced images. Perfusion is an important set of physiological parameters that reflect angiogenesis. In cancer, heightened angiogenesis is a key process in the growth and spread of tumorous masses. An automatic classification technique using recovered perfusion may prove to be a highly accurate diagnostic tool. Such a classification system would supplement existing histopathological tests, and help physicians to choose the most optimal treatment protocol. Perfusion is obtained through deconvolution of signal intensity series and a pharmacokinetic model. However, many computational problems complicate the accurate-consistent recovery of perfusion. The high time-resolution acquisition of images decreases signal-to-noise, producing poor deconvolution solutions. The delivery of contrast agent as a function of time must also be determined or sampled before deconvolution can proceed. Some regions of the body, such as the brain, provide a nearby artery to serve as this arterial input function. Poor estimates can lead to an over or under estimation of perfusion. Breast tissue is an example of one tissue region where a clearly defined artery is not present. This proposes a new method of using recovered perfusion and spatial information in an automated classifier. This classifier grades suspected lesions as benign or malignant. This method can be integrated into a computer-aided diagnostic system to enhance the value of medical imagery.