Ultrasound-guided Intervention for Prostate Cancer Detection

Loading...
Thumbnail Image
Date
2014-11-19
Authors
Khojaste Galesh-Khale, Amir
Keyword
Abstract
Abstract
Prostate cancer is the second most common cancer diagnosed in North American men. It can be managed with a survival rate of over 90% if diagnosed early. Although diagnosis of prostate cancer is important, however not all the diagnosed cancers are life-threatening. Hence, accurate prognosis of prostate cancer is crucial to avoid over-treatment of patients with indolent disease. Recently, a novel ultrasound tissue typing technique was proposed that utilizes ultrasound RF data acquired from a stationary location of the tissue over time. The goal in this thesis is to present the feasibility of characterizing aggressive prostate cancer using ultrasound RF time series. We pursue this goal by analyzing data from two ex vivo and one in vivo studies involving prostatectomy patients.In almost all of the ultrasound-based interventions, calibration of the ultrasound probes is of crucial importance to determine the location of image plane in a global coordinate system. Calibration is routinely performed by imaging a geometrically known object. The calibration problem is then solved by relating the feature locations in this object in the ultrasound images and their true locations. Recently a calibration technique was proposed that eliminated the need for fabricating complex phantoms. This technique is based on imaging a flat plate that is immersed into a water tank. In this thesis, we also evaluate the robustness of this calibration by performing extensive experimental evaluations. In the ex vivo studies, in a cross-validation framework, areas under the accumulated receiver operating characteristic curve (AUC) of 0.8 and 0.85 were obtained from 15 and 6 patients, respectively. The results are confirmed by performing a gold-standard pathology to ultrasound registration. We also provide likelihood maps showing the probability and extent of higher grade cancer in the entire cancerous area of each patient. In the in vivo study on prostatectomy cases, an AUC of 0.88 is achieved on characterizing higher grade prostate cancer. The experimental evaluations of single wall approach yielded reproducibility accuracy of 3.14 ± 1.5 mm. Having a well-defined hand motion pattern for ultrasound probe movements and digitizing the wall plane first can improve the robustness of this calibration framework.
External DOI