Computer Aided Liver Lesion Detection Algorithm

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): The use of computer aided diagnosis (CAD) is growing in medical imaging. Currently CAD is used in mammography and to a lesser extent in chest imaging. Imaging of the liver is performed for the work-up and evaluation of hepatic neoplasms and other malignancies which may spread to the liver is growing and may be amenable to CAD. Computed Tomography (CT) on the liver is the most frequently used diagnostic modality to screen the liver in patients with malignancy. With higher resolution, thinner collimation CT scanners an increasing number of images are generated and need to be interpreted by Radiologists. Concurrent with the use of thinner CT sections, is the improvement in hepatic resection, which is currently considered the standard therapy for colorectal metastases isolated to the liver. Precise size, number and localization of liver lesions are necessary to aid the surgeon in preoperative planning for hepatic resection. In patients with other malignancies the choice of definitive therapy relies in part upon the exclusion of distant disease, including to the liver. In this proposed research work, we will develop tools for CAD of the liver including advanced computerized methods to automatically delineate liver contours using gradient vector flow (GVF) snake with a modified edge-map for the contour initiation, computer aided detection of metastases in the extracted livers by a local density minimum algorithm, and segmentation of liver metastases with a novel constrained region-growing algorithm. This proposal will capitalize on a database of 145 patients with metastatic colorectal cancer, who were previously enrolled in an RO1 funded grant "Utility of Whole-Body 18F-Fluorodeoxyglucose Positron Emission Tomography (PET) In the Preoperative Assessment of Patients with Hepatic Colorectal Metastases". This database provides an outstanding opportunity to correlate preoperative CT scanning with surgical and pathological findings, and will therefore serve as an ideal test bed for the development and verification of algorithms for the automatic detection and segmentation of liver metastases. The development of a successful automated liver lesion detection algorithm will greatly improve the efficiency of detectability. It has the potential to aid the radiologist in detecting more lesions and thus improve staging for patients. Clearly this is the first step in a process of development and if successful will lead to further efforts in automated detection and characterization.