Robust automated detection, segmentation, and classification of hepatic tumors from CT data

The manuscript presents the automated detection and segmentation of hepatic tumors from abdominal CT images with variable acquisition parameters. After obtaining an initial segmentation of the liver, optimized graph cuts segment the liver tumor candidates using shape and enhancement constraints. One hundred and fifty-seven features are computed for the tumor candidates and support vector machines are used to select features and separate true and false detections. Training and testing are performed using leave-one-patientout on 14 patients with a total of 79 tumors. After selection, the feature space is reduced to eight. The resulting sensitivity for tumor detection was 100% at 2.3 false positives/case. For the true tumors, 74.1% overlap and 1.6mm average surface distance were recorded between the ground truth and the results of the automated method. Results from test data demonstrate the method's robustness to analyze livers from difficult clinical cases to allow the diagnoses and temporal monitoring of patients with hepatic cancer.