Automatic liver tumor segmentation from CT scans with knowledge-based constraints

Automatic hepatic tumor segmentation is a crucial step for diagnosis and surgery planning. This paper presents a new fully automatic technique to segment the tumors in liver structure with no interaction from user. Contrast enhancement is applied to the slices of segmented liver, then adding each image to itself to have a white image with some pepper noise and tumors as dark gray spots. After applying Gaussian smoothing, Isodata is used to threshold the tumor in the slice. In order to eliminate erroneous segmentation, discriminative rule based on diagnostic knowledge on liver cancer shape is applied along with a 3-D consistency check is performed based on three-dimensional information that a lesion mass cannot appear in a single slice independently. Tests are performed on 9 abdominal datasets and promising result shows that sensitivity and specificity for automatic liver tumor segmentation are 87% and 99% respectively.