An expert system for electronic cleansing of contrast in CT colonography images

Clinical interpretation of polyp using Computed Tomography Colonography (CTC) expects a clean colon without any residuals. Electronic cleansing is an image post processing technique in CTC which removes the left out oral contrast from the colon. There are few post processing consequences like soft tissue degradation, incomplete cleansing and pseudo soft tissue structures which results in false positives. The objective of the study was to remove the contrast from colon without losing the tissue details. This paper proposes a novel method to solve the first two problems using a multi-step algorithm. It uses a new edge model based method which involves a) colon segmentation, b) priori information of selective Hounsfield Unit (HU) of different colonic materials at specific tube voltages, c) subtracting the contrast, d) decomposing the materials within colon based on selective HU values, e) removing boundary between air and contrast, f) and applying filters to clean minute particles of improperly tagged endoluminal fluids which appear as noise. The main finding of the study was, soft tissue structures were absolutely preserved after removing contrast. The technique was applied on fecal tagged patient's dataset (40 patients, ∼500 images/patient) where the contrast was not completely removed from the colon. The results were validated by radiologists for any artefacts due to image processing.