Authors:
B. Ganeshan1, K. A. Miles2, R. C. D. Young1, C. R. Chatwin1; 1Falmer/UK, 2Brighton/UK
Purpose
Can computer-analysis of liver texture during CT provide biomarkers for patients with colorectal cancer?
Colorectal cancer patients entering surveillance programs do not represent a uniform population of equal risk of recurrence. Identification of predictive factors that are linked to outcomes may allow modification of surveillance strategies for sub-groups of patients and the need for research in this area has been highlighted by an expert panel of ASCO [1]. A number of laboratory-derived predictive factors have been identified for patients with colorectal cancer [2,3] but relatively little attention has been given to biomarkers derived from diagnostic images. Recent developments in physiologic and molecular imaging techniques have provided new opportunities for the use of imaging as a biomarker [4].
Relevant physiologic imaging techniques that have been demonstrated to provide prognostic information for patients with colorectal cancer include Doppler ultrasound of the liver and quantitative analysis of hepatic contrast enhancement on CT [5,6]. It is likely that these modalities detect the perturbation of hepatic hemodynamics that has been shown to be associated with micrometastases [7,8]. However, such techniques are additional to routine imaging protocols and have not been widely adopted in surveillance programs.
CT densitometry for measurement of hepatic x-ray attenuation, as used previously in the diagnosis of Wilson’s disease, iron overload and steatosis [9-11], is an analysis methodology that could be applied to the CT images acquired routinely. However, measurements of hepatic attenuation are not sensitive to the local spatial variations in image brightness that produce features such as coarseness and regularity. The human visual system has difficulties in discriminating such textural information but these more subtle features can be quantified using a computer-based numeric manipulation of digital images known as texture analysis [12]. Most texture analysis studies of CT images have been based on segmentation and classification of visible focal lesions into benign and malignant and recognition of different organs using wavelet techniques and artificial neural network based decision algorithms [13-19]. However it is more complex and challenging to distinguish diagnostic patients groups from visually normal areas of the liver of patients following resection of colorectal cancer. Previous studies have shown potential for hepatic CT texture to differ between normal livers and apparently normal areas of tissue within livers bearing tumours and may reflect hepatic vascularity [20,21].
The purpose of this study was to assess the potential for computer-analysis of liver texture during CT to provide biomarkers for patients with colorectal cancer and to identify possible biological correlates for relevant texture features.