Automated measurement of tumour extent in patients with colorectal liver metastases from X-ray computed tomography

Abstract

Tumour burden is a significant prognostic factor in patients with colorectal liver metastases, and can be staged according to percentage hepatic replacement (PHR), but accurate measurements of PHR are difficult to obtain. We have used an automated computerized technique to isolate tumour and measure PHR from computed tomography images of 22 patients. Tumour identification was achieved by establishing local histogram-based thresholds at liver-tumour boundaries displaying clear bimodal histogram distributions. There was a significant association between measured PHR and independent visual estimates of PHR stage defined in quartiles (p less than 0.001). When the measured value of PHR differed from the PHR stage, the visual estimate was always higher. The accuracy of PHR measurement was assessed on stimulated liver computed tomography images. It seems likely that visual estimates of PHR from computed tomography images tend to overestimate tumour extent. The automated technique is probably more accurate and will be useful for assessing and monitoring patients.