Comparison of semi-automated and manual methods to measure the volume of liver tumours on MDCT images

To retrospectively compare semi-automated and manual volume measurements of malignant liver tumours and inter- and intra-observer variability using commercially available software. This study was performed on 60 consecutive patients with untreated liver metastases (30) and HCCs (30), i.e. 92 lesions (49 metastases, 43 HCCs) using hepatic MDCT. Lesion volumes were manually measured independently by two radiologists and semi-automatically by the same two radiologists and a technician. Those measurements were repeated on 20 patients (10 metastases and 10 HCCs) a week later. An independent operator timed all the measurements. Using the Spearman correlation coefficient and Bland-Altman plots, statistical analyses were performed. Liver lesion volumes obtained with semi-automated and manual methods were well correlated (Spearman, r = 0.98 and 0.91). Their agreement was high for intra-observer measurements with the semi-automated method (Spearman, r = 0.91 and 0.94). The agreement was lower for inter-observer measurements with both methods (Spearman, r = 0.87 for semi-automated and 0.91 for manual). The semi-automated method significantly reduced the post-processing duration (23s +/- 19s vs. 33s +/- 11s, p value < 0.0001). In our study, semi-automated volume analysis of malignant liver tumours correlated well with the manual method. Furthermore, the semi-automated volume analysis was significantly quicker.