Technical limitation of semi-automated volumetric analysis using CT in patients with Malignant Pleural Mesothelioma

Abstract

Introduction: Volumetric tumour measurement of malignant pleural mesothelioma (MPM) using contrast-enhanced CT could be used to assess MPM tumour stage as an alternative to RECIST response reporting. Aims and Objectives: We attempted to deploy a recently reported semi-automated method of volumetric MPM tumour assessment developed using Magnetic Resonance Imaging (MRI) (Armato S, Lung Cancer In press) on CT images acquired in the same patients. The MRI method requires tumour regions to be grown from seed points placed within a pre-defined volume of parietal pleural tumour, based on differential tumour enhancement after contrast. Methods: 23 patients with MPM were studied. 15 regions of interest (ROI) were placed on visible areas of pleural tumour. Similar ROIs were placed on intercostal muscle, pleural fluid, diaphragm and liver and Hounsfield Units (HU) generated for all ROIs. Pleural volume contours were then drawn every 8–10 axial CT slices and propagated throughout the volume (Myrian, Intrasense, France). Tumour regions were then grown within these volumes based on the measured range HU of pleural tumour. Results: The median (range) HU of pleural tumour, intercostal muscle and diaphragm overlapped: 49 (15–127), 21 (4–55) and 38 (7–65), respectively. Inter-slice interpolation artefacts were common and tumour region growing often resulted in ‘bleeding’ of the tumour regions into adjacent normal structures. Conclusions: It was not possible to accurately deploy a semi-automated volumetric analysis method for MPM developed using MRI on contrast-enhanced CT images. Reduced definition between tumour and adjacent structures was a major contributor, manifested as overlapping HU values.