Computer-aided volumetric assessment of malignant pleural mesothelioma on CT using a random walk-based method

Abstract

Objective: The aim of this study is to assess the performance of a computer-aided semi-automated algorithm we have adapted for the purpose of segmenting malignant pleural mesothelioma (MPM) on CT. Methods: Forty-five CT scans were collected from 15 patients (M:F = 10:5, mean age 62.8 years) in a multi-centre clinical drug trial. A computer-aided random walk-based algorithm was applied to segment the tumour; the results were then compared to radiologist-drawn contours and correlated with measurements made using the MPM-adapted Response Evaluation Criteria in Solid Tumour (modified RECIST). Results: A mean accuracy (Sørensen–Dice index) of 0.825 (95% CI [0.758, 0.892]) was achieved. Compared to a median measurement time of 68.1 min (range [40.2, 102.4]) for manual delineation, the median running time of our algorithm was 23.1 min (range [10.9, 37.0]). A linear correlation (Pearson’s correlation coefficient: 0.6392, p< 0.05) was established between the changes in modified RECIST and computed tumour volume. Conclusion: Volumetric tumour segmentation offers a potential solution to the challenges in quantifying MPM. Computer-assisted methods such as the one presented in this study facilitate this in an accurate and time-efficient manner and provide additional morphological information about the tumour’s evolution over time.