Deep convolutional neural networks for the automated segmentation of malignant pleural mesothelioma on computed tomography scans

Abstract

Tumor volume has been a topic of interest in the staging, prognostic evaluation, and treatment response assessment of malignant pleural mesothelioma (MPM). Deep convolutional neural networks (CNNs) were trained separately for the left and right hemithoraces on the task of differentiating between pleural thickening and normal thoracic tissue on computed tomography (CT) scans. A total of 4259 and 6192 axial sections containing segmented tumor were used to train the left-hemithorax CNN and the right-hemithorax CNN, respectively. Two distinct test sets of 131 sections from the CT scans of 43 patients were used to evaluate segmentation performance by calculating the Dice similarity coefficient (DSC) between deep CNN-generated tumor segmentations and reference tumor segmentations provided by a total of eight observers. Median DSC values ranged from 0.662 to 0.800 over the two test sets when comparing deep CNN-generated segmentations with observer reference segmentations. The deep CNN-based method achieved significantly higher DSC values for all three observers on the test set that allowed direct comparisons with a previously published automated segmentation method of MPM tumor on CT scans (p < 0.0005). A deep CNN was implemented for the automated segmentation of MPM tumor on CT scans, showing superior performance to a previously published method.

Deep learning-based segmentation of malignant pleural mesothelioma tumor on computed tomography scans: application to scans demonstrating pleural effusion

“…38,42 The median DSC value for radiologist interobserver comparisons was found to range from 0.65 to 0.81 across the two test sets of our previous study on the deep learning-based segmentation of mesothelioma. 23 Across both test sets of the present study, the overlap of deep learning-predicted tumor segmentations with radiologist tumor contours remained on par with radiologist interobserver overlap achieved on the two test sets of our previous study. Despite these encouraging results, the present method remains to be clinically validated through an observer study, whereby the segmentation performance of the method would be assessed by radiologists experienced in the measurement and assessment of mesothelioma tumor.…”

“…Our previous study on the deep learning-based segmentation of mesothelioma ("2018 Method") showed a significantly improved segmentation performance when compared with a prior stepwise mesothelioma segmentation method; however, this deep learning-based method did not adequately exclude pleural effusion from tumor contours. 22,23 Compared with the 2018 Method, the present deep CNN-based mesothelioma segmentation method showed significantly greater overlap with radiologist-provided reference tumor contours on a test set of 94 CT sections (i.e., the "Tumor and Effusion Test Set") that all exhibited both tumor and pleural effusion. The agreement between deep CNN-predicted tumor contours and observerprovided reference tumor contours on this test set, as evaluated using the AHD metric, was found to be significantly higher for the present method when compared with the 2018 Method.…”

“…20,21 Previous work on the automated segmentation of mesothelioma tumor on CT scans includes a stepwise method that aimed to isolate pleural thickening in patients through the segmentation of the lungs and the ribcage 22 and our study that used a deep CNN-based segmentation method based on the U-Net architecture for the automated segmentation of mesothelioma on CT scans ("2018 Method"). 23 In our prior deep learning-based study, the 2018 Method showed superior performance to the nondeep learning-based method on an independent test set of CT scans of mesothelioma patients with radiologist-provided tumor contours. It was found, however, that the 2018 Method showed increased disagreement with observer-provided tumor contours in the presence of pleural effusion.…”

“…Results of the present method were compared with (1) a set of manual tumor outlines constructed by three thoracic radiologists on CT sections of two test sets not included in the training set and (2) the output of the 2018 Method on the same test sets. 23…”

“…This test set was created through the combination of the two test sets of our prior published study on the deep learning-based segmentation of mesothelioma (i.e., the "2018 Method"). 23 To simplify the analysis of the present study, a single set of reference tumor contours for Test Set 2 was constructed using only the reference contours provided by the single observer with the highest average interobserver Dice similarity coefficient (DSC) value on each of the test sets of our prior study (observers A and 4 of our prior study). Exclusion criteria for Test Set 2 included prominent calcifications, chest-wall invasion, surgical intervention, a mean DSC value across all observers ≤0.5, and lack of observer agreement on the laterality of disease.…”