Automated Segmentation of Sacral Chordomas and Surrounding Muscles Using Deep Learning Ensemble.

Researchers

Journal

Modalities

Models

Abstract

The manual segmentation of organ structures in radiation oncology treatment planning is a time-consuming and highly skilled task, particularly when treating rare tumors like sacral chordomas. This study evaluates the performance of automated deep learning (DL) models in accurately segmenting the gross tumor volume (GTV) and surrounding muscle structures of sacral chordomas.An expert radiation oncologist contoured five muscle structures (Gluteus Maximus, Gluteus Medius, Gluteus Minimus, Paraspinal, Piriformis) and sacral chordoma GTV on CT images from 48 patients. We trained six DL auto-segmentation models based on 3D U-Net and Residual 3D U-Net architectures. We then implemented an average and an optimally weighted average ensemble to improve prediction performance. We evaluated algorithms with the average and standard deviation of the Volumetric Dice Similarity Coefficient (VDSC), Surface Dice Similarity Coefficient (SDSC) with 2 and 3 mm thresholds, and Average Symmetric Surface Distance (ASSD). One independent expert radiation oncologist assessed the clinical viability of the DL contours and determined the necessary amount of editing before they could be used in clinical practice.Quantitatively, the ensembles performed the best across all structures. The optimal ensemble (VDSC, ASSD) was (85.5±6.4,2.6±0.8; GTV), (94.4±1.5,1.0±0.4; Gluteus Maximus), (92.6±0.9,0.9±0.1; Gluteus Medius), (85.0±2.7,1.1±0.3; Gluteus Minimus), (92.1±1.5,0.8±0.2; Paraspinal), and (78.3±5.7,1.5±0.6; Piriformis). The qualitative evaluation suggested that the best model could reduce the total muscle and tumor delineation time to a 19-minute average.Our methodology produces expert-level muscle and sacral chordoma tumor segmentation using DL and ensemble modeling. It can substantially augment the streamlining and accuracy of treatment planning and represents a critical step towards automated delineation of the Clinical Target Volume (CTV) in sarcoma and other disease sites.Copyright © 2023 Elsevier Inc. All rights reserved.