Deep cross-modality (MR-CT) educed distillation learning for cone beam CT lung tumor segmentation

Abstract

Purpose: Despite the widespread availability of in-treatment room cone beam computed tomography (CBCT) imaging, due to the lack of reliable segmentation methods, CBCT is only used for gross set up corrections in lung radiotherapies. Accurate and reliable auto-segmentation tools could potentiate volumetric response assessment and geometry-guided adaptive radiation therapies. Therefore, we developed a new deep learning CBCT lung tumor segmentation method.

Methods: The key idea of our approach called cross-modality educed distillation (CMEDL) is to use magnetic resonance imaging (MRI) to guide a CBCT segmentation network training to extract more informative features during training. We accomplish this by training an end-to-end network comprised of unpaired domain adaptation (UDA) and cross-domain segmentation distillation networks (SDNs) using unpaired CBCT and MRI datasets. UDA approach uses CBCT and MRI that are not aligned and may arise from different sets of patients. The UDA network synthesizes pseudo MRI from CBCT images. The SDN consists of teacher MRI and student CBCT segmentation networks. Feature distillation regularizes the student network to extract CBCT features that match the statistical distribution of MRI features extracted by the teacher network and obtain better differentiation of tumor from background. The UDA network was implemented with a cycleGAN improved with contextual losses separately on Unet and dense fully convolutional segmentation networks (DenseFCN). Performance comparisons were done against CBCT only using 2D and 3D networks. We also compared against an alternative framework that used UDA with MR segmentation network, whereby segmentation was done on the synthesized pseudo MRI representation. All networks were trained with 216 weekly CBCTs and 82 T2-weighted turbo spin echo MRI acquired from different patient cohorts. Validation was done on 20 weekly CBCTs from patients not used in training. Independent testing was done on 38 weekly CBCTs from patients not used in training or validation. Segmentation accuracy was measured using surface Dice similarity coefficient (SDSC) and Hausdroff distance at 95th percentile (HD95) metrics.

Results: The CMEDL approach significantly improved (p < 0.001) the accuracy of both Unet (SDSC of 0.83 ± 0.08; HD95 of 7.69 ± 7.86 mm) and DenseFCN (SDSC of 0.75 ± 0.13; HD95 of 11.42 ± 9.87 mm) over CBCT only 2DUnet (SDSC of 0.69 ± 0.11; HD95 of 21.70 ± 16.34 mm), 3D Unet (SDSC of 0.72 ± 0.20; HD95 15.01 ± 12.98 mm), and DenseFCN (SDSC of 0.66 ± 0.15; HD95 of 22.15 ± 17.19 mm) networks. The alternate framework using UDA with the MRI network was also more accurate than the CBCT only methods but less accurate the CMEDL approach.

Conclusions: Our results demonstrate feasibility of the introduced CMEDL approach to produce reasonably accurate lung cancer segmentation from CBCT images. Further validation on larger datasets is necessary for clinical translation.

Keywords: CBCT segmentation; MR informed segmentation; adversarial deep learning; distillation learning; lung tumors.