A generalization performance study on the boosting radiotherapy dose calculation engine based on super-resolution

Abstract

Purpose: During the radiation treatment planning process, one of the time-consuming procedures is the final high-resolution dose calculation, which obstacles the wide application of the emerging online adaptive radiotherapy techniques (OLART). There is an urgent desire for highly accurate and efficient dose calculation methods. This study aims to develop a dose super resolution-based deep learning model for fast and accurate dose prediction in clinical practice.

Method: A Multi-stage Dose Super-Resolution Network (MDSR Net) architecture with sparse masks module and multi-stage progressive dose distribution restoration method were developed to predict high-resolution dose distribution using low-resolution data. A total of 340 VMAT plans from different disease sites were used, among which 240 randomly selected nasopharyngeal, lung, and cervix cases were used for model training, and the remaining 60 cases from the same sites for model benchmark testing, and additional 40 cases from the unseen site (breast and rectum) was used for model generalizability evaluation. The clinical calculated dose with a grid size of 2 mm was used as baseline dose distribution. The input included the dose distribution with 4 mm grid size and CT images. The model performance was compared with HD U-Net and cubic interpolation methods using Dose-volume histograms (DVH) metrics and global gamma analysis with 1%/1 mm and 10% low dose threshold. The correlation between the prediction error and the dose, dose gradient, and CT values was also evaluated.

Results: The prediction errors of MDSR were 0.06-0.84% of D_mean indices, and the gamma passing rate was 83.1-91.0% on the benchmark testing dataset, and 0.02-1.03% and 71.3-90.3% for the generalization dataset respectively. The model performance was significantly higher than the HD U-Net and interpolation methods (p < 0.05). The mean errors of the MDSR model decreased (monotonously by 0.03-0.004%) with dose and increased (by 0.01-0.73%) with the dose gradient. There was no correlation between prediction errors and the CT values.

Conclusion: The proposed MDSR model achieved good agreement with the baseline high-resolution dose distribution, with small prediction errors for DVH indices and high gamma passing rate for both seen and unseen sites, indicating a robust and generalizable dose prediction model. The model can provide fast and accurate high-resolution dose distribution for clinical dose calculation, particularly for the routine practice of OLART.

Keywords: adaptive radiotherapy; deep learning; dose calculation; generalization performance; super resolution.

Figure 1

MDSR-Net model architecture. (a) the diagram of MDSR-Net. The model consists of the patch subnetwork with an encoder-decoder architecture and original resolution subnetwork with a single-scale feature pipeline architecture. The green arrow represents the cross-stage feature fusion operation and the purple arrow of the supervised boost attention mechanism (BAM). (b) the architecture of an encoder-decoder subnetwork. (c) the architecture of SMR module. (d) the diagram of a BAM module. (e) the diagram of SMG module (e.1) and SMC module (e.2).

Figure 2

Comparison of the performance of MDSR, HD U-Net, and interpolation models on the benchmark testing dataset. * refers to p < 0.05, ** refers to p < 0.01, *** refers to p < 0.001, **** refers to p < 0.0001 comparing to the MDSR model. (a) errors with standard deviation (SD) of D_mean index for cervix cases. (b) D_95% and D_2% indices for cervix cases. (c) D_mean index for lung cases. (d) D_95% and D_2% indices for lung cases. (e) D_mean index for NPC cases. (f) D_95% and D_2% indices for NPC cases.

Figure 3

The average gamma passing rate of MDSR, HD U-Net and interpolation models on the benchmark and generation testing datasets. * refers to p < 0.05, ** refers to p < 0.01, *** refers to p < 0.001, **** refers to p < 0.0001 comparing to the MDSR model. (a) the average gamma passing rate with standard deviation (SD) on the benchmark testing dataset. (b) the average passing rate with SD on the generation testing dataset.

Figure 4

The prediction results for MDSR, HD U-Net, and interpolation models on the simulation application dataset. * refers to p < 0.05, ** refers to p < 0.01, *** refers to p < 0.001, **** refers to p < 0.0001 comparing to the MDSR model. (a) errors with standard deviation (SD) of D_mean index for rectum cases. (b) errors with SD of D_95% and D_2% for rectum cases. (c) errors with SD of D_mean for breast cases. (d) errors with SD of D_95% and D_2% for breast cases.

Figure 5

The results of errors distribution pattern analysis. (a) the correlation between the relative dose and mean errors with standard error (SE) (in 10% dose interval). (b) the correlation between relative dose gradient and mean errors with SE in 10% dose gradient interval. (c) the correlation between relative CT values and mean errors with SE in 10% CT values interval.

Figure 6

The results of gamma analysis of ablation experiment for evaluating the contribution from the introduced modules. * refers to p < 0.05, ** refers to p < 0.01, *** refers to p < 0.001, **** refers to p < 0.0001 comparing to the MDSR model. (a) the average gamma passing rate with standard deviation (SD) on the benchmark testing dataset. (b) the average passing rate with SD on the generation testing dataset.