A Comparative Study of Deep Learning Dose Prediction Models for Cervical Cancer Volumetric Modulated Arc Therapy

Purpose: Deep learning (DL) is widely used in dose prediction for radiation oncology, multiple DL techniques comparison is often lacking in the literature. To compare the performance of 4 state-of-the-art DL models in predicting the voxel-level dose distribution for cervical cancer volumetric modulated arc therapy (VMAT). Methods and Materials: A total of 261 patients' plans for cervical cancer were retrieved in this retrospective study. A three-channel feature map, consisting of a planning target volume (PTV) mask, organs at risk (OARs) mask, and CT image was fed into the three-dimensional (3D) U-Net and its 3 variants models. The data set was randomly divided into 80% as training-validation and 20% as testing set, respectively. The model performance was evaluated on the 52 testing patients by comparing the generated dose distributions against the clinical approved ground truth (GT) using mean absolute error (MAE), dose map difference (GT-predicted), clinical dosimetric indices, and dice similarity coefficients (DSC). Results: The 3D U-Net and its 3 variants DL models exhibited promising performance with a maximum MAE within the PTV 0.83% ± 0.67% in the UNETR model. The maximum MAE among the OARs is the left femoral head, which reached 6.95% ± 6.55%. For the body, the maximum MAE was observed in UNETR, which is 1.19 ± 0.86%, and the minimum MAE was 0.94 ± 0.85% for 3D U-Net. The average error of the Dmean difference for different OARs is within 2.5 Gy. The average error of V40 difference for the bladder and rectum is about 5%. The mean DSC under different isodose volumes was above 90%. Conclusions: DL models can predict the voxel-level dose distribution accurately for cervical cancer VMAT treatment plans. All models demonstrated almost analogous performance for voxel-wise dose prediction maps. Considering all voxels within the body, 3D U-Net showed the best performance. The state-of-the-art DL models are of great significance for further clinical applications of cervical cancer VMAT.

Could tumour volume and major and minor axis based on CTA statistical anatomy improve the pre-operative T-stage in oesophageal cancer?

OBJECTIVES

To statistically study the 3D shape of oesophageal cancer (EC) and its spatial relationships based on computed tomography angiography (CTA) 3D reconstruction, to determine its relationship with T-stages, and to create an optimal T-stage diagnosis protocol based on CTA calculation.

METHODS

Pre-operative CTA images of 155 patients with EC were retrospectively collected and divided into four groups: T1-T4. We used Amira software to segment and 3D reconstruct the EC, oesophagus, aorta, pericardium and peripheral lymph nodes and measured their surface area, volume, major axis, minor axis, longitudinal length, roughness and relationship to the aorta of the EC. One-way ANOVA, independent sample t-test, ROC, etc., were performed and critical values between different T-stages were calculated. We also invited two radiologists to evaluate the measurements.

RESULTS

There were no significant differences in EC longitudinal length, roughness score and relationship with the aorta between the different T-stages of EC. There were significant differences in EC surface area, EC volume and mean major and minor axis among the different T-stages. The volumes of the T1-T4 tumours were 12,934.36 ± 7739.25, 23,095.27 ± 14,975.67, 37,577.98 ± 36,085.64 and 58,579.25 ± 41,073.96 mm³ separately (p < 0.05), and the T1-T4 volume cut-off values were 11,712.00, 19,809.00 and 44,103.50 mm³ separately. For comparison with radiologists, the AUC value of our measurements was 0.704, which was higher than the radiologists of AUC = 0.630.

CONCLUSIONS

EC volume, major and minor axis can be used as important factors for surgeons in the T-stage diagnosis of EC, which helps to improve prognosis and treatment decisions after CTA.