RESUMEN
Objective:Topredict the three-dimensional dose distribution of regions of interest (ROI) with brachytherapy for cervical cancer based on U-Net fully convolutional network, and evaluate the accuracy of prediction model.Methods:First, 100 cases of cervical cancer intracavity combined with interstitial implantation were selected as the entire research data set, and divided into the training set ( n=72), validation set ( n=8), and test set ( n=20). Then the U-Net was used to construct two models based on whether the uterine tandem and the implantation needles were included as the distinguishing factors. Finally, dose distribution of 20 cases in the test set were predicted using the trained model, and comparative analysis was performed. The performance of the model was jointly evaluated by , and the mean absolute deviation (MAD). Results:Compared with the model without the uterine tandem and the implantation needles, the of the rectum was increased by (16.83±1.82) cGy ( P<0.05), and the or of the other ROI were not different significantly (all P>0.05). The MAD of the high-risk clinical target volume, rectum, sigmoid, small bowel, and bladder was increased by (11.96±3.78) cGy, (11.43±0.54) cGy, (24.08±1.65) cGy, (17.04±7.17) cGy and (9.52±4.35) cGy, respectively (all P<0.05). The MAD of the intermediate-risk clinical target volume was decreased by (120.85±29.78) cGy ( P<0.05). The mean value of MAD for all ROI was decreased by (7.8±53) cGy ( P<0.05), which was closer to the actual plan. Conclusions:U-Net fully convolutional network can be used to predict three-dimensional dose distribution of patients with cervical cancer undergoing brachytherapy. Combining the uterine tube with the implantation needles as the input parameters yields more accurate predictions than a single use of the ROI structure as the input.
RESUMEN
Objective To establish a three-dimensional (3D) dose prediction model,which can predict multiple organs simultaneously in a single model and automatically learn the effect of the geometric anatomical structure on dose distribution.Methods Clinical radiotherapy plans of patients diagnosed with the same type of tumors were collected and retrospectively analyzed.For every plan,each organs at risk (OAR) voxel was regarded as the study sample and its deposited dose was considered as the dosimetric feature.A regularized multi-task learning method than could learn the relationship among different tasks was employed to establish the relationship matrix among tasks and the correlation between geometric structure and dose distribution among organs.In this experiment,the spinal cord,brainstem and bilateral parotids involved in the intensity-modulated radiotherapy (IMRT) plan of 15 nasopharyngeal cancer patients were utilized to establish the multi-organ prediction model.The relative percentage error between the predicted dose of voxel and the clinical planning dose was calculated to assess the feasibility of the model.Results Ten cases receiving IMRT plans were utilized as the training data,and the remaining five cases were used as the test data.The test results demonstrated a higher prediction accuracy and less data demand.And the average voxel dose errors among the spinal cord,brainstem and the left and right parotids were (2.01±0.02)％,(2.65± 0.02) ％,(2.45± 0.02) ％ and (2.55± 0.02) ％,respectively.Conclusion The proposed model can accurately predict the dose of multiple organs in a single model and avoid the establishment of multiple single-organ prediction models,laying a solid foundation for patient-specific plan quality control and knowledge-based treatment planning.