(Sunday, 7/29/2018) 3:00 PM - 6:00 PM
Room: Exhibit Hall
Purpose: To demonstrate knowledge-based 3D dose distribution in patients with nasopharyngeal carcinoma treated with tomotherapy for automatic planning.
Methods: A retrospective study was conducted in 41 cases of nasopharyngeal carcinoma treated with helical Tomotherapy. The prescription doses to pGTV was 67.5Gy in 30 fractions. The minimum distance between each voxel point and each OAR was considered as the feature of this point, and the dose corresponding to the voxel was taken as the target value. If the voxel is in a target area or OARs, the minimum distance corresponding to the corresponding structure would be negative. The voxel size of this study was 1.95mmx1.95mmx3.00mm. The Eigenvalues and target values of 30 patients were randomly selected to establish a point-dose prediction model using machine learning methods. The prediction model was used to get the dose prediction for the remaining 11 patients. Median filter was done to make the dose distribution of each layer in the predicted dose by using the template of 7 × 7. The clinical dose distribution was compared to validate the prediction model and the prediction accuracy was evaluated. All the data extraction, model establishment and model validation in the study were done using MATLAB.
Results: A total of 14.97 million data were collected from 30 patients to build a predictive model. The predicted dose distribution was close to and partially better than the clinical dose distribution. The dose difference of left, right eyes and spinal cord in 11 patient plans was significant (p<0.05). No noticeable changes were found in targets and other OARs between the predicted and planned dose.
Conclusion: The results achieved using this method are generally satisfactory. Preliminary studies have shown that it is possible to clinically provide a good objective function for automatic optimization with helical tomotherapy.
Funding Support, Disclosures, and Conflict of Interest: Chinese Natural Science Fund(No. 61601012) Natural Key R&D Program of China(No.2017YFC0112100)
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
