Curtailing patient-specific IMRT QA procedures from 2D dose error distribution

A deep learning-based prediction model for gamma evaluation in patient-specific quality assurance

PURPOSE

Patient-specific quality assurance (QA) measurement is conducted to confirm the accuracy of dose delivery. However, measurement is time-consuming and places a heavy workload on the medical physicists and radiological technologists. In this study, we proposed a prediction model for gamma evaluation, based on deep learning. We applied the model to a QA measurement dataset of prostate cancer cases to evaluate its practicality.

METHODS

Sixty pretreatment verification plans from prostate cancer patients treated using intensity modulated radiation therapy were collected. Fifteen-layer convolutional neural networks (CNN) were developed to learn the sagittal planar dose distributions from a RT-3000 QA phantom (R-TECH.INC., Tokyo, Japan). The percentage gamma passing rate (GPR) was measured using GAFCHROMIC EBT3 film (Ashland Specialty Ingredients, Covington, USA). The input training data also included the volume of the PTV (planning target volume), rectum, and overlapping region, measured in cm3 , and the monitor unit values for each field. The network produced predicted GPR values at four criteria: 2%(global)/2 mm, 3%(global)/2 mm, 2%(global)/3 mm, and 3%(global)/3 mm. Adam, an algorithm for first-order gradient-based optimization of stochastic objective functions, was used for learning and for optimizing the CNN-based model. Fivefold cross-validation was applied to validate the performance of the proposed method. Forty cases were used for training and validation set in fivefold cross-validation, and the remaining 20 cases were used for the test set. The predicted and measured GPR values were compared.

RESULTS

A linear relationship was found between the measured and predicted values, for each of the four criteria. Spearman rank correlation coefficients in validation set between measured and predicted GPR values at four criteria were 0.73 at 2%/2 mm, 0.72 at 3%/2 mm, 0.74 at 2%/3 mm, and 0.65 at 3%/3 mm, respectively (P < 0.01). The Spearman rank correlation coefficients in the test set were 0.62 (P < 0.01) at 2%/2 mm, 0.56 (P < 0.01) at 3%/2 mm, 0.51 (P = 0.02) at 2%/3 mm, and 0.32 (P = 0.16) at 3%/3 mm. These results demonstrated a strong or moderate correlation between the predicted and measured values.

CONCLUSIONS

We developed a CNN-based prediction model for patient-specific QA of dose distribution in prostate treatment. Our results suggest that deep learning may provide a useful prediction model for gamma evaluation of patient-specific QA in prostate treatment planning.

A clinically relevant IMRT QA workflow: Design and validation

PURPOSE

The purpose of this study was to determine clinically relevant pass/question/fail criteria for gamma analysis of intensity-modulated radiation therapy quality assurance (IMRT QA) plans, identify which plans should be further analyzed with dose-volume histogram (DVH) metrics, and create a workflow for performing that DVH-based analysis.

METHODS

A total of 11 plans, 5 prostate and 6 head/neck, were selected to represent known good plans based on their high-passing rate using conventional IMRT QA criteria. These were modified by moving the programmed MLC positions to underdose the target or overdose important structures by varying amounts. Commercially available hardware/software was used to measure and analyze all plans (76 total) using 4%/3 mm, 3%/3 mm, 3%/2 mm, and 2%/2 mm gamma criteria. Two receiver operator characteristic (ROC) curves per criterion were created to assess effective passing rates. One ROC curve was to find a higher threshold that determined a clear pass and the second to find a lower threshold to determine a clear failure. Plans between these two thresholds need DVH-based analysis to assess the clinical consequence of the dose difference. The modified plans were analyzed in the planning system and reconstructed in commercially available DVH-based analysis software to access the accuracy and usefulness of the software.

RESULTS

Analysis of the ROC curves showed optimal pass and fail thresholds for plan error detection per criterion to achieve clinically relevant sensitivity and specificity. Based on measurement uncertainty and pass/fail ranges, 3%/2 mm gamma criteria with a pass threshold of 95% and a fail threshold of 90% were most optimal. DVH analysis showed good agreement with all reconstructed plans except where the changes to the MLC patterns caused the periphery of the target to be underdosed. For questionable plans, comparing the organ-specific DVHs to the physician-provided planning constraints proved to be an efficient and effective workflow since plans for which the target dose was slightly high or where organs at risk were underdosed could be released for the treatment without consulting the physician for a clinical decision.

CONCLUSION

This work indicates the potential for appreciable improvement in error detection for IMRT QA. Using effective pass/fail thresholds to determine plans that need DVH-based analysis minimizes the need for excessive, time-consuming, analysis, and making use of the dosimetric constraints of the plan minimizes the burden on physicians. Overall, DVH-based analysis is a powerful tool that can provide substantial insight over the traditional approach that does not provide structure-specific data.