Purpose We sought to develop machine learning models to detect multileaf collimator (MLC) modeling errors with the use of radiomic features of fluence maps measured in patient‐specific quality assurance (QA) for intensity‐modulated radiation therapy (IMRT) with an electric portal imaging device (EPID). Methods Fluence maps measured with EPID for 38 beams from 19 clinical IMRT plans were assessed. Plans with various degrees of error in MLC modeling parameters [i.e., MLC transmission factor (TF) and dosimetric leaf gap (DLG)] and plans with an MLC positional error for comparison were created. For a total of 152 error plans for each type of error, we calculated fluence difference maps for each beam by subtracting the calculated maps from the measured maps. A total of 837 radiomic features were extracted from each fluence difference map, and we determined the number of features used for the training dataset in the machine learning models by using random forest regression. Machine learning models using the five typical algorithms [decision tree, k‐nearest neighbor (kNN), support vector machine (SVM), logistic regression, and random forest] for binary classification between the error‐free plan and the plan with the corresponding error for each type of error were developed. We used part of the total dataset to perform fourfold cross‐validation to tune the models, and we used the remaining test dataset to evaluate the performance of the developed models. A gamma analysis was also performed between the measured and calculated fluence maps with the criteria of 3%/2 and 2%/2 mm for all of the types of error. Results The radiomic features and its optimal number were similar for the models for the TF and the DLG error detection, which was different from the MLC positional error. The highest sensitivity was obtained as 0.913 for the TF error with SVM and logistic regression, 0.978 for the DLG error with kNN and SVM, and 1.000 for the MLC positional error with kNN, SVM, and random forest. The highest specificity was obtained as 1.000 for the TF error with a decision tree, SVM, and logistic regression, 1.000 for the DLG error with a decision tree, logistic regression, and random forest, and 0.909 for the MLC positional error with a decision tree and logistic regression. The gamma analysis showed the poorest performance in which sensitivities were 0.737 for the TF error and the DLG error and 0.882 for the MLC positional error for 3%/2 mm. The addition of another type of error to fluence maps significantly reduced the sensitivity for the TF and the DLG error, whereas no effect was observed for the MLC positional error detection. Conclusions Compared to the conventional gamma analysis, the radiomics‐based machine learning models showed higher sensitivity and specificity in detecting a single type of the MLC modeling error and the MLC positional error. Although the developed models need further improvement for detecting multiple types of error, radiomics‐based IMRT QA was shown to be a promising approach for detecting the MLC modeli...

No abstract

In conventional stereotactic radiosurgery (SRS), treatment of multiple brain metastases using multiple isocenters is time-consuming resulting in long dose delivery times for patients. A single-isocenter technique has been developed which enables the simultaneous irradiation of multiple targets at one isocenter. This technique requires accurate positioning of the patient to ensure optimal dose coverage. We evaluated the effect of six degrees of freedom (6DoF) setup errors in patient setups on SRS dose distributions for multiple brain metastases using a single-isocenter technique. We used simulated spherical gross tumor volumes (GTVs) with diameters ranging from 1.0 to 3.0 cm. The distance from the isocenter to the target's center was varied from 0 to 15 cm. We created dose distributions so that each target was entirely covered by 100% of the prescribed dose. The target's position vectors were rotated from 0°-2.0°and translated from 0-1.0 mm with respect to the three axes in space. The reduction in dose coverage for the targets for each setup error was calculated and compared with zero setup error. The calculated margins for the GTV necessary to satisfy the tolerance values for loss of GTV coverage of 3% to 10% were defined as coverage-based margins. In addition, the maximum isocenter to target distance for different 6DoF setup errors was calculated to satisfy the tolerance values. The dose coverage reduction and coverage-based margins increased as the target diameter decreased, and the distance and 6DoF setup error increased. An increase in setup error when a single-isocenter technique is used may increase the risk of missing the tumor; this risk increases with increasing distance from the isocenter and decreasing tumor size.

To achieve an accurate stopping power ratio (SPR) prediction in particle therapy treatment planning, we previously proposed a simple conversion to the SPR from dual-energy (DE) computed tomography (CT) data via electron density and effective atomic number (Z eff) calibration (DEEDZ-SPR). This study was conducted to carry out an initial implementation of the DEEDZ-SPR conversion method with a clinical treatment planning system (TPS; VQA, Hitachi Ltd., Tokyo) for proton beam therapy. Consequently, this paper presents a proton therapy plan for an anthropomorphic phantom to evaluate the stability of the dose calculations obtained by the DEEDZ-SPR conversion against the variation of the calibration phantom size. Dual-energy x-ray CT images were acquired using a dual-source CT (DSCT) scanner. A single-energy CT (SECT) scan using the same DSCT scanner was also performed to compare the DEEDZ-SPR conversion with the SECT-based SPR (SECT-SPR) conversion. The scanner-specific parameters necessary for the SPR calibration were obtained from the CT images of tissue substitutes in a calibration phantom. Two calibration phantoms with different sizes (a 33 cm diameter phantom and an 18 cm diameter phantom) were used for the SPR calibrations to investigate the beam-hardening effect on dosimetric uncertainties. Each set of calibrated SPR data was applied to the proton therapy plan designed using the VQA TPS with a pencil beam algorithm for the anthropomorphic phantom. The treatment plans with the SECT-SPR conversion exhibited discrepancies between the dose distributions and the dose-volume histograms (DVHs) of the 33 cm and 18 cm phantom calibrations. In contrast, the corresponding dose distributions and the DVHs obtained using the DEEDZ-SPR conversion method coincided almost perfectly with each other. The DEEDZ-SPR conversion appears to be a promising method for providing proton dose plans that are stable against the size variations of the calibration phantom and the patient.

Background & AimsA new real‐time tracking radiotherapy (RTRT) system, the SyncTraX FX4 (Shimadzu, Kyoto, Japan), consisting of four X‐ray tubes and four ceiling‐mounted flat panel detectors (FPDs) combined with a linear accelerator, was installed at Uonuma Kikan Hospital (Niigata, Japan) for the first time worldwide. In addition to RTRT, the SyncTraX FX4 system enables bony structure‐based patient verification. Here we provide the first report of this system's clinical commissioning for intracranial stereotactic radiotherapy (SRT).Materials & MethodsA total of five tests were performed for the commissioning: evaluations of (1) the system's image quality; (2) the imaging and treatment coordinate coincidence; and (3) the localization accuracy of cone‐beam computed tomography (CBCT) and SyncTraX FX4; (4) the measurement of air kerma; (5) an end‐to‐end test.Results & DiscussionThe tests revealed the following. (1) All image quality evaluation items satisfied each acceptable criterion in all FPDs. (2) The maximum offsets among the centers were ≤0.40 mm in all combinations of the FPD and X‐ray tubes (preset). (3) The isocenter localization discrepancies between CBCT and preset #3 in the SyncTraX FX4 system were 0.29 ± 0.084 mm for anterior‐posterior, −0.19 ± 0.13 mm for superior‐inferior, 0.076 ± 0.11 mm for left‐right, −0.11 ± 0.066° for rotation, −0.14 ± 0.064° for pitch, and 0.072±0.058° for roll direction. the Pearson's product‐moment correlation coefficient between the two systems was >0.98 in all directions. (4) The mean air kerma value for preset #3 was 0.11 ± 0.0002 mGy in predefined settings (80 kV, 200 mA, 50 msec). (5) For 16 combinations of gantry and couch angles, median offset value in all presets was 0.31 mm (range 0.14–0.57 mm).ConclusionOur results demonstrate a competent performance of the SyncTraX FX4 system in terms of the localization accuracy for intracranial SRT.