Development of a virtual source model for Monte Carlo-based independent dose calculation for varian linac.

Monte Carlo (MC) independent dose calculations are often based on phase-space files (PSF), as they can accurately represent particle characteristics. PSF generally are large and create a bottleneck in computation time. In addition, the number of independent particles is limited by the PSF, preventing further reduction of statistical uncertainty. The purpose of this study is to develop and validate a virtual source model (VSM) to address these limitations. Particles from existing PSF for the Varian TrueBeam medical linear accelerator 6X, 6XFFF, 10X, and 10XFFF beam configurations were tallied, analyzed, and used to generate a dual-source photon VSM that includes electron contamination. The particle density distribution, kinetic energy spectrum, particle direction, and the correlations between characteristics were computed. The VSM models for each beam configuration were validated with water phantom measurements as well as clinical test cases against the original PSF. The new VSM requires 67 MB of disk space for each beam configuration, compared to 50 GB for the PSF from which they are based and effectively remove the bottleneck set by the PSF. At 3% MC uncertainty, the VSM approach reduces the calculation time by a factor of 14 on our server. MC doses obtained using the VSM approach were compared against PSF-generated doses in clinical test cases and measurements in a water phantom using a gamma index analysis. For all tests, the VSMs were in excellent agreement with PSF doses and measurements (>90% passing voxels between doses and measurements). Results of this study indicate the successful derivation and implementation of a VSM model for Varian Linac that significantly saves computation time without sacrificing accuracy for independent dose calculation.

Investigation of absolute dose calibration accuracy for TomoTherapy using real water.

A systematic bias in TomoTherapy output calibration was reported by the Imaging and Radiation Oncology Core Houston (IROC-H) after analyzing intensity-modulated radiation therapy (IMRT) credentialing results from hundreds of TomoTherapy units. Multiple theories were developed to explain this observation. One theory was that the use of a solid water "cheese" phantom instead of real water in the calibration measurement was the culprit. A phantom filled with distilled water was built to investigate whether our TomoTherapy was miscalibrated due to the use of a solid water phantom. A miscalibration of -1.47% was detected on our TomoTherapy unit. It is found that despite following the vendor's updated recommendation on computed tomography (CT) number to density calibration, the cheese phantom was still mapped to a density of 1.028 g/cm3 , rather than the 1.01 g/cm3 value reported in literature. When the density of the cheese phantom was modified to 1.01 g/cm3 in the treatment planning system, the measurement also indicated that our TomoTherapy machine was miscalibrated by -1.52%, agreeing with the real water phantom findings. Our single-institution finding showed that the cheese phantom density assignment can introduce greater than 1% errors in the TomoTherapy absolute dose calibration. It is recommended that the absolute dose calibration for TomoTherapy be performed either in real water or in the cheese phantom with the density in TPS overridden as 1.01 g/cm3 .

Contour subregion error detection methodology using deep learning auto-segmentation.

BACKGROUND: Inaccurate manual organ delineation is one of the high-risk failure modes in radiation treatment. Numerous automated contour quality assurance (QA) systems have been developed to assess contour acceptability; however, manual inspection of flagged cases is a time-consuming and challenging process, and can lead to users overlooking the exact error location. PURPOSE: Our aim is to develop and validate a contour QA system that can effectively detect and visualize subregional contour errors, both qualitatively and quantitatively. METHODS/MATERIALS: A novel contour subregion error detection (CSED) system was developed using subregional surface distance discrepancies between manual and deep learning auto-segmentation (DLAS) contours. A validation study was conducted using a head and neck public dataset containing 339 cases and evaluated according to knowledge-based pass criteria derived from a clinical training dataset of 60 cases. A blind qualitative evaluation was conducted, comparing the results from the CSED system with manual labels. Subsequently, the CSED-flagged cases were re-examined by a radiation oncologist. RESULTS: The CSED system could visualize the diverse types of subregional contour errors qualitatively and quantitatively. In the validation dataset, the CSED system resulted in true positive rates (TPR) of 0.814, 0.800, and 0.771; false positive rates (FPR) of 0.310, 0.267, and 0.298; and accuracies of 0.735, 0.759, and 0.730, for brainstem and left and right parotid contours, respectively. The CSED-assisted manual review caught 13 brainstem, 19 left parotid, and 21 right parotid contour errors missed by conventional human review. The TPR/FPR/accuracy of the CSED-assisted manual review improved to 0.836/0.253/0.784, 0.831/0.171/0.830, and 0.808/0.193/0.807 for each structure, respectively. Further, the time savings achieved through CSED-assisted review improved by 75%, with the time for review taking 24.81 ± 12.84, 26.75 ± 10.41, and 28.71 ± 13.72 s for each structure, respectively. CONCLUSIONS: The CSED system enables qualitative and quantitative detection, localization, and visualization of manual segmentation subregional errors utilizing DLAS contours as references. The use of this system has been shown to help reduce the risk of high-risk failure modes resulting from inaccurate organ segmentation.

Contouring quality assurance methodology based on multiple geometric features against deep learning auto-segmentation.

BACKGROUND: Contouring error is one of the top failure modes in radiation treatment. Multiple efforts have been made to develop tools to automatically detect segmentation errors. Deep learning-based auto-segmentation (DLAS) has been used as a baseline for flagging manual segmentation errors, but those efforts are limited to using only one or two contour comparison metrics. PURPOSE: The purpose of this research is to develop an improved contouring quality assurance system to identify and flag manual contouring errors. METHODS AND MATERIALS: DLAS contours were used as a reference to compare with manually segmented contours. A total of 27 geometric agreement metrics were determined from the comparisons between the two segmentation approaches. Feature selection was performed to optimize the training of a machine learning classification model to identify potential contouring errors. A public dataset with 339 cases was used to train and test the classifier. Four independent classifiers were trained using five-fold cross validation, and the predictions from each classifier were ensembled using soft voting. The trained model was validated on a held-out testing dataset. An additional independent clinical dataset with 60 cases was used to test the generalizability of the model. Model predictions were reviewed by an expert to confirm or reject the findings. RESULTS: The proposed machine learning multiple features (ML-MF) approach outperformed traditional nonmachine-learning-based approaches that are based on only one or two geometric agreement metrics. The machine learning model achieved recall (precision) values of 0.842 (0.899), 0.762 (0.762), 0.727 (0.842), and 0.773 (0.773) for Brainstem, Parotid_L, Parotid_R, and mandible contours, respectively compared to 0.526 (0.909), 0.619 (0.765), 0.682 (0.882), 0.773 (0.568) for an approach based solely on Dice similarity coefficient values. In the external validation dataset, 66.7, 93.3, 94.1, and 58.8% of flagged cases were confirmed to have contouring errors by an expert for Brainstem, Parotid_L, Parotid_R, and mandible contours, respectively. CONCLUSIONS: The proposed ML-MF approach, which includes multiple geometric agreement metrics to flag manual contouring errors, demonstrated superior performance in comparison to traditional methods. This method is easy to implement in clinical practice and can help to reduce the significant time and labor costs associated with manual segmentation and review.

Estimating breast tissue-specific DNA methylation age using next-generation sequencing data.

BACKGROUND: DNA methylation (DNAm) age has been widely accepted as an epigenetic biomarker for biological aging. Emerging evidence suggests that DNAm age can be tissue-specific and female breast tissue ages faster than other parts of the body. The Horvath clock, which estimates DNAm age across multiple tissues, has been shown to be poorly calibrated in breast issue. We aim to develop a model to estimate breast tissue-specific DNAm age. METHODS: Genome-wide DNA methylation sequencing data were generated for 459 normal, 107 tumor, and 45 paired adjacent-normal breast tissue samples. We determined a novel set of 286 breast tissue-specific clock CpGs using penalized linear regression and developed a model to estimate breast tissue-specific DNAm age. The model was applied to estimate breast tissue-specific DNAm age in different breast tissue types and in tumors with distinct clinical characteristics to investigate cancer-related aging effects. RESULTS: Our estimated breast tissue-specific DNAm age was highly correlated with chronological age (r = 0.88; p = 2.9 × 10-31) in normal breast tissue. Breast tumor tissue samples exhibited a positive epigenetic age acceleration, where DNAm age was on average 7 years older than respective chronological age (p = 1.8 × 10-8). In age-matched analyses, tumor breast tissue appeared 12 and 13 years older in DNAm age than adjacent-normal and normal breast tissue (p = 4.0 × 10-6 and 1.0 × 10-6, respectively). Both HER2+ and hormone-receptor positive subtypes demonstrated significant acceleration in DNAm ages (p = 0.04 and 3.8 × 10-6, respectively), while no apparent DNAm age acceleration was observed for triple-negative breast tumors. We observed a non-linear pattern of epigenetic age acceleration with breast tumor grade. In addition, early-staged tumors showed a positive epigenetic age acceleration (p = 0.003) while late-staged tumors exhibited a non-significant negative epigenetic age acceleration (p = 0.10). CONCLUSIONS: The intended applications for this model are wide-spread and have been shown to provide biologically meaningful results for cancer-related aging effects in breast tumor tissue. Future studies are warranted to explore whether breast tissue-specific epigenetic age acceleration is predictive of breast cancer development, treatment response, and survival as well as the clinical utility of whether this model can be extended to blood samples.