A feasibility study to predict 3D dose delivery accuracy for IMRT using DenseNet with log files.

OBJECTIVE: This study aims to explore the feasibility of DenseNet in the establishment of a three-dimensional (3D) gamma prediction model of IMRT based on the actual parameters recorded in the log files during delivery. METHODS: A total of 55 IMRT plans (including 367 fields) were randomly selected. The gamma analysis was performed using gamma criteria of 3% /3 mm (Dose Difference/Distance to Agreement), 3% /2 mm, 2% /3 mm, and 2% /2 mm with a 10% dose threshold. In addition, the log files that recorded the gantry angle, monitor units (MU), multi-leaf collimator (MLC), and jaws position during delivery were collected. These log files were then converted to MU-weighted fluence maps as the input of DenseNet, gamma passing rates (GPRs) under four different gamma criteria as the output, and mean square errors (MSEs) as the loss function of this model. RESULTS: Under different gamma criteria, the accuracy of a 3D GPR prediction model decreased with the implementation of stricter gamma criteria. In the test set, the mean absolute error (MAE) of the prediction model under the gamma criteria of 3% /3 mm, 2% /3 mm, 3% /2 mm, and 2% /2 mm was 1.41, 1.44, 3.29, and 3.54, respectively; the root mean square error (RMSE) was 1.91, 1.85, 4.27, and 4.40, respectively; the Sr was 0.487, 0.554, 0.573, and 0.506, respectively. There was a correlation between predicted and measured GPRs (P <  0.01). Additionally, there was no significant difference in the accuracy between the validation set and the test set. The accuracy in the high GPR group was high, and the MAE in the high GPR group was smaller than that in the low GPR group under four different gamma criteria. CONCLUSIONS: In this study, a 3D GPR prediction model of patient-specific QA using DenseNet was established based on log files. As an auxiliary tool for 3D dose verification in IMRT, this model is expected to improve the accuracy and efficiency of dose validation.

Predicting the error magnitude in patient-specific QA during radiotherapy based on ResNet.

BACKGROUND: The error magnitude is closely related to patient-specific dosimetry and plays an important role in evaluating the delivery of the radiotherapy plan in QA. No previous study has investigated the feasibility of deep learning to predict error magnitude. OBJECTIVE: The purpose of this study was to predict the error magnitude of different delivery error types in radiotherapy based on ResNet. METHODS: A total of 34 chest cancer plans (172 fields) of intensity-modulated radiation therapy (IMRT) from Eclipse were selected, of which 30 plans (151 fields) were used for model training and validation, and 4 plans including 21 fields were used for external testing. The collimator misalignment (COLL), monitor unit variation (MU), random multi-leaf collimator shift (MLCR), and systematic MLC shift (MLCS) were introduced. These dose distributions of portal dose predictions for the original plans were defined as the reference dose distribution (RDD), while those for the error-introduced plans were defined as the error-introduced dose distribution (EDD). Different inputs were used in the ResNet for predicting the error magnitude. RESULTS: In the test set, the accuracy of error type prediction based on the dose difference, gamma distribution, and RDD + EDD was 98.36%, 98.91%, and 100%, respectively; the root mean squared error (RMSE) was 1.45-1.54, 0.58-0.90, 0.32-0.36, and 0.15-0.24; the mean absolute error (MAE) was 1.06-1.18, 0.32-0.78, 0.25-0.27, and 0.11-0.18, respectively, for COLL, MU, MLCR and MLCS. CONCLUSIONS: In this study, error magnitude prediction models with dose difference, gamma distribution, and RDD + EDD are established based on ResNet. The accurate prediction of the error magnitude under different error types can provide reference for error analysis in patient-specific QA.

Image-based features in machine learning to identify delivery errors and predict error magnitude for patient-specific IMRT quality assurance.

OBJECTIVE: To identify delivery error type and predict associated error magnitude by image-based features using machine learning (ML). METHODS: In this study, a total of 40 thoracic plans (including 208 beams) were selected, and four error types with different magnitudes were introduced into the original plans, including 1) collimator misalignment (COLL), 2) monitor unit (MU) variation, 3) systematic multileaf collimator misalignment (MLCS), and 4) random MLC misalignment (MLCR). These dose distributions of portal dose predictions for the original plans were defined as the reference dose distributions (RDD), while those for the error-introduced plans were defined as the error-introduced dose distributions (EDD). Both distributions were calculated for all beams with portal dose image prediction (PDIP). Besides, 14 image-based features were extracted from RDD and EDD of portal dose predictions to obtain the feature vectors. In addition, a random forest was adopted for the multiclass classification task, and regression prediction for error magnitude. RESULTS: The top five features extracted with the highest weight included 1) the relative displacement in the x direction, 2) the ratio of the absolute minimum residual error to the maximal RDD value, 3) the product of the maximum and minimum residuals, 4) the ratio of the absolute maximum residual error to the maximal RDD value, and 5) the ratio of the absolute mean residual value to the maximal RDD value. The relative displacement in the x direction had the highest weight. The overall accuracy of the five-class classification model was 99.85% for the validation set and 99.30% for the testing set. This model could be applied to the classification of the error-free plan, COLL, MU, MLCS, and MLCR with an accuracy of 100%, 98.4%, 99.9%, 98.0%, and 98.3%, respectively. MLCR had the worst performance in error magnitude prediction (70.1-96.6%), while others had better performance in error magnitude prediction (higher than 93%). In the error magnitude prediction, the mean absolute error (MAE) between predicted error magnitude and actual error ranged from 0.03 to 0.33, with the root mean squared error (RMSE) varying from 0.17 to 0.56 for the validation set. The MAE and RMSE ranged from 0.03 to 0.50 and 0.44 to 0.59 for the test set, respectively. CONCLUSION: It could be demonstrated in this study that the image-based features extracted from RDD and EDD can be employed to identify different types of delivery errors and accurately predict error magnitude with the assistance of ML techniques. They can be used to associate traditional gamma analysis with clinically based analysis for error classification and magnitude prediction in patient-specific IMRT quality assurance.

Investigation of total skin helical tomotherapy using a 3D-printed total skin bolus.

OBJECTIVE: To investigate the effectiveness of using a 3D-printed total skin bolus in total skin helical tomotherapy for the treatment of mycosis fungoides. MATERIALS AND METHODS: A 65-year-old female patient with a 3-year history of mycosis fungoides underwent treatment using an in-house desktop fused deposition modelling printer to create a total skin bolus made of a 5-mm-thick flexible material, which increased the skin dose through dose building. The patient's scan was segmented into upper and lower sections, with the division line placed 10 cm above the patella. The prescription was to deliver 24 Gy over 24 fractions, given 5 times per week. The plan parameters consisted of a field width of 5 cm, pitch of 0.287 and modulation factor of 3. The complete block was placed 4 cm away from the planned target region to reduce the area of the internal organs at risk, especially the bone marrow. Dose delivery accuracy was verified using point dose verification with a "Cheese" phantom (Gammex RMI, Middleton, WI), 3D plane dose verification with ArcCHECK (Model 1220, Sun Nuclear, Melbourne, FL), and multipoint film dose verification. Megavoltage computed tomography guidance was also utilized to ensure the accuracy of the setup and treatment. RESULTS: A 5-mm-thick 3D-printed suit was used as a bolus to achieve a target volume coverage of 95% of the prescribed dose. The conformity index and homogeneity index of the lower segment were slightly better than those of the upper segment. As the distance from the skin increased, the dose to the bone marrow gradually decreased, and the dose to other organs at risk remained within clinical requirements. The point dose verification deviation was less than 1%, the 3D plane dose verification was greater than 90%, and the multipoint film dose verification was less than 3%, all of which confirmed the accuracy of the delivered dose. The total treatment time was approximately 1.5 h, which included 0.5 h of wearing the 3D-printed suit and 1 h with the beam on. Patients only experienced mild fatigue, nausea or vomiting, low-grade fever, and grade III bone marrow suppression. CONCLUSION: The use of a 3D-printed suit for total skin helical tomotherapy can result in a uniform dose distribution, short treatment time, simple implementation process, good clinical outcomes, and low toxicity. This study presents an alternative treatment approach that can potentially yield improved clinical outcomes for mycosis fungoides.

Clinical application and improvement of a CNN-based autosegmentation model for clinical target volumes in cervical cancer radiotherapy.

OBJECTIVE: Clinical target volume (CTV) autosegmentation for cervical cancer is desirable for radiation therapy. Data heterogeneity and interobserver variability (IOV) limit the clinical adaptability of such methods. The adaptive method is proposed to improve the adaptability of CNN-based autosegmentation of CTV contours in cervical cancer. METHODS: This study included 400 cervical cancer treatment planning cases with CTV delineated by radiation oncologists from three hospitals. The datasets were divided into five subdatasets (80 cases each). The cases in datasets 1, 2, and 3 were delineated by physicians A, B, and C, respectively. The cases in datasets 4 and 5 were delineated by multiple physicians. Dataset 1 was divided into training (50 cases), validation (10 cases), and testing (20 cases) cohorts, and they were used to construct the pretrained model. Datasets 2-5 were regarded as host datasets to evaluate the accuracy of the pretrained model. In the adaptive process, the pretrained model was fine-tuned to measure improvements by gradually adding more training cases selected from the host datasets. The accuracy of the autosegmentation model on each host dataset was evaluated using the corresponding test cases. The Dice similarity coefficient (DSC) and 95% Hausdorff distance (HD_95) were used to evaluate the accuracy. RESULTS: Before and after adaptive improvements, the average DSC values on the host datasets were 0.818 versus 0.882, 0.763 versus 0.810, 0.727 versus 0.772, and 0.679 versus 0.789, which are improvements of 7.82%, 6.16%, 6.19%, and 16.05%, respectively. The average HD_95 values were 11.143 mm versus 6.853 mm, 22.402 mm versus 14.076 mm, 28.145 mm versus 16.437 mm, and 33.034 mm versus 16.441 mm, which are improvements of 37.94%, 37.17%, 41.60%, and 50.23%, respectively. CONCLUSION: The proposed method improved the adaptability of the CNN-based autosegmentation model when applied to host datasets.

Influencing Factors of Total Skin Irradiation With Helical Tomotherapy.

Purpose: To investigate the influencing factors of total skin irradiation (TSI) with helical tomotherapy for guiding the clinical selection of the suitable parameters and optimizing the plan quality and efficiency. Materials and Methods: Six patients with mycosis fungoides (MF) who received TSI were retrospectively selected. They were all dressed with 5 mm thick diving suits during the CT scan and treatment as a bolus to increase the superficial dose through buildup. The dose prescription was 24 Gy in 20 fractions and 5 times per week. During the planned pretreatment, Ring0, Ring1, Ring2, Ring3, and Ring4 of 1 cm thick away from the planning target volume (PTV) at the distances of 0, 1, 2, 3, and 4 cm and other normal tissues (NTs) were generated, respectively. The auxiliary structures were completely blocked during planning; while the field widths were 5 and 2.5 cm, the pitches were 0.287 and 0.215, the modulation factors were 4 and 3, and the other parameters remained consistent. Finally, the dose parameters of PTV and auxiliary structures, as well as the beam on time (BOT) and gantry period, were compared and analyzed. Results: when the auxiliary structures were completely blocked with distance to PTV (dPTV) above 3 cm were used, the mean dose (Dmean), conformity index (CI), and heterogeneity index (HI) of the PTV met the clinical requirements. As the dPTV gradually increased, the BOT decreased while the volume of normal tissue that received excessive radiation increased correspondingly. If the dPTV was less than 3 cm, the clinical requirements were not met. The field widths (FWs), pitches, and modulation factors (MFs) had no effect on PTVmean and the HI. The FW of 2.5 cm was slightly better than 5 cm for the CI. The FW and MF had a significant impact on the BOT, which gradually increased with decreasing FW and increasing MF. Pitch had no effect on the BOT. Conclusion: During planning with TSI patients, dPTV is the key factor that has a significant influence on the plan quality. We found that the plan with the dPTV above 3 cm can meet clinical objectives. The BOT increases as the dPTV increases. The FWs also have an effect on the CI and BOT. Therefore, it is necessary to comprehensively balance these factors to optimize the quality and efficiency of the plan. We also found that different MFs and pitches have no obvious effect on the results.

Clinical study of total bone marrow combined with total lymphatic irradiation pretreatment based on tomotherapy in hematopoietic stem cell transplantation of acute leukemia.

Objective: Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is an effective method for the treatment of refractory and relapsed acute leukemia, and the preconditioning methods before transplantationis one of the important factors affecting the survival of patients. Radiotherapy combined with chemotherapy is the most commonly used preconditioning method before transplantation. This study evaluated the safety and efficacy of total bone marrow combined with total lymphatic irradiation as a preconditioning method before hematopoietic stem cell transplantation. Methods: Seventeen patients with acute leukemia who were admitted to our center from 2016 to 2020 were selected. The median age was 17 years (8-35). The target area for TMLI includes the total bone marrow and total lymphatic space, and the organs at risk include the lens, lungs, kidneys, intestine, heart, and liver. The patients received a total bone marrow and lymphatic irradiation preconditioning regimen, the related acute adverse reactions were graded, and the prognosis of the patients after transplantation was observed. Results: During patient preconditioning, only grade 1-2 toxicity was observed, and grade 3-4 toxicity did not occur. Except for one patient whose platelets were not engrafted, all the other patients were successfully transplanted. The median time of neutrophil implantation was 14 d (9-15 d), and the median time of platelet implantation was 14 d (13-21 d). With a median follow-up of 9 months (2-48), 4 relapses occurred, 3 died, and 10 leukemia patients survived and were disease-free. One-year overall survival was 69.8%, cumulative recurrence was 19.5%, disease-free-survival was 54.2%. Conclusion: The Allo-HSCT pretreatment regimen of total bone marrow combined with total lymphatic irradiation is safe and effective in the treatment of malignant hematological diseases. Total bone marrow combined with total lymphatic irradiation may completely replace total body irradiation, and the clinically observed incidence of acute toxicity is not high.

Deep Learning for Patient-Specific Quality Assurance: Predicting Gamma Passing Rates for IMRT Based on Delivery Fluence Informed by log Files.

Objectives: In this study, we propose a deep learning-based approach to predict Intensity-modulated radiation therapy (IMRT) quality assurance (QA) gamma passing rates using delivery fluence informed by log files. Methods: A total of 112 IMRT plans for chest cancers were planned and measured by portal dosimetry equipped on TrueBeam linac. The convolutional neural network (CNN) based learning model was trained using delivery fluence as inputs and gamma passing rates (GPRs) of 4 different criteria (3%/3 mm, 2%/3 mm, 3%/2 mm, and 2%/2 mm) as outputs. Model performance for both validation and test sets was assessed using mean absolute error (MAE), mean squared error (MSE), root MSE (RMSE), Spearman rank correlation coefficients (Sr), and Determination coefficient (R2) between the measured and predicted GPR values. Results: In the test set, the MAE of the prediction model were 0.402, 0.511, 1.724, and 2.530, the MSE were 0.640, 0.986, 6.654, and 9.508, the RMSE were 0.800, 0.993, 2.580, and 3.083, the Sr were 0.643, 0.684, 0.821, and 0.824 (P < .001) and the R2 were 0.4110, 0.4666, 0.6677, and 0.6769 for 3%/3 mm, 3%/2 mm, 2%/3 mm, and 2%/2 mm, respectively. The MAE and RMSE of the prediction model decreased with stricter gamma criteria while the Sr and R2 between measured and predicted GPR values increased. Conclusions: The CNN prediction model based on delivery fluence informed by log files could accurately predict IMRT QA passing rates for different gamma criteria. It could reduce QA workload and improve efficiency in pretreatment QA. Our results suggest that the CNN prediction model based on delivery fluence informed by log files may be a promising tool for the gamma evaluation of IMRT QA.

Virtual Patient-Specific Quality Assurance of IMRT Using UNet++: Classification, Gamma Passing Rates Prediction, and Dose Difference Prediction.

The dose verification in radiotherapy quality assurance (QA) is time-consuming and places a heavy workload on medical physicists. To provide a clinical tool to perform patient specific QA accurately, the UNet++ is investigated to classify failed or pass fields (the GPR lower than 85% is considered "failed" while the GPR higher than 85% is considered "pass"), predict gamma passing rates (GPR) for different gamma criteria, and predict dose difference from virtual patient-specific quality assurance in radiotherapy. UNet++ was trained and validated with 473 fields and tested with 95 fields. All plans used Portal Dosimetry for dose verification pre-treatment. Planar dose distribution of each field was used as the input for UNet++, with QA classification results, gamma passing rates of different gamma criteria, and dose difference were used as the output. In the test set, the accuracy of the classification model was 95.79%. The mean absolute error (MAE) were 0.82, 0.88, 2.11, 2.52, and the root mean squared error (RMSE) were 1.38, 1.57, 3.33, 3.72 for 3%/3mm, 3%/2 mm, 2%/3 mm, 2%/2 mm, respectively. The trend and position of the predicted dose difference were consistent with the measured dose difference. In conclusion, the Virtual QA based on UNet++ can be used to classify the field passed or not, predict gamma pass rate for different gamma criteria, and predict dose difference. The results show that UNet++ based Virtual QA is promising in quality assurance for radiotherapy.

Technical note: factors affecting dose distribution in the overlap region of two-segment total body irradiation by helical tomotherapy.

OBJECTIVE: To assess the effects of various treatment planning parameters to identify the optimal gap distance for precise two-segment total body irradiation (TBI) using helical tomotherapy (HT) with fixed jaw mode. METHODS AND MATERIALS: Data of a treatment plan for 8 acute leukemia patients (height range: 109-130 cm) were analyzed. All patients underwent total-body computed tomography (CT) with 5-mm slice thickness. A lead wire, placed at 10 cm above the patella, was used to mark the boundary between the two segments. Target volumes and organs at risk were delineated using a Varian Eclipse 10.0 physician's workstation. Different distances between the lead wire and the boundary of the two targets were used. CT images were transferred to the HT workstation to design the treatment plans, by adjusting parameters, including the field width (FW; 2.5 cm, and 5 cm), pitch (0.287 and 0.430), modulation factor (1.8). The plans were superimposed to analyze the dose distributions in the overlap region when varying target gap distances, FWs, pitches to determine the optimal combinations. RESULTS: The pitch did not affect the dose distribution in the overlap region. The dose distribution in the overlap region was mostly homogeneous when the target gap distance was equal to the FW. Increased FW diminished the effect of the target gap distance on the heterogeneous index of the overlap region. CONCLUSIONS: In two-segment TBI treatments by HT with Helix mode, a gap distance equal to the FW may achieve optimal dose distribution in the overlap region.

VirtualDose-IR: a cloud-based software for reporting organ doses in interventional radiology.

A cloud-based software, VirtualDose-IR (Virtual Phantoms Inc., Albany, New York, USA), designed to report organ doses and effective doses for a diverse patient population from interventional radiology (IR) procedures has been developed and tested. This software is based on a comprehensive database of Monte Carlo-generated organ dose built with a set of 21 anatomically realistic patient phantoms. The patient types included in this database are both male and female people with different ages reflecting reference adults, obese people with different BMIs and pregnant women at different gestational stages. Selectable parameters such as patient type, tube voltage, filtration thickness, beam direction, field size, and irradiation site are also considered in VirtualDose-IR. The software has been implemented using the 'Software as a Service (SaaS)' delivery concept permitting simultaneous multi-user, multi-platform access without requiring local installation. The patient doses resulting from different target sites and patient populations were reported using the VirtualDose-IR system. The patient doses under different source to surface distances (SSD) and beam angles calculated by VirtualDose-IR and Monte Carlo simulations were compared. For most organs, the dose differences between VirtualDose-IR results and Monte Carlo results were less than 0.3 mGy at 15 000 mGy * cm2 kerma-area product (KAP). The organ dose results were compared with measurement data previously reported in literatures. The doses to organs that were located within the irradiation field match closely with experimental measurement data. The differences in the effective dose values between calculated using VirtualDose-IR and those measured were less than 2.5%. The dose errors of most organs between VirtualDose-IR and literature results were less than 40%. These results validate the accuracy of organ doses reported by VirtualDose-IR. With the inclusion of pre-specified clinical IR examination parameters (such as beam direction, target location, field of view and beam quality) and the latest anatomically realistic patient phantoms in Monte Carlo simulations, VirtualDose-IR provides users with accurate dose information in order to systematically compare, evaluate, and optimize IR plans.

DEVELOPMENT OF A SET OF MESH-BASED AND AGE-DEPENDENT CHINESE PHANTOMS AND APPLICATION FOR CT DOSE CALCULATIONS.

Phantoms for organ dose calculations are essential in radiation protection dosimetry. This article describes the development of a set of mesh-based and age-dependent phantoms for Chinese populations using reference data recommended by the Chinese government and by the International Atomic Energy Agency (IAEA). Existing mesh-based RPI adult male (RPI-AM) and RPI adult female (RPI-AF) phantoms were deformed to form new phantoms according to anatomical data for the height and weight of Chinese individuals of 5 years old male, 5 years old female, 10 years old male, 10 years old female,15 years old male, 15 years old female, adult male and adult female-named USTC-5 M, USTC-5F, USTC-10M, USTC-10F, USTC-15M, USTC-15F, USTC-AM and USTC-AF, respectively. Following procedures to ensure the accuracy, more than 120 organs/tissues in each model were adjusted to match the Chinese reference parameters and the mass errors were within 0.5%. To demonstrate the usefulness, these new set of phantoms were combined with a fully validated model of the GE LightSpeed Pro 16 multi-detector computed tomography (MDCT) scanner and the GPU-based ARCHER Monte Carlo code to compute organ doses from CT examinations. Organ doses for adult models were then compared with the data of RPI-AM and RPI-AF under the same conditions. The absorbed doses and the effective doses of RPI phantoms are found to be lower than these of the USTC adult phantoms whose body sizes are smaller. Comparisons for the doses among different ages and genders were also made. It was found that teenagers receive more radiation doses than adults do. Such Chinese-specific phantoms are clearly better suited in organ dose studies for the Chinese individuals than phantoms designed for western populations. As already demonstrated, data derived from age-specific Chinese phantoms can help CT operators and designers to optimize image quality and doses.

MONTE CARLO SIMULATION OF OUT-OF-FIELD ORGAN DOSES AND CANCER RISK IN TANZANIA FOR RADIATION THERAPY OF UNILATERAL RETINOBLASTOMA USING A 60Co UNIT.

The use of 60Co teletherapy unit for the treatment of unilateral retinoblastoma (Rb) patients is a very common procedure in many developing countries including Tanzania. The aim of this study was to estimate organ-specific absorbed doses from an external beam radiation therapy 60Co unit for unilateral Rb and to assess the risks of the patients developing a secondary primary cancer. The absorbed dose estimations were based on a Monte Carlo method and a set of age-dependent computational male phantoms. The estimated doses were used to calculate the secondary cancer risks in out-of-field organs using the Biological Effects of Ionising Radiation VII risk models. The survival information and baseline cancer risks were based on relevant statistics for the Tanzanian population. The resulting out-of-field organ doses data showed that organs which are close to the target volume, such as the brain, salivary glands and thyroid glands, received the highest absorbed dose from scattered photons during the treatment of Rb. It was also found that the resulting photons dose to specific organs depends on the patient's age. Younger patients are more sensitive to radiation and also received higher dose contributions from the treatment head due to a larger part of the body exposed to the photon radiation. In all sites considered, the overall risks associated with radiation-induced secondary cancer were relatively lower than the baseline risks. Thus, the results in this article can help to provide good estimations of radiation-induced secondary cancer after radiation treatment of unilateral Rb using 60Co teletherapy unit in Tanzania and other developing countries.