DVH Analytics: A DVH database for clinicians and researchers

Phase I quality control framework for monitoring organ-at-risk dose

Objective.The aim of this work was to develop a Phase I control chart framework for the recently proposed multivariate risk-adjusted Hotelling'sT2chart. Although this control chart alone can identify most patients receiving extreme organ-at-risk (OAR) dose, it is restricted by underlying distributional assumptions, making it sensitive to extreme observations in the sample, as is typically found in radiotherapy plan quality data such as dose-volume histogram (DVH) points. This can lead to slightly poor-quality plans that should have been identified as out-of-control (OC) to be signaled in-control (IC).Approach. We develop a robust iterative control chart framework to identify all OC patients with abnormally high OAR dose and improve them via re-optimization to achieve an IC sample prior to establishing the Phase I control chart, which can be used to monitor future treatment plans.Main Results. Eighty head-and-neck patients were used in this study. After the first iteration, P14, P67, and P68 were detected as OC for high brainstem dose, warranting re-optimization aimed to reduce brainstem dose without worsening other planning criteria. The DVH and control chart were updated after re-optimization. On the second iteration, P14, P67, and P68 were IC, but P40 was identified as OC. After re-optimizing P40's plan and updating the DVH and control chart, P40 was IC, but P14* (P14's re-optimized plan) and P62 were flagged as OC. P14* could not be re-optimized without worsening target coverage, so only P62 was re-optimized. Ultimately, a fully IC sample was achieved. Multiple iterations were needed to identify and improve all OC patients, and to establish a more robust control limit to monitor future treatment plans.Significance. The iterative procedure resulted in a fully IC sample of patients. With this sample, a more robust Phase I control chart that can monitor OAR doses of new plans was established.

Determining risk and predictors of head and neck cancer treatment-related lymphedema: A clinicopathologic and dosimetric data mining approach using interpretable machine learning and ensemble feature selection

Background and purpose

The ability to determine the risk and predictors of lymphedema is vital in improving the quality of life for head and neck (HN) cancer patients. However, selecting robust features is challenging due to the multicollinearity and high dimensionality of radiotherapy (RT) data. This study aims to overcome these challenges using an ensemble feature selection technique with machine learning (ML).

Materials and methods

Thirty organs-at-risk, including bilateral cervical lymph node levels, were contoured, and dose-volume data were extracted from 76 HN treatment plans. Clinicopathologic data was collected. Ensemble feature selection was used to reduce the number of features. Using the reduced features as input to ML and competing risk models, internal and external lymphedema prediction capability was evaluated with the ML models, and time to lymphedema event and risk stratification were estimated using the risk models.

Results

Two ML models, XGBoost and random forest, exhibited robust prediction performance. They achieved average F1-scores and AUCs of 84 ± 3.3 % and 79 ± 11.9 % (external lymphedema), and 64 ± 12 % and 78 ± 7.9 % (internal lymphedema). Predictive ML and risk models identified common predictors, including bulky node involvement, high dose to various lymph node levels, and lymph nodes removed during surgery. At 180 days, removing 0-25, 26-50, and > 50 lymph nodes increased external lymphedema risk to 72.1 %, 95.6 %, and 57.7 % respectively (p = 0.01).

Conclusion

Our approach, involving the reduction of HN RT data dimensionality, resulted in effective ML models for HN lymphedema prediction. Predictive dosimetric features emerged from both predictive and competing risk models. Consistency with clinicopathologic features from other studies supports our methodology.

Artificial intelligence and machine learning: Definition of terms and current concepts in critical care research

With increasing computing power, artificial intelligence (AI) and machine learning (ML) have prospered, which facilitate the analysis of large datasets, especially those found in critical care. It is important to define these terminologies, to inform a standardized approach to critical care research. This manuscript hopes to clarify these terms with examples from medical literature. Three major components that are required for a successful ML implementation: (i) reliable dataset, (ii) ML algorithm, and (iii) unbiased model evaluation, are discussed. A reliable dataset can be structured or unstructured with limited noise, outliers, and missing values. ML, a subset of AI, is typically focused on supervised or unsupervised learning tasks in which the output is based on inputs and derived from iterative pattern recognition algorithms, while AI is the overall ability of a machine to "think" or mimic human behavior; and to analyze data free from human influence. Even with successful implementation, advanced AI and ML algorithms have faced challenges in adoption into practice, mainly due to their lack of interpretability, which hinders trust, buy-in, and engagement from clinicians. Consequently, traditional algorithms, such as linear and logistic regression, that may have reduced predictive power but are highly interpretable, continue to be widely used.

End-to-end framework for automated collection of large multicentre radiotherapy datasets demonstrated in a Danish Breast Cancer Group cohort

Large Digital Imaging and Communications in Medicine (DICOM) datasets are key to support research and the development of machine learning technology in radiotherapy (RT). However, the tools for multi-centre data collection, curation and standardisation are not readily available. Automated batch DICOM export solutions were demonstrated for a multicentre setup. A Python solution, Collaborative DICOM analysis for RT (CORDIAL-RT) was developed for curation, standardisation, and analysis of the collected data. The setup was demonstrated in the DBCG RT-Nation study, where 86% (n = 7748) of treatments in the inclusion period were collected and quality assured, supporting the applicability of the end-to-end framework.

Predictors of Pneumonitis in Combined Thoracic Stereotactic Body Radiotherapy and Immunotherapy

PURPOSE

Thoracic stereotactic body radiotherapy (SBRT) is associated with high rates of local control but carries a risk of pneumonitis. Immunotherapy is a standard treatment for patients with metastatic disease but can also cause pneumonitis. To evaluate the feasibility and safety of thoracic SBRT with systemic immunotherapy, clinical outcomes of patients treated with immune checkpoint blockade (ICB) and SBRT on prospective trials were reviewed.

METHODS AND MATERIALS

Three consecutive phase 1 trials of combination SBRT and ICB conducted between 2016-2020 for widely metastatic solid tumors were reviewed. The protocols mandated adherence to NRG BR001/BR002 OAR constraints, resulting in <100% coverage of some target volumes. ICB was administered either sequentially (within 7 days after completion of SBRT) or concurrently (before or at the start of SBRT), depending on protocol. Endpoints included pneumonitis, dose-volume constraints, local failure, and overall survival (OS). The cumulative incidence estimator and Kaplan-Meier method were used.

RESULTS

123 patients met eligibility with 311 metastases irradiated. The most common histologies included non-small cell lung cancer (33%) and colorectal cancer (12%). Median follow up was 12 months. The overall rate of grade 3+ pneumonitis was 8.1%. 1-year local failure was 3.6%. Established dosimetric parameters were significantly associated with the development of pneumonitis (p<0.05). In most patients, the lungs were not challenged with high doses of radiation, defined as receiving ≥75% of the maximum for a given lung dose-volume constraint. Patients who were challenged were not found to have a significantly higher risk of pneumonitis.

CONCLUSIONS

In the largest series of thoracic SBRT and immunotherapy, local control was excellent with acceptable toxicity and support the conclusion that established dose-volume constraints for the lung are safe. However, these results highlight the potential value in reporting of OARs being challenged with doses approaching protocol specified limits.

Automated scripting of the dosimetric evaluation of adaptive versus non-adaptive radiotherapy

Objective. To quantify the benefit of adaptive radiotherapy over non-adaptive radiotherapy it is useful to extract and compare dosimetric features of patient treatments in both scenarios. This requires Image-Guided Radiotherapy (IGRT) matching of baseline planning to adaptive fraction imaging, followed by extraction of relevant dose metrics. This can be impractical to retrospectively perform manually for multiple patients.Approach. Here we present an algorithm for automatic IGRT matching of baseline planning with fraction imaging and performing automated dosimetric feature extraction from adaptive and non-adaptive treatment plans, thereby allowing comparison of the two scenarios. This workflow can be done in an entirely automated way via scripting solutions given structure and dose Digital Imaging and Communications in Medicine (DICOM) files from baseline and adaptive fractions. We validate this algorithm against the results of manual IGRT matching. We also demonstrate automated dosimetric feature extraction. Lastly, we combine these two scripting solutions to extract daily adaptive and non-adaptive radiotherapy dosimetric features from an initial cohort of patients treated on an MRI guided linear accelerator (MR-LINAC).Results.Our results demonstrate that automated feature extraction and IGRT matching was successful and comparable to results performed by a manual operator. We have therefore demonstrated a method for easy analysis of patients treated on an adaptive radiotherapy platform.Significance.We believe that this scripting solution can be used for quantifying the benefits of adaptive therapy and for comparing adaptive therapy against various non-adaptive IGRT scenarios (e.g. 6 degree of freedom couch rotation).

The impact of nutritional counseling on thyroid disorders in head and neck cancer patients after (chemo)radiotherapy: results from a prospective interventional trial

Objective

To analyze the impact of nutritional counseling on the development of hypothyroidism after (chemo)radiotherapy in head and neck cancer patients to propose a new normal tissue complication probability (NTCP) model.

Materials and methods

At baseline, at the end of (chemo)radiotherapy, and during follow-up, thyroid-stimulating hormone (TSH) with free thyroxin (fT3 and fT4), nutritional status, and nutrient intake were prospectively analyzed in 46 out of 220 screened patients. Patients received (chemo)radiotherapy within an intervention (individual nutritional counseling every 2 weeks during therapy) and a control group (no nutritional counseling).

Results

Overall median follow-up was 16.5 [IQR: 12; 22] months. Fourteen patients (30.4%) presented with hypothyroidism after 13.5 [8.8; 17] months. During (chemo)radiotherapy, nutritional status worsened in the entire cohort: body mass index (p < 0.001) and fat-free mass index (p < 0.001) decreased, calorie deficit (p = 0.02) increased, and the baseline protein intake dropped (p = 0.028). The baseline selenium intake (p = 0.002) increased until the end of therapy. Application of the NTCP models by Rønjom, Cella, and Boomsma et al. resulted in good performance of all three models, with an AUC ranging from 0.76 to 0.78. Our newly developed NTCP model was based on baseline TSH and baseline ferritin. Model performance was good, receiving an AUC of 0.76 (95% CI: 0.61–0.87), with a sensitivity of 57.1% and specificity of 96.9% calculated for a Youden index of 0.73 (p = 0.004; area = 0.5).

Conclusion

Baseline TSH and ferritin act as independent predictors for radiotherapy-associated hypothyroidism. The exclusion of such laboratory chemistry parameters in future NTCP models may result in poor model performance.

Customization of a Model For Knowledge-Based Planning to Achieve Ideal Dose Distributions in Volume Modulated arc Therapy for Pancreatic Cancers

Purpose:

To evaluate customizing a knowledge-based planning (KBP) model using dosimetric analysis for volumetric modulated arc therapy for pancreatic cancer.

Materials and Methods:

The first model (M1) using 56 plans and the second model (M2) using 31 plans were created in the first 7 months of the study. The ratios of volume of both kidneys overlapping the expanded planning target volume to the total volume of both kidneys (V_overlap/V_whole) were calculated in all cases to customize M1. Regression lines were derived from V_overlap/V_whole and mean dose to both kidneys. The third model (M3) was created using 30 plans which data put them below the regression line. For validation, KBP was performed with the three models on 21 patients.

Results:

V₁₈ of the left kidney for M1 plans was 7.3% greater than for clinical plans. Dmean of the left kidney for M2 plans was 2.2% greater than for clinical plans. There was no significant difference between all kidney doses in M3 and clinical plans. Dmean of the left kidney for M2 plans was 2.2% greater than for clinical plans. Dmean to both kidneys did not differ significantly between the three models in validation plans with V_overlap/V_whole lower than average. In plans with larger than average volumes, the Dmean of validation plans created by M3 was significantly lower for both kidneys by 1.7 and 0.9 Gy than with M1 and M2, respectively.

Conclusions:

Selecting plans to register in a model by analyzing dosimetry and geometry is an effective means of improving the KBP model.

Split-VMAT technique to control the deep inspiration breath hold time for breast cancer radiotherapy

BACKGROUND

To improve split-VMAT technique by optimizing treatment delivery time for deep-inspiration breath hold (DIBH) radiotherapy in left-sided breast cancer patients, when automatic beam-interruption devices are not available.

METHODS

Ten consecutive patients were treated with an eight partial arcs (8paVMAT) plan, standard of care in our center. A four partial arcs (4paVMAT) plan was also created and actual LINAC outputs were measured, to evaluate whether there was a dosimetric difference between both techniques and potential impact on the delivered dose. Subsequently, ten other patients were consecutively treated with a 4paVMAT plan to compare the actual treatment delivery time between both techniques. The prescribed dose was 40.05 Gy/15 fractions on the PTV breast (breast or thoracic wall), lymph nodes (LN) and intramammary lymph node chain (IMN). Treatment delivery time, PTVs coverage, conformity index (CI), organs at risk (OAR) dose, monitor units (MU), and gamma index were compared.

RESULTS

Both split-VMAT techniques resulted in similar dose coverage for the PTV Breast and LN, and similar CI. For PTV IMN we observed a 5% increased coverage for the volume receiving ≥ 36 Gy with 4paVMAT, with an identical volume receiving ≥ 32 Gy. There was no difference for the OAR sparing, with the exception of the contralateral organs: there was a 0.6 Gy decrease for contralateral breast mean (p ≤ 0.01) and 1% decrease for the volume of right lung receiving ≥ 5 Gy (p = 0.024). Overall, these results indicate a modest clinical benefit of using 4paVMAT in comparison to 8paVMAT. An increase in the number of MU per arc was observed for the 4paVMAT technique, as expected, while the total number of MU remained comparable for both techniques. All the plans were measured with the Delta4 phantom and passed the gamma index criteria with no significant differences. Finally, the main difference was seen for the treatment delivery time: there was a significant decrease from 8.9 to 5.4 min for the 4paVMAT plans (p < .05).

CONCLUSIONS

This study is mainly of interest for centers who are implementing the DIBH technique without automatic beam-holding devices and who therefore may require to manually switch the beam on and off during breast DIBH treatment. Split-VMAT technique with 4 partial arcs significantly reduces the treatment delivery time compared to 8 partial arcs, without compromising the target coverage and the OAR sparing. The technique decreases the number of breath holds per fraction, resulting in a shorter treatment session.

Treatment plan quality control using multivariate control charts

PURPOSE

Statistical process control tools such as control charts were recommended by the American Association of Physicists in Medicine (AAPM) Task Group 218 for radiotherapy quality assurance. However, the tools needed to analyze multivariate, correlated data that are often encountered in treatment plan quality measures, are lacking. In this study, we develop quality control tools that can model multivariate plan quality measures with correlations and account for patient-specific risk factors, without adding a significant burden to clinical workflow.

METHODS AND MATERIALS

A multivariate, quality control chart is developed that includes a risk-adjustment model, Hotelling's T² statistic, and principal component analysis (PCA). Principal component analysis accounts for correlations among a set of organ-at-risk (OAR) dose-volume histogram (DVH) points that serves as proxies for plan quality. Risk-adjustment models estimate the principal components from PCA using a set of patient- and treatment-specific risk factors. The resulting residuals from the risk-adjustment models are used to compute the Hotelling's T² statistic; the corresponding multivariate control chart is then plotted based on the beta distribution followed by the statistic. Further, the box-cox transformation is used to account for non-normality in DVH points. We investigate the application of the proposed methodology via three multivariate control charts - a conventional chart that ignores risk-adjustment and PCA, a risk-adjusted chart ignoring PCA, and a PCA-based, risk-adjusted chart. These control charts are evaluated on 69 head-and-neck cases.

RESULTS

The conventional multivariate control chart fails to account for important patient-specific risk factors, including volumes and cross-sectional areas of the tumor and OARs and distances in-between. This failure leads to a larger number of false alarms. While the multivariate risk-adjusted control chart is able to reduce false alarms, it fails to account for correlations in DVH points. The multivariate PCA-based, risk-adjusted control chart can detect unusual plans after accounting for the correlations. By replanning, improvements are shown on an unusual plan identified by both risk-adjusted methods.

CONCLUSIONS

The multivariate risk-adjusted control chart developed here enables quality control of plans prior to delivery. This methodology is generic and can be readily applied for other radiotherapy quality assurance protocols, such as gamma analysis pass rates.

Two-Level Factorial Pre-TomoBreast Pilot Study of Tomotherapy and Conventional Radiotherapy in Breast Cancer: Post Hoc Utility of a Mean Absolute Dose Deviation Penalty Score

Background:

A 2-level factorial pilot study was conducted in 2007 just before starting a randomized clinical trial comparing tomotherapy and conventional radiotherapy (CR) to reduce cardiac and pulmonary adverse effects in breast cancer, considering tumor laterality (left/right), target volume (with/without nodal irradiation), surgery (tumorectomy/mastectomy), and patient position (prone/supine). The study was revisited using a penalty score based on the recently developed mean absolute dose deviation (MADD).

Methods:

Eight patients with a unique combination of laterality, nodal coverage, and surgery underwent dual tomotherapy and CR treatment planning in both prone and supine positions, providing 32 distinct combinations. The penalty score was applied using the weighted sum of the MADDs. The Lenth method for unreplicated 2-level factorial design was used in the analysis.

Results:

The Lenth analysis identified nodal irradiation as the active main effect penalizing the dosimetry by 1.14 Gy (P = 0.001). Other significant effects were left laterality (0.94 Gy), mastectomy (0.61 Gy), and interactions between left mastectomy (0.89 Gy) and prone mastectomy (0.71 Gy), with P-values between 0.005 and 0.05. Tomotherapy provided a small reduction in penalty (reduction of 0.54 Gy) through interaction with nodal irradiation (P = 0.080). Some effects approached significance with P-values > 0.05 and ≤ 0.10 for interactions of prone × mastectomy × left (0.60 Gy), nodal irradiation × mastectomy (0.59 Gy), and prone × left (0.55 Gy) and the main effect prone (0.52 Gy).

Conclusions:

The historical dosimetric analysis previously revealed the feasibility of tomotherapy, but a conclusion could not be made. The MADD-based score is promising, and a new analysis highlights the impact of factors and hierarchy of priorities that need to be addressed if major gains are to be attained.

A Risk-Adjusted Control Chart to Evaluate Intensity Modulated Radiation Therapy Plan Quality

Purpose

This study aimed to develop a quality control framework for intensity modulated radiation therapy plan evaluations that can account for variations in patient- and treatment-specific risk factors.

Methods and Materials

Patient-specific risk factors, such as a patient's anatomy and tumor dose requirements, affect organs-at-risk (OARs) dose-volume histograms (DVHs), which in turn affects plan quality and can potentially cause adverse effects. Treatment-specific risk factors, such as the use of chemotherapy and surgery, are clinically relevant when evaluating radiation therapy planning criteria. A risk-adjusted control chart was developed to identify unusual plan quality after accounting for patient- and treatment-specific risk factors. In this proof of concept, 6 OAR DVH points and average monitor units serve as proxies for plan quality. Eighteen risk factors are considered for modeling quality: planning target volume (PTV) and OAR cross-sectional areas; volumes, spreads, and surface areas; minimum and centroid distances between OARs and the PTV; 6 PTV DVH points; use of chemotherapy; and surgery. A total of 69 head and neck cases were used to demonstrate the application of risk-adjusted control charts, and the results were compared with the application of conventional control charts.

Results

The risk-adjusted control chart remains robust to interpatient variations in the studied risk factors, unlike the conventional control chart. For the brainstem, the conventional chart signaled 4 patients with unusual (out-of-control) doses to 2% brainstem volume. However, the adjusted chart did not signal any plans after accounting for their risk factors. For the spinal cord doses to 2% brainstem volume, the conventional chart signaled 2 patients, and the adjusted chart signaled a separate patient after accounting for their risk factors. Similar adjustments were observed for the other DVH points when evaluating brainstem, spinal cord, ipsilateral parotid, and average monitor units. The adjustments can be directly attributed to the patient- and treatment-specific risk factors.

Conclusions

A risk-adjusted control chart was developed to evaluate plan quality, which is robust to variations in patient- and treatment-specific parameters.

Dosimetric variations for high-risk prostate cancer by VMAT plans due to patient’s weight changes

Purpose:

The aim of this study is to investigate the impact of anatomical changes in prostate cancer patients on the target coverage when using 6 MV beams-VMAT therapy and to propose strategies that allow us to evaluate the dose or correct it by normalization without having to perform a new simulation.

Methods and materials:

Ten patients of high-risk prostate cancer were chosen for the study. All test plans were delivered using the same isocenter and monitor units as the original plan and compared against the original unedited plan. The expansion and contraction of body contours due to size changes was mimicked by increasing and decreasing the body contour with depths of −2, −1·5, …, 1·5, 2 cm, in the anterior, and both lateral directions of the patient. A total of 90 plans were evaluated, 9 for each patient. Dose-volume histogram statistics were extracted from each plan and normalized to prescription dose.

Results:

Weight changes resulted in considerable dose modifications to the target and critical structures. Plans were found to be varied with 2·9% ± 0·3% per cm SSD change for VMAT treatment with a correlation index close to one. Therefore, doses variations were linear to the changes of depth. Gamma index evaluation was performed for the 10 renormalized plans. All of them passed criteria of 3%/3 mm in at least 98.2% of points. Eight of them passed criteria in 99% points. Gamma index 4%/4 mm passed 100% points in all patients for the chosen region of interest.

Conclusions:

The dosimetry estimation presented in this study shows important data for the radiation oncology staff to justify whether a CT rescan is necessary or not when a patient experiences weight changes during treatment. Based on the results of our study, discrepancies between real dose and planned dose were >5% for 1·7 cm of difference in external contour in the anterior and both lateral directions of the patient.

Conversion of computational human phantoms into DICOM-RT for normal tissue dose assessment in radiotherapy patients

Radiotherapy (RT) treatment planning systems (TPS) are designed for the fast calculation of dose to the tumor bed and nearby organs at risk using x-ray computed tomography (CT) images. However, CT images for a patient are typically available for only a small portion of the body, and in some cases, such as for retrospective epidemiological studies, no images may be available at all. When dose to organs that lie out-of-scan must be estimated, a convenient alternative for the unknown patient anatomy is to use a matching whole-body computational phantom as a surrogate. The purpose of the current work is to connect such computational phantoms to commercial RT TPS for retrospective organ dose estimation. A custom software with graphical user interface (GUI), called the DICOM-RT Generator, was developed in MATLAB to convert voxel computational phantoms into the digital imaging and communications in medicine radiotherapy (DICOM-RT) format, compatible with commercial TPS. DICOM CT image sets for the phantoms are created via a density-to-Hounsfield unit (HU) conversion curve. Accompanying structure sets containing the organ contours are automatically generated by tracing binary masks of user-specified organs on each phantom CT slice. The software was tested on a library of body size-dependent phantoms, the International Commission on Radiological Protection reference phantoms, and a canine voxel phantom, taking only a few minutes per conversion. The resulting DICOM-RT files were tested on several commercial TPS. As an example application, a library of converted phantoms was used to estimate organ doses for members of the National Wilms Tumor Study (NWTS) cohort. The converted phantom library, in DICOM format, and a standalone MATLAB-compiled executable of the DICOM-RT Generator are available for others to use for research purposes (http://ncidose.cancer.gov).

Automated data extraction and report analysis in computer-aided radiology audit: practice implications from post-mortem paediatric imaging

AIM

To determine local departmental adherence to the paediatric post-mortem magnetic resonance imaging (MRI) protocols, using a customised automated computational approach.

MATERIALS AND METHODS

A retrospective review of 460 whole-body post-mortem MRI examinations performed at Great Ormond Street Hospital for Children over a 5.5-year period was assessed for adherence to a full or abbreviated imaging sequence protocol. A simple computer program was developed to batch process DICOM (digital imaging and communications in medicine) files, extracting imaging sequence details, followed by natural language processing (NLP) of authorised reports to automate information extraction of diagnostic image quality.

RESULTS

The program was able to extract study parameters from the entire dataset (approximately 80 GB of data) in a few hours, and retrieve information on diagnostic image quality using NLP with an overall diagnostic accuracy for data extraction of 96.7% (445/460, 95% confidence interval [CI]: 94.7-98%). The full imaging protocol was adhered to in 305/460 (66.3%) cases, and an abbreviated protocol in 140/460 (30.4%) cases. Overall, 423/460 (91.9%) of studies were of diagnostic quality. These included 298/305 (97.7%) of the full protocol, 111/140 (79.3%) of the abbreviated protocol. In only five cases were the examinations non-diagnostic for all body systems, all of whom weighed <100 g (24.7-72 g) and imaged using the abbreviated protocol.

CONCLUSION

The present study demonstrated a successful application of an automated approach for data collection for audit and quality assessment purposes using paediatric post-mortem imaging as a specific example. Re-audit of these data following change implementation will be straightforward now that the automated workflow is clearly established.