Implementation of a knowledge-based decision support system for treatment plan auditing through automation.

BACKGROUND: Independent auditing is a necessary component of a comprehensive quality assurance (QA) program and can also be utilized for continuous quality improvement (QI) in various radiotherapy processes. Two senior physicists at our institution have been performing a time intensive manual audit of cross-campus treatment plans annually, with the aim of further standardizing our planning procedures, updating policies and guidelines, and providing training opportunities of all staff members. PURPOSE: A knowledge-based automated anomaly-detection algorithm to provide decision support and strengthen our manual retrospective plan auditing process was developed. This standardized and improved the efficiency of the assessment of our external beam radiotherapy (EBRT) treatment planning across all eight campuses of our institution. METHODS: A total of 843 external beam radiotherapy plans for 721 lung patients from January 2020 to March 2021 were automatically acquired from our clinical treatment planning and management systems. From each plan, 44 parameters were automatically extracted and pre-processed. A knowledge-based anomaly detection algorithm, namely, "isolation forest" (iForest), was then applied to the plan dataset. An anomaly score was determined for each plan using recursive partitioning mechanism. Top 20 plans ranked with the highest anomaly scores for each treatment technique (2D/3D/IMRT/VMAT/SBRT) including auto-populated parameters were used to guide the manual auditing process and validated by two plan auditors. RESULTS: The two auditors verified that 75.6% plans with the highest iForest anomaly scores have similar concerning qualities that may lead to actionable recommendations for our planning procedures and staff training materials. The time to audit a chart was approximately 20.8 min on average when done manually and 14.0 min when done with the iForest guidance. Approximately 6.8 min were saved per chart with the iForest method. For our typical internal audit review of 250 charts annually, the total time savings are approximately 30 hr per year. CONCLUSION: iForest effectively detects anomalous plans and strengthens our cross-campus manual plan auditing procedure by adding decision support and further improve standardization. Due to the use of automation, this method was efficient and will be used to establish a standard plan auditing procedure, which could occur more frequently.

Technical Assessment of an Automated Treatment Planning on Dose Escalation of Pancreas Stereotactic Body Radiotherapy.

BACKGROUND: Stereotactic body radiotherapy has been suggested to provide high rates of local control for locally advanced pancreatic cancer. However, the close proximity of highly radiosensitive normal tissues usually causes the labor-intensive planning process and may impede further escalation of the prescription dose. PURPOSE: The present study aims to evaluate the consistency and efficiency of Pinnacle Auto-Planning for pancreas stereotactic body radiotherapy with original prescription and escalated prescription. METHODS: Twenty-four patients with pancreatic cancer treated with stereotactic body radiotherapy were studied retrospectively. The prescription is 40 Gy over 5 consecutive fractions. Most of patients (n = 21) also had 3 other different dose-level targets (6 Gy/fraction, 5 Gy/fraction, and 4 Gy/fraction). Two types of plans were generated by Pinnacle Auto-Planning with the original prescription (8 Gy/fraction, 6 Gy/fraction, 5 Gy/fraction, and 4 Gy/fraction) and escalated prescription (9 Gy/fraction, 7 Gy/fraction, 6 Gy/fraction, and 5 Gy/fraction), respectively. The same Auto-Planning template, including beam geometry, intensity-modulated radiotherapy objectives and intensity-modulated radiotherapy optimization parameters, were utilized for all the auto-plans in each prescription group. The intensity-modulated radiotherapy objectives do not include any manually created structures. Dosimetric parameters including percentage volume of PTV receiving 100% of the prescription dose, percentage volume of PTV receiving 93% of the prescription dose, and consistency of the dose-volume histograms of the target volumes were assessed. Dmax and D1 cc of highly radiosensitive organs were also evaluated. RESULTS: For all the pancreas stereotactic body radiotherapy plans with the original or escalated prescriptions, auto-plans met institutional dose constraints for critical organs, such as the duodenum, small intestine, and stomach. Furthermore, auto-plans resulted in acceptable planning target volume coverage for all targets with different prescription levels. All the plans were generated in a one-attempt manner, and very little human intervention is necessary to achieve such plan quality. CONCLUSIONS: Pinnacle3 Auto-Planning consistently and efficiently generate acceptable treatment plans for multitarget pancreas stereotactic body radiotherapy with or without dose escalation and may play a more important role in treatment planning in the future.

HOX transcript antisense RNA is elevated in gastric carcinogenesis and regulated by the NF-κB pathway.

The expression pattern of HOX transcript antisense RNA (HOTAIR) in the progression of gastric cancer and the regulation of its expression are still unclear. In the current study, HOTAIR expressions in gastric tissues collected from patients with superficial gastritis, atrophic gastritis, atypical hyperplasia, and gastric cancer as well as normal controls was quantitatively examined. The results showed that the expression of HOTAIR was higher in gastric cancer than in normal tissues, but reached the highest level in atrophic gastritis, suggesting that HOTAIR may be involved in the molecular process of nonresolving inflammation. Then tumor necrosis factor-α-induced protein-8 like-2 (TIPE2), a known gene associated with nonresolving inflammation, was overexpressed and the results showed that the promotion in TIPE2 expression triggered HOTAIR reduction, this result was further verified by microarray analysis and TIPE2 knockout mice. Subsequently, the data obtained from HOTAIR knockdown experiment showed that it significantly enhanced colony forming capability and inhibited p27 expression in AGS cells. Furthermore, deletion constructs and luciferase-based activity assays indicated that the -475 to -443bp region of HOTAIR promoter contained a crucial regulatory element. Transcription factor prediction with software TRANSFAC revealed that nuclear factor-κB signaling protein p65 had a binding site in this region and might have roles in HOTAIR expression. The binding of phosphor-p65 to HOTAIR promoter was verified by chromatin immunoprecipitation, and succeeding experiment results demonstrated that p65 reduction by p65 small interfering RNA and TIPE2 overexpression also decreased HOTAIR expression. Conclusively, our results suggest that HOTAIR was associated with nonresolving inflammation, and its expression is regulated by p65.

Application of statistical and computational methodology to predict brainstem dosimetry for trigeminal neuralgia stereotactic radiosurgery.

OBJECTIVES: To apply advanced statistical and computational methodology in evaluating the impact of anatomical and technical variables on normal tissue dosimetry of trigeminal neuralgia (TN) stereotactic radiosurgery (SRS). METHODS: Forty patients treated with LINAC-based TN SRS with 90 Gy maximum dose were randomly selected for the study. Parameters extracted from the treatment plans for the study included three dosimetric output variables: the maximum dose to the brainstem (BSmax), the volume of brainstem that received at least 10 Gy (V10BS), and the volume of normal brain that received at least 12 Gy (V12). We analyzed five anatomical variables: the incidence angle of the nerve with the brainstem surface (A), the nerve length (L), the nerve width as measured both axially (WA) and sagittally (WS), the distance measured along the nerve between the isocenter and the brainstem surface (D), and one technical variable: the utilized cone size (CS). Univariate correlation was calculated for each pair among all parameters. Multivariate regression models were fitted for the output parameters using the optimal input parameters selected by the Gaussian graphic model LASSO. Repeated twofold cross-validations were used to evaluate the models. RESULTS: Median BSmax, V10BS, and V12 for the 40 patients were 35.7 Gy, 0.14 cc, and 1.28 cc, respectively. Median A, L, WA, WS, D, and CS were 43.7°, 8.8 mm, 2.8 mm, 2.7 mm, 4.8 mm, and 6 mm, respectively. Of the three output variables, BSmax most strongly correlated with the input variables. Specifically, it had strong, negative correlations with the input anatomical variables and a positive correlation with CS. The correlation between D and BSmax at -0.51 was the strongest correlation between single input and output parameters, followed by that between CS and V10BS at 0.45, and that between A and BSmax at -0.44. V12 was most correlated with cone size alone, rather than anatomy. LASSO identified an optimal 3-feature combination of A, D, and CS for BSmax and V10BS prediction. Using cross-validation, the multivariate regression models with the three selected features yielded stronger correlations than the correlation between the BSmax and V10BS themselves. CONCLUSIONS: For the first time, an advanced statistical and computational methodology was applied to study the impact of anatomical and technical variables on TN SRS. The variables were found to impact brainstem doses, and reasonably strong correlation models were established using an optimized 3-feature combination including the nerve incidence angle, cone size, and isocenter-brainstem distance.

Using weighted power mean for equivalent square estimation.

PURPOSE: Equivalent Square (ES) enables the calculation of many radiation quantities for rectangular treatment fields, based only on measurements from square fields. While it is widely applied in radiotherapy, its accuracy, especially for extremely elongated fields, still leaves room for improvement. In this study, we introduce a novel explicit ES formula based on Weighted Power Mean (WPM) function and compare its performance with the Sterling formula and Vadash/Bjärngard's formula. METHODS: The proposed WPM formula is ESWPMa,b=waα+1-wbα1/α for a rectangular photon field with sides a and b. The formula performance was evaluated by three methods: standard deviation of model fitting residual error, maximum relative model prediction error, and model's Akaike Information Criterion (AIC). Testing datasets included the ES table from British Journal of Radiology (BJR), photon output factors (Scp ) from the Varian TrueBeam Representative Beam Data (Med Phys. 2012;39:6981-7018), and published Scp data for Varian TrueBeam Edge (J Appl Clin Med Phys. 2015;16:125-148). RESULTS: For the BJR dataset, the best-fit parameter value α = -1.25 achieved a 20% reduction in standard deviation in ES estimation residual error compared with the two established formulae. For the two Varian datasets, employing WPM reduced the maximum relative error from 3.5% (Sterling) or 2% (Vadash/Bjärngard) to 0.7% for open field sizes ranging from 3 cm to 40 cm, and the reduction was even more prominent for 1 cm field sizes on Edge (J Appl Clin Med Phys. 2015;16:125-148). The AIC value of the WPM formula was consistently lower than its counterparts from the traditional formulae on photon output factors, most prominent on very elongated small fields. CONCLUSION: The WPM formula outperformed the traditional formulae on three testing datasets. With increasing utilization of very elongated, small rectangular fields in modern radiotherapy, improved photon output factor estimation is expected by adopting the WPM formula in treatment planning and secondary MU check.

A comprehensive dosimetric study on switching from a Type-B to a Type-C dose algorithm for modern lung SBRT.

BACKGROUND: Type-C dose algorithms provide more accurate dosimetry for lung SBRT treatment planning. However, because current dosimetric protocols were developed based on conventional algorithms, its applicability for the new generation algorithms needs to be determined. Previous studies on this issue used small sample sizes and reached discordant conclusions. Our study assessed dose calculation of a Type-C algorithm with current dosimetric protocols in a large patient cohort, in order to demonstrate the dosimetric impacts and necessary treatment planning steps of switching from a Type-B to a Type-C dose algorithm for lung SBRT planning. METHODS: Fifty-two lung SBRT patients were included, each planned using coplanar VMAT arcs, normalized to D95% = prescription dose using a Type-B algorithm. These were compared against three Type-C plans: re-calculated plans (identical plan parameters), re-normalized plans (D95% = prescription dose), and re-optimized plans. Dosimetric endpoints were extracted and compared among the four plans, including RTOG dosimetric criteria: (R100%, R50%, D2cm, V105%, and lung V20), PTV Dmin, Dmax, Dmean, V% and D90%, PTV coverage (V100%), homogeneity index (HI), and Paddick conformity index (PCI). RESULTS: Re-calculated Type-C plans resulted in decreased PTV Dmin with a mean difference of 5.2% and increased Dmax with a mean difference of 3.1%, similar or improved RTOG dose compliance, but compromised PTV coverage (mean D95% and V100% reduction of 2.5 and 8.1%, respectively). Seven plans had >5% D95% reduction (maximum reduction = 16.7%), and 18 plans had >5% V100% reduction (maximum reduction = 60.0%). Re-normalized Type-C plans restored target coverage, but yielded degraded plan conformity (average PCI reduction 4.0%), and RTOG dosimetric criteria deviation worsened in 11 plans, in R50%, D2cm, and R100%. Except for one case, re-optimized Type-C plans restored RTOG compliance achieved by the original Type-B plans, resulting in similar dosimetric values but slightly higher target dose heterogeneity (mean HI increase = 13.2%). CONCLUSIONS: Type-B SBRT lung plans considerably overestimate target coverage for some patients, necessitating Type-C re-normalization or re-optimization. Current RTOG dosimetric criteria appear to remain appropriate.

Still equivalent for dose calculation in the Monte Carlo era? A comparison of free breathing and average intensity projection CT datasets for lung SBRT using three generations of dose calculation algorithms.

PURPOSE: Inhomogeneity dose modeling and respiratory motion description are two critical technical challenges for lung stereotactic body radiotherapy, an important treatment modality for small size primary and secondary lung tumors. Recent studies revealed lung density-dependent target dose differences between Monte Carlo (Type-C) algorithm and earlier algorithms. Therefore, this study aimed to investigate the equivalence of the two most popular CT datasets for treatment planning, free breathing (FB) and average intensity projection (AIP) CTs, using Type-C algorithms, and comparing with two older generation algorithms (Type-A and Type-B). METHODS: Twenty patients (twenty-one lesions) were planned using a Type-A algorithm on the FB CT. Lung was contoured separately on FB and AIP CTs and compared. Dose comparison was obtained between the two CTs using four commercial dose algorithms including one Type-A (Pencil Beam Convolution - PBC), one Type-B (Analytical Anisotropic Algorithm - AAA), and two Type-C algorithms (Voxel Monte Carlo - VMC and Acuros External Beam - AXB). For each algorithm, the dosimetric parameters of the target (PTV, Dmin , Dmax , Dmean , D95, and D90) and lung (V5, V10, V20, V30, V35, and V40) were compared between the two CTs using the Wilcoxon signed rank test. Correlation between dosimetric differences and density differences for each algorithm were studied using linear regression and Spearman correlation, in which both global and local density differences were evaluated. RESULTS: Although the lung density differences on FB and AIP CTs were statistically significant (P = 0.003), the magnitude was small at 1.21 ± 1.45%. Correspondingly, for the two Type-C algorithms, target and lung dosimetric differences were small in magnitude and statistically insignificant (P > 0.05) for all but one instance, similar to the findings for the older generation algorithms. Nevertheless, a significant correlation was shown between the dosimetric and density differences for Type-C and Type-B algorithms, but not for the Type-A algorithm. CONCLUSIONS: With the capability to more accurately model inhomogeneity, Monte Carlo (Type-C) algorithms are sensitive to respiration-induced local and global tissue density changes and exhibit a strong correlation between dosimetric and density differences. However, FB and AIP CTs may still be considered equivalent for dose calculation in the Monte Carlo era, due to the small magnitude of lung density differences between these two datasets.

Automatic planning on hippocampal avoidance whole-brain radiotherapy.

Mounting evidence suggests that radiation-induced damage to the hippocampus plays a role in neurocognitive decline for patients receiving whole-brain radiotherapy (WBRT). Hippocampal avoidance whole-brain radiotherapy (HA-WBRT) has been proposed to reduce the putative neurocognitive deficits by limiting the dose to the hippocampus. However, urgency of palliation for patients as well as the complexities of the treatment planning may be barriers to protocol enrollment to accumulate further clinical evidence. This warrants expedited quality planning of HA-WBRT. Pinnacle3 Automatic treatment planning was designed to increase planning efficiency while maintaining or improving plan quality and consistency. The aim of the present study is to evaluate the performance of the Pinnacle3 Auto-Planning on HA-WBRT treatment planning. Ten patients previously treated for brain metastases were selected. Hippocampal volumes were contoured on T1 magnetic resonance (MR) images, and planning target volumes (PTVs) were generated based on RTOG0933. The following 2 types of plans were generated by Pinnacle3 Auto-Planning: the one with 2 coplanar volumetric modulated arc therapy (VMAT) arcs and the other with 9-field noncoplanar intensity-modulated radiation therapy (IMRT). D2% and D98% of PTV were used to calculate homogeneity index (HI). HI and Paddick Conformity index (CI) of PTV as well as D100% and Dmax of the hippocampus were used to evaluate the plan quality. All the auto-plans met the dose coverage and constraint objectives based on RTOG0933. The auto-plans eliminated the necessity of generating pseudostructures by the planners, and it required little manual intervention which expedited the planning process. IMRT quality assurance (QA) results also suggest that all the auto-plans are practically acceptable on delivery. Pinnacle3 Auto-Planning generates acceptable plans by RTOG0933 criteria without time-consuming planning process. The expedited quality planning achieved by Auto-Planning (AP) may facilitate protocol enrollment of patients to further investigate the hippocampal-sparing effect and be used to ensure timely start of palliative treatment in future clinical practice.

Evaluating inter-campus plan consistency using a knowledge based planning model.

BACKGROUND AND PURPOSE: We investigate whether knowledge based planning (KBP) can identify systematic variations in intensity modulated radiotherapy (IMRT) plans between multiple campuses of a single institution. MATERIAL AND METHODS: A KBP model was constructed from 58 prior main campus (MC) esophagus IMRT radiotherapy plans and then applied to 172 previous patient plans across MC and 4 regional sites (RS). The KBP model predicts DVH bands for each organ at risk which were compared to the previously planned DVHs for that patient. RESULTS: RS1's plans were the least similar to the model with less heart and stomach sparing, and more variation in liver dose, compared to MC. RS2 produced plans most similar to those expected from the model. RS3 plans displayed more variability from the model prediction but overall, the DVHs were no worse than those of MC. RS4 did not present any statistically significant results due to the small sample size (n=11). CONCLUSIONS: KBP can retrospectively highlight subtle differences in planning practices, even between campuses of the same institution. This information can be used to identify areas needing increased consistency in planning output and subsequently improve consistency and quality of care.

Interobserver variability in radiation therapy plan output: Results of a single-institution study.

PURPOSE: We investigated the sources of variability in radiation therapy treatment plan output between planners within a single institution. METHODS AND MATERIALS: Forty treatment planners across 5 campuses of an institution created a plan on the same thoracic esophagus patient computed tomography scan and structure set. Plans were scored and ranked based on the planner's adherence to an ordered list of target dose coverage and normal tissue evaluation criteria. A runs test was used to identify whether any of the studied planner qualities influenced the ranking. Spearman rank correlation was used to investigate whether plan score correlated with years of experience or planned monitor units. RESULTS: The distribution of scores, ranging from 80.24 to 135.89, was negatively skewed (mean, 128.7; median, 131.5). No statistically significant relationship between plan score and campus (P = .193), job title (P = .174), previous outside experience (P = .611), or number of gantry angles (P = .156) was discovered. No statistical correlation between plan score and monitor unit or years of experience was found. CONCLUSIONS: Despite clear and established critical organ dose criteria and well-documented planning guidelines, planning variation still occurs, even among members of the same institution. Because plan consistency does not seem to significantly correlate with experience, career path, or campus, investigation into alternate methods beyond additional education and training to reduce this variation, such as knowledge-based planning or advanced optimization techniques, is necessary.