The impact of heterogeneity correction on dosimetric parameters that predict for radiation pneumonitis

Application of auto-planning in radiotherapy for breast cancer after breast-conserving surgery

To evaluate the quality of planning target volume (PTV) and organs at risk (OAR) generated by the manual Pinnacle planning (manP) and Auto-Planning (AP) modules and discuss the feasibility of AP in the application of radiotherapy for patients with breast cancer. Thirty patients who underwent breast-conserving therapy were randomly selected. The Philips Pinnacle 9.10 treatment planning system was used to design the manP and AP modules for PTV and OAR distribution on the same computed tomography. A physician compared the plans in terms of dosimetric parameters and monitor units (MUs) using blind qualitative scoring. Statistical differences were evaluated using paired two-sided Wilcoxon's signed-rank test. On comparing the plans of AP and manP modules, the conformal index (P < 0.01) and D₅₀ (P = 0.04) of PTV in the AP group was lower than those in the manP group, while D₁ was higher (P = 0.03). In terms of dosimetry of OAR, ipsilateral lung V_{20 Gy} (P < 0.01), V_{10 Gy} (P < 0.01), V_{5 Gy} (P < 0.05), and D_mean (P < 0.01) of the AP group were better than those of the manP group. Heart V_{40 Gy} and D_mean of all patients with breast cancer in the AP group were lower than those in the manP group (P < 0.01). Moreover, 12 patients with left breast cancer had the same results (P < 0.01). The MU value of the intensity-modulated radiation therapy module designed using two different methods was higher in the AP group than in the manP group (P = 0.32), although there was no statistical significance. The AP module almost had an equal quality of PTV and dose distribution as the manP module, and its OAR was less irradiated.

Dosimetric factors and Lyman normal-tissue complication modelling analysis for predicting radiation-induced lung injury in postoperative breast cancer radiotherapy: a prospective study

To investigate the relationship between dosimetric factors, including Lyman normal-tissue complication (NTCP) parameters and radiation-induced lung injury (RILI), in postoperative breast cancer patients treated by intensity modulated radiotherapy (IMRT). 109 breast cancer patients who received IMRT between January 2012 and December 2013 were prospectively enrolled. A maximum likelihood analysis yielded the best estimates for Lyman NTCP parameters. Ten patients were diagnosed with RILI (primarily Grade 1 or Grade 2 RILI); the rate of RILI was 9.17% (10/109). Multivariate analysis demonstrated that ipsilateral lung V20 was an independent predictor (P=0.001) of RILI. Setting V20=29.03% as the cut-off value, the prediction of RILI achieved high accuracy (94.5%), with a sensitivity of 80% and specificity of 96%. The NTCP model parameters for 109 patients were m=0.437, n=0.912, and TD50(1)=17.211 Gy. The sensitivity of the modified Lyman NTCP model to predict the RILI was 90% (9/10), the specificity was 69.7% (69/99), and the accuracy was 71.6% (78/109). The RILI rate of the NTCP<9.62% in breast cancer patients was 1.43% (1/70), but the RILI rate of the NTCP>9.62% in patients with breast cancer was 23.08% (9/39), (P=0.001). In conclusion, V20 is an independent predictive factor for RILI in patients with breast cancer treated by IMRT; V20=29.03% could be a useful dosimetric parameter to predict the risk of RILI. The Lyman NTCP model parameters of the new value (m=0.437, n=0.912, TD50 (1) =17.211 Gy) can be used as an effective biological index to evaluate the risk of RILI.

Dosimetric and clinical predictors of radiation-induced lung toxicity in esophageal carcinoma

Aims and background

Radiation-induced lung toxicity occurs frequently in patients with esophageal carcinoma. This study aims to evaluate the clinical and three-dimensional dosimetric parameters associated with lung toxicity after radiotherapy for esophageal carcinoma.

Methods and study design

The records of 56 patients treated for esophageal carcinoma were reviewed. The Radiation Therapy Oncology Group criteria for grading of lung toxicity were followed. Spearman's correlation test, the chi-square test and logistic regression analyses were used for statistical analysis.

Results

Ten of the 56 patients developed acute toxicity. The toxicity grades were grade 2 in 7 patients and grade 3 in 3 patients; none of the patients developed grade 4 or worse toxicity. One case of toxicity occurred during radiotherapy and 9 occurred 2 weeks to 3 months after radiotherapy. The median time was 2.0 months after radiotherapy. Fourteen patients developed late irradiated lung injury, 3 after 3.5 months, 7 after 9 months, and 4 after 14 months. Radiographic imaging demonstrated patchy consolidation (n = 5), atelectasis with parenchymal distortion (n = 6), and solid consolidation (n = 3). For acute toxicity, the irradiated esophageal volume, number of fields, and most dosimetric parameters were predictive. For late toxicity, chemotherapy combined with radiotherapy and other dosimetric parameters were predictive. No obvious association between the occurrence of acute and late injury was observed.

Conclusions

The percent of lung tissue receiving at least 25 Gy (V25), the number of fields, and the irradiated length of the esophagus can be used as predictors of the risk of acute toxicity. Lungs V30, as well as chemotherapy combined with radiotherapy, are predictive of late lung injury.

Using machine learning to predict radiation pneumonitis in patients with stage I non-small cell lung cancer treated with stereotactic body radiation therapy

To develop a patient-specific 'big data' clinical decision tool to predict pneumonitis in stage I non-small cell lung cancer (NSCLC) patients after stereotactic body radiation therapy (SBRT). 61 features were recorded for 201 consecutive patients with stage I NSCLC treated with SBRT, in whom 8 (4.0%) developed radiation pneumonitis. Pneumonitis thresholds were found for each feature individually using decision stumps. The performance of three different algorithms (Decision Trees, Random Forests, RUSBoost) was evaluated. Learning curves were developed and the training error analyzed and compared to the testing error in order to evaluate the factors needed to obtain a cross-validated error smaller than 0.1. These included the addition of new features, increasing the complexity of the algorithm and enlarging the sample size and number of events. In the univariate analysis, the most important feature selected was the diffusion capacity of the lung for carbon monoxide (DLCO adj%). On multivariate analysis, the three most important features selected were the dose to 15 cc of the heart, dose to 4 cc of the trachea or bronchus, and race. Higher accuracy could be achieved if the RUSBoost algorithm was used with regularization. To predict radiation pneumonitis within an error smaller than 10%, we estimate that a sample size of 800 patients is required. Clinically relevant thresholds that put patients at risk of developing radiation pneumonitis were determined in a cohort of 201 stage I NSCLC patients treated with SBRT. The consistency of these thresholds can provide radiation oncologists with an estimate of their reliability and may inform treatment planning and patient counseling. The accuracy of the classification is limited by the number of patients in the study and not by the features gathered or the complexity of the algorithm.

Effect of normal lung definition on lung dosimetry and lung toxicity prediction in radiation therapy treatment planning

PURPOSE

This study aimed to compare lung dose-volume histogram (DVH) parameters such as mean lung dose (MLD) and the lung volume receiving ≥20 Gy (V20) of commonly used definitions of normal lung in terms of tumor/target subtraction and to determine to what extent they differ in predicting radiation pneumonitis (RP).

METHODS AND MATERIALS

One hundred lung cancer patients treated with definitive radiation therapy were assessed. The gross tumor volume (GTV) and clinical planning target volume (PTVc) were defined by the treating physician and dosimetrist. For this study, the clinical target volume (CTV) was defined as GTV with 8-mm uniform expansion, and the PTV was defined as CTV with an 8-mm uniform expansion. Lung DVHs were generated with exclusion of targets: (1) GTV (DVHG); (2) CTV (DVHC); (3) PTV (DVHP); and (4) PTVc (DVHPc). The lung DVHs, V20s, and MLDs from each of the 4 methods were compared, as was their significance in predicting radiation pneumonitis of grade 2 or greater (RP2).

RESULTS

There are significant differences in dosimetric parameters among the various definition methods (all Ps<.05). The mean and maximum differences in V20 are 4.4% and 12.6% (95% confidence interval 3.6%-5.1%), respectively. The mean and maximum differences in MLD are 3.3 Gy and 7.5 Gy (95% confidence interval, 1.7-4.8 Gy), respectively. MLDs of all methods are highly correlated with each other and significantly correlated with clinical RP2, although V20s are not. For RP2 prediction, on the receiver operating characteristic curve, MLD from DVHG (MLDG) has a greater area under curve of than MLD from DVHC (MLDC) or DVHP (MLDP). Limiting RP2 to 30%, the threshold is 22.4, 20.6, and 18.8 Gy, for MLDG, MLDC, and MLDP, respectively.

CONCLUSIONS

The differences in MLD and V20 from various lung definitions are significant. MLD from the GTV exclusion method may be more accurate in predicting clinical significant radiation pneumonitis.

Prediction of radiation pneumonitis in lung cancer patients: a systematic review

PURPOSE

Factors prediction in the development of radiation pneumonitis (RP) remains unclear. A meta-analysis about this was performed.

MATERIALS

Articles were searched in February 2012 from PubMed, EMBASE, Cochrane Library and CNKI (Chinese Journal Full-text Database) using the keywords "lung cancer," "radiation pneumonitis" or "radiation lung injury." The outcome was the RP incidence. We pooled the data using RevMan 5.1 software and tested the statistical heterogeneity.

RESULTS

We included the following factors: age, gender, weight loss, smoking history, complications, performance status, pre-radiation therapy (RT) pulmonary function, TNM, histological type, tumor location, pre-RT surgery, RT combined with chemotherapy (RCT), RT/RCT combined with amifostine, plasma end/pre-RT TGF-β1 ratio and irradiation volume. The significant risk factors for RP ≥ grade 2 were patients with chronic lung disease, tumor located in the middle or lower lobe, without pre-RT surgery, RCT, plasma end/pre-RT TGF-β1 ratio ≥1 and gross tumor volume (GTV). Following factors were identified significant for RP, including tumor located not in the upper lobe, smokers, combined with chronic lung diseases or diabetes mellitus, low pre-RT pulmonary function, RCT, RT/RCT without amifostine and plasma end/pre-RT TGF-β1 ratio ≥1. Dose-volume parameters included the average of mean lung dose (MLD) of disease lung, GTV and V (5), V (10) (≥34 %), V (20) (≥25 %), V (30) (≥18 %) of bilateral lung.

CONCLUSIONS

More attention should be paid to the levels of patients' pulmonary function, plasma TGF-β1 and dose-volume histogram (DVH). Rigorous studies are needed to identify the relationship between the above-mentioned factors and RP ≥grade 1 or 3.

Impact of tissue heterogeneity corrections in stereotactic body radiation therapy treatment plans for lung cancer

This study aims at evaluating the impact of tissue heterogeneity corrections on dosimetry of stereotactic body radiation therapy treatment plans. Four-dimensional computed tomography data from 15 low stage non-small cell lung cancer patients was used. Treatment planning and dose calculations were done using pencil beam convolution algorithm of Varian Eclipse system with Modified Batho Power Law for tissue heterogeneity. Patient plans were generated with 6 MV co-planar non-opposing four to six field beams optimized with tissue heterogeneity corrections to deliver a prescribed dose of 60 Gy in three fractions to at least 95% of the planning target volume, keeping spinal cord dose <10 Gy. The same plans were then regenerated without heterogeneity correction by recalculating previously optimized treatment plans keeping identical beam arrangements, field fluences and monitor units. Compared with heterogeneity corrected plans, the non-corrected plans had lower average minimum, mean, and maximum tumor doses by 13%, 8%, and 6% respectively. The results indicate that tissue heterogeneity is an important determinant of dosimetric optimization of SBRT plans.

Tissue heterogeneity in IMRT dose calculation for lung cancer

The aim of this study was to evaluate the differences in accuracy of dose calculation between 3 commonly used algorithms, the Pencil Beam algorithm (PB), the Anisotropic Analytical Algorithm (AAA), and the Collapsed Cone Convolution Superposition (CCCS) for intensity-modulated radiation therapy (IMRT). The 2D dose distributions obtained with the 3 algorithms were compared on each CT slice pixel by pixel, using the MATLAB code (The MathWorks, Natick, MA) and the agreement was assessed with the γ function. The effect of the differences on dose-volume histograms (DVHs), tumor control, and normal tissue complication probability (TCP and NTCP) were also evaluated, and its significance was quantified by using a nonparametric test. In general PB generates regions of over-dosage both in the lung and in the tumor area. These differences are not always in DVH of the lung, although the Wilcoxon test indicated significant differences in 2 of 4 patients. Disagreement in the lung region was also found when the Γ analysis was performed. The effect on TCP is less important than for NTCP because of the slope of the curve at the level of the dose of interest. The effect of dose calculation inaccuracy is patient-dependent and strongly related to beam geometry and to the localization of the tumor. When multiple intensity-modulated beams are used, the effect of the presence of the heterogeneity on dose distribution may not always be easily predictable.

Analysis of clinical and dosimetric factors associated with severe acute radiation pneumonitis in patients with locally advanced non-small cell lung cancer treated with concurrent chemotherapy and intensity-modulated radiotherapy

Background

To evaluate the association between the clinical, dosimetric factors and severe acute radiation pneumonitis (SARP) in patients with locally advanced non-small cell lung cancer (LANSCLC) treated with concurrent chemotherapy and intensity-modulated radiotherapy (IMRT).

Methods

We analyzed 94 LANSCLC patients treated with concurrent chemotherapy and IMRT between May 2005 and September 2006. SARP was defined as greater than or equal 3 side effects and graded according to Common Terminology Criteria for Adverse Events (CTCAE) version 3.0.

The clinical and dosimetric factors were analyzed. Univariate and multivariate logistic regression analyses were performed to evaluate the relationship between clinical, dosimetric factors and SARP.

Results

Median follow-up was 10.5 months (range 6.5-24). Of 94 patients, 11 (11.7%) developed SARP. Univariate analyses showed that the normal tissue complication probability (NTCP), mean lung dose (MLD), relative volumes of lung receiving more than a threshold dose of 5-60 Gy at increments of 5 Gy (V5-V60), chronic obstructive pulmonary disease (COPD) and Forced Expiratory Volume in the first second (FEV1) were associated with SARP (p < 0.05). In multivariate analysis, NTCP value (p = 0.001) and V10 (p = 0.015) were the most significant factors associated with SARP. The incidences of SARP in the group with NTCP > 4.2% and NTCP ≤4.2% were 43.5% and 1.4%, respectively (p < 0.01). The incidences of SARP in the group with V10 ≤50% and V10 >50% were 5.7% and 29.2%, respectively (p < 0.01).

Conclusions

NTCP value and V10 are the useful indicators for predicting SARP in NSCLC patients treated with concurrent chemotherapy and IMRT.

Analysis of dose-volume histogram parameters for radiation pneumonitis after definitive concurrent chemoradiotherapy for esophageal cancer

PURPOSE

To evaluate dose-volume histogram (DVH) parameters as predictors of radiation pneumonitis (RP) in esophageal cancer patients treated with definitive concurrent chemoradiotherapy.

PATIENTS AND METHODS

Thirty-seven esophageal cancer patients treated with radiotherapy with concomitant chemotherapy consisting of 5-fluorouracil and cisplatin were reviewed. Radiotherapy was delivered at 2 Gy per fraction to a total of 60 Gy. For most of the patients, two weeks of interruption was scheduled after 30 Gy. The percentage of lung volume receiving more than 5-50 Gy in increments of 5 Gy (V5-V50, respectively), and the mean lung dose (MLD) were analyzed.

RESULTS

Ten (27%) patients developed RP of grade 2; 2 (5%), grade 3; 0 (0%), grade 4; and 1 (3%), grade 5. By univariate analysis, all DVH parameters (i.e., V5-V50 and MLD) were significantly associated with grade 2 RP (p < 0.01). The incidences of grade 2 RP were 13%, 33%, and 78% in patients with V20s of 24%, 25-36%, and 37%, respectively. The optimal V20 threshold to predict symptomatic RP was 30.5% according to the receiver operating characteristics curve analysis.

CONCLUSION

DVH parameters were predictors of symptomatic RP and should be considered in the evaluation of treatment planning for esophageal cancer.

Combining multiple models to generate consensus: application to radiation-induced pneumonitis prediction

The fusion of predictions from disparate models has been used in several fields to obtain a more realistic and robust estimate of the "ground truth" by allowing the models to reinforce each other when consensus exists, or, conversely, negate each other when there is no consensus. Fusion has been shown to be most effective when the models have some complementary strengths arising from different approaches. In this work, we fuse the results from four common but methodologically different nonlinear multivariate models (Decision Trees, Neural Networks, Support Vector Machines, Self-Organizing Maps) that were trained to predict radiation-induced pneumonitis risk on a database of 219 lung cancer patients treated with radiotherapy (34 with Grade 2+ postradiotherapy pneumonitis). Each model independently incorporated a small number of features from the available set of dose and nondose patient variables to predict pneumonitis; no two models had all features in common. Fusion was achieved by simple averaging of the predictions for each patient from all four models. Since a model's prediction for a patient can be dependent on the patient training set used to build the model, the average of several different predictions from each model was used in the fusion (predictions were made by repeatedly testing each patient with a model built from different cross-validation training sets that excluded the patient being tested). The area under the receiver operating characteristics curve for the fused cross-validated results was 0.79, with lower variance than the individual component models. From the fusion, five features were extracted as the consensus among all four models in predicting radiation pneumonitis. Arranged in order of importance, the features are (1) chemotherapy; (2) equivalent uniform dose (EUD) for exponent a=1.2 to 3; (3) EUD for a=0.5 to 1.2, lung volume receiving >20-30 Gy; (4) female sex; and (5) squamous cell histology. To facilitate ease of interpretation and prospective use, the fused outcome results for the patients were fitted to a logistic probability function.

Analysis of radiation pneumonitis risk using a generalized Lyman model

PURPOSE

To introduce a version of the Lyman normal-tissue complication probability (NTCP) model adapted to incorporate censored time-to-toxicity data and clinical risk factors and to apply the generalized model to analysis of radiation pneumonitis (RP) risk.

METHODS AND MATERIALS

Medical records and radiation treatment plans were reviewed retrospectively for 576 patients with non-small cell lung cancer treated with radiotherapy. The time to severe (Grade >/=3) RP was computed, with event times censored at last follow-up for patients not experiencing this endpoint. The censored time-to-toxicity data were analyzed using the standard and generalized Lyman models with patient smoking status taken into account.

RESULTS

The generalized Lyman model with patient smoking status taken into account produced NTCP estimates up to 27 percentage points different from the model based on dose-volume factors alone. The generalized model also predicted that 8% of the expected cases of severe RP were unobserved because of censoring. The estimated volume parameter for lung was not significantly different from n = 1, corresponding to mean lung dose.

CONCLUSIONS

NTCP models historically have been based solely on dose-volume effects and binary (yes/no) toxicity data. Our results demonstrate that inclusion of nondosimetric risk factors and censored time-to-event data can markedly affect outcome predictions made using NTCP models.

Developing and evaluating stereotactic lung RT trials: what we should know about the influence of inhomogeneity corrections on dose

Purpose

To investigate the influence of inhomogeneity corrections on stereotactic treatment plans for non-small cell lung cancer and determine the dose delivered to the PTV and OARs.

Materials and methods

For 26 patients with stage-I NSCLC treatment plans were optimized with unit density (UD), an equivalent pathlength algorithm (EPL), and a collapsed-cone (CC) algorithm, prescribing 60 Gy to the PTV. After optimization the first two plans were recalculated with the more accurate CC algorithm. Dose parameters were compared for the three different optimized plans. Dose to the target and OARs was evaluated for the recalculated plans and compared with the planned values.

Results

For the CC algorithm dose constraints for the ratio of the 50% isodose volume and the PTV, and the V_{20 Gy} are harder to fulfill. After recalculation of the UD and EPL plans large variations in the dose to the PTV were observed. For the unit density plans, the dose to the PTV varied from 42.1 to 63.4 Gy for individual patients. The EPL plans all overestimated the PTV dose (average 48.0 Gy). For the lungs, the recalculated V_{20 Gy} was highly correlated to the planned value, and was 12% higher for the UD plans (R² = 0.99), and 15% lower for the EPL plans (R² = 0.96).

Conclusion

Inhomogeneity corrections have a large influence on the dose delivered to the PTV and OARs for SBRT of lung tumors. A simple rescaling of the dose to the PTV is not possible, implicating that accurate dose calculations are necessary for these treatment plans in order to prevent large discrepancies between planned and actually delivered doses to individual patients.

Using patient data similarities to predict radiation pneumonitis via a self-organizing map

This work investigates the use of the self-organizing map (SOM) technique for predicting lung radiation pneumonitis (RP) risk. SOM is an effective method for projecting and visualizing high-dimensional data in a low-dimensional space (map). By projecting patients with similar data (dose and non-dose factors) onto the same region of the map, commonalities in their outcomes can be visualized and categorized. Once built, the SOM may be used to predict pneumonitis risk by identifying the region of the map that is most similar to a patient's characteristics. Two SOM models were developed from a database of 219 lung cancer patients treated with radiation therapy (34 clinically diagnosed with Grade 2+ pneumonitis). The models were: SOM(all) built from all dose and non-dose factors and, for comparison, SOM(dose) built from dose factors alone. Both models were tested using ten-fold cross validation and Receiver Operating Characteristics (ROC) analysis. Models SOM(all) and SOM(dose) yielded ten-fold cross-validated ROC areas of 0.73 (sensitivity/specificity = 71%/68%) and 0.67 (sensitivity/specificity = 63%/66%), respectively. The significant difference between the cross-validated ROC areas of these two models (p < 0.05) implies that non-dose features add important information toward predicting RP risk. Among the input features selected by model SOM(all), the two with highest impact for increasing RP risk were: (a) higher mean lung dose and (b) chemotherapy prior to radiation therapy. The SOM model developed here may not be extrapolated to treatment techniques outside that used in our database, such as several-field lung intensity modulated radiation therapy or gated radiation therapy.

Investigation of the support vector machine algorithm to predict lung radiation-induced pneumonitis

The purpose of this study is to build and test a support vector machine (SVM) model to predict for the occurrence of lung radiation-induced Grade 2+ pneumonitis. SVM is a sophisticated statistical technique capable of separating the two categories of patients (with/without pneumonitis) using a boundary defined by a complex hypersurface. Despite the complexity, the SVM boundary is only minimally influenced by outliers that are difficult to separate. By contrast, the simple hyperplane boundary computed by the more commonly used and related linear discriminant analysis method is heavily influenced by outliers. Two SVM models were built using data from 219 patients with lung cancer treated using radiotherapy (34 diagnosed with pneumonitis). One model (SVM(all)) selected input features from all dose and non-dose factors. For comparison, the other model (SVM(dose)) selected input features only from lung dose-volume factors. Model predictive ability was evaluated using ten-fold cross-validation and receiver operating characteristics (ROC) analysis. For the model SVM(all), the area under the cross-validated ROC curve was 0.76 (sensitivity/specificity = 74%/75%). Compared to the corresponding SVM(dose) area of 0.71 (sensitivity/specificity = 68%/68%), the predictive ability of SVM(all) was improved, indicating that non-dose features are important contributors to separating patients with and without pneumonitis. Among the input features selected by model SVM(all), the two with highest importance for predicting lung pneumonitis were: (a) generalized equivalent uniform doses close to the mean lung dose, and (b) chemotherapy prior to radiotherapy. The model SVM(all) is publicly available via internet access.

A neural network model to predict lung radiation-induced pneumonitis

A feed-forward neural network was investigated to predict the occurrence of lung radiation-induced Grade 2+ pneumonitis. The database consisted of 235 patients with lung cancer treated using radiotherapy, of whom 34 were diagnosed with Grade 2+ pneumonitis at follow-up. The network was constructed using an algorithm that alternately grew and pruned it, starting from the smallest possible network, until a satisfactory solution was found. The weights and biases of the network were computed using the error back-propagation approach. Momentum and variable leaning techniques were used to speed convergence. Using the growing/pruning approach, the network selected features from 66 dose and 27 non-dose variables. During network training, the 235 patients were randomly split into ten groups of approximately equal size. Eight groups were used to train the network, one group was used for early stopping training to prevent overfitting, and the remaining group was used as a test to measure the generalization capability of the network (cross-validation). Using this methodology, each of the ten groups was considered, in turn, as the test group (ten-fold cross-validation). For the optimized network constructed with input features selected from dose and non-dose variables, the area under the receiver operating characteristics (ROC) curve for cross-validated testing was 0.76 (sensitivity: 0.68, specificity: 0.69). For the optimized network constructed with input features selected only from dose variables, the area under the ROC curve for cross-validation was 0.67 (sensitivity: 0.53, specificity: 0.69). The difference between these two areas was statistically significant (p = 0.020), indicating that the addition of non-dose features can significantly improve the generalization capability of the network. A network for prospective testing was constructed with input features selected from dose and non-dose variables (all data were used for training). The optimized network architecture consisted of six input nodes (features), four hidden nodes, and one output node. The six input features were: lung volume receiving > 16 Gy (V16), generalized equivalent uniform dose (gEUD) for the exponent a = 1 (mean lung dose), gEUD for the exponent a = 3.5, free expiratory volume in 1 s (FEV1), diffusion capacity of carbon monoxide (DLCO%), and whether or not the patient underwent chemotherapy prior to radiotherapy. The significance of each input feature was individually evaluated by omitting it during network training and gauging its impact by the consequent deterioration in cross-validated ROC area. With the exception of FEV1 and whether or not the patient underwent chemotherapy prior to radiotherapy, all input features were found to be individually significant (p < 0.05). The network for prospective testing is publicly available via internet access.

Predicting lung radiotherapy-induced pneumonitis using a model combining parametric Lyman probit with nonparametric decision trees

PURPOSE

To develop and test a model to predict for lung radiation-induced Grade 2+ pneumonitis.

METHODS AND MATERIALS

The model was built from a database of 234 lung cancer patients treated with radiotherapy (RT), of whom 43 were diagnosed with pneumonitis. The model augmented the predictive capability of the parametric dose-based Lyman normal tissue complication probability (LNTCP) metric by combining it with weighted nonparametric decision trees that use dose and nondose inputs. The decision trees were sequentially added to the model using a "boosting" process that enhances the accuracy of prediction. The model's predictive capability was estimated by 10-fold cross-validation. To facilitate dissemination, the cross-validation result was used to extract a simplified approximation to the complicated model architecture created by boosting. Application of the simplified model is demonstrated in two example cases.

RESULTS

The area under the model receiver operating characteristics curve for cross-validation was 0.72, a significant improvement over the LNTCP area of 0.63 (p = 0.005). The simplified model used the following variables to output a measure of injury: LNTCP, gender, histologic type, chemotherapy schedule, and treatment schedule. For a given patient RT plan, injury prediction was highest for the combination of pre-RT chemotherapy, once-daily treatment, female gender and lowest for the combination of no pre-RT chemotherapy and nonsquamous cell histologic type. Application of the simplified model to the example cases revealed that injury prediction for a given treatment plan can range from very low to very high, depending on the settings of the nondose variables.

CONCLUSIONS

Radiation pneumonitis prediction was significantly enhanced by decision trees that added the influence of nondose factors to the LNTCP formulation.

The impact of breathing motion versus heterogeneity effects in lung cancer treatment planning

The purpose of this study is to investigate the effects of tissue heterogeneity and breathing-induced motion/deformation on conformal treatment planning for pulmonary tumors and to compare the magnitude and the clinical importance of changes induced by these effects. Treatment planning scans were acquired at normal exhale/inhale breathing states for fifteen patients. The internal target volume (ITV) was defined as the union of exhale and inhale gross tumor volumes uniformly expanded by 5 mm. Anterior/posterior opposed beams (AP/PA) and three-dimensional (3D)-conformal plans were designed using the unit-density exhale ("static") dataset. These plans were further used to calculate (a) density-corrected ("heterogeneous") static dose and (b) heterogeneous cumulative dose, including breathing deformations. The DPM Monte Carlo code was used for dose computations. For larger than coin-sized tumors, relative to unit-density plans, tumor and lung doses increased in the heterogeneity-corrected plans. In comparing cumulative and static plans, larger normal tissue complication probability changes were observed for tumors with larger motion amplitudes and uncompensated breathing-induced hot/cold spots in lung. Accounting for tissue heterogeneity resulted in average increases of 9% and 7% in mean lung dose (MLD) for the 6 MV and 15 MV photon beams, respectively. Breathing-induced effects resulted in approximately 1% and 2% average decreases in MLD from the static value, for the 6 and 15 MV photon beams, respectively. The magnitude of these effects was not found to correlate with the treatment plan technique, i.e., AP/PA versus 3D-CRT. Given a properly designed ITV, tissue heterogeneity effects are likely to have a larger clinical significance on tumor and normal lung treatment evaluation metrics than four-dimensional respiratory-induced changes.