LASSO NTCP predictors for the incidence of xerostomia in patients with head and neck squamous cell carcinoma and nasopharyngeal carcinoma

Artificial intelligence-based predictive model for guidance on treatment strategy selection in oral and maxillofacial surgery

Application of deep learning (DL) and machine learning (ML) is rapidly increasing in the medical field. DL is gaining significance for medical image analysis, particularly, in oral and maxillofacial surgeries. Owing to the ability to accurately identify and categorize both diseased and normal soft- and hard-tissue structures, DL has high application potential in the diagnosis and treatment of tumors and in orthognathic surgeries. Moreover, DL and ML can be used to develop prediction models that can aid surgeons to assess prognosis by analyzing the patient's medical history, imaging data, and surgical records, develop more effective treatment strategies, select appropriate surgical modalities, and evaluate the risk of postoperative complications. Such prediction models can play a crucial role in the selection of treatment strategies for oral and maxillofacial surgeries. Their practical application can improve the utilization of medical staff, increase the treatment accuracy and efficiency, reduce surgical risks, and provide an enhanced treatment experience to patients. However, DL and ML face limitations, such as data drift, unstable model results, and vulnerable social trust. With the advancement of social concepts and technologies, the use of these models in oral and maxillofacial surgery is anticipated to become more comprehensive and extensive.

Utilization of machine learning methods for prediction of acute and late rectal toxicity due to curative prostate radiotherapy

OBJECTIVE

Rectal toxicity is one of the primary dose-limiting side effects of prostate cancer radiotherapy, and consequential impairment on quality of life in these patients with long survival is an important problem. In this study, we aimed to evaluate the possibility of predicting rectal toxicity with artificial intelligence model which was including certain dosimetric parameters.

MATERIALS AND METHODS

One hundred and thirty-seven patients with a diagnosis of prostate cancer who received curative radiotherapy for prostate +/- pelvic lymphatics were included in the study. The association of the clinical data and dosimetric data between early and late rectal toxicity reported during follow-up was evaluated. The sample size was increased to 274 patients by synthetic data generation method. To determine suitable models, 15 models were studied with machine learning algorithms using Python 2.3, Pycaret library. Random forest classifier was used with to detect active variables.

RESULTS

The area under the curve and accuracy were found to be 0.89-0.97 and 95%-99%, respectively, with machine learning algorithms. The sensitivity values for acute and toxicity were found to be 0.95 and 0.99, respectively.

CONCLUSION

Early or late rectal toxicity can be predicted with a high probability via dosimetric and physical data and machine learning algorithms of patients who underwent prostate +/- pelvic radiotherapy. The fact that rectal toxicity can be predicted before treatment, which may result in limiting the dose and duration of treatment, makes us think that artificial intelligence can enter our daily practice in a short time in this sense.

Predict nutrition-related adverse outcomes in head and neck cancer patients undergoing radiotherapy: A systematic review

BACKGROUND

Acute nutrition-related adverse outcomes are common in head and neck cancer patients undergoing radiotherapy. Predictive models can assist in identifying high-risk patients to enable targeted intervention. We aimed to systematically evaluate predictive models for predicting severe acute nutritional symptoms, insufficient intake, tube feeding, sarcopenia, and weight loss.

METHODS

We searched PubMed, Web of Science, EBSCO, Embase, WanFang, CNKI, and SinoMed. We selected studies developing predictive models for the aforementioned outcomes. Data were extracted using a predefined checklist. Risk of bias and applicability assessment were assessed using the Prediction model Risk of Bias Assessment Tool. A narrative synthesis was conducted to summarize the model characteristics, risk of bias, and performance.

RESULTS

A total of 2941 studies were retrieved and 19 were included. Study outcome measure were different symptoms (n = 11), weight loss (n = 5), tube feeding (n = 3), and symptom or tube feeding (n = 1). Predictive factors mainly encompassed sociodemographic data, disease-related data, and treatment-related data. Seventeen studies reported area under the curve or C-index values ranging from 0.610 to 0.96, indicating moderate to good predictive performance. However, candidate predictors were incomplete, outcome measures were diverse, and the risk of bias was high. Most of them used traditional model development methods, and only two used machine learning.

CONCLUSIONS

Most current models showed moderate to good predictive performance. However, predictors are incomplete, outcome are inconsistent, and the risk of bias is high. Clinicians could carefully select the models with better model performance from the available models according to their actual conditions. Future research should include comprehensive and modifiable indicators and prioritize well-designed and reported studies for model development.

Development of a risk prediction model for radiation dermatitis following proton radiotherapy in head and neck cancer using ensemble machine learning

PURPOSE

This study aims to develop an ensemble machine learning-based (EML-based) risk prediction model for radiation dermatitis (RD) in patients with head and neck cancer undergoing proton radiotherapy, with the goal of achieving superior predictive performance compared to traditional models.

MATERIALS AND METHODS

Data from 57 head and neck cancer patients treated with intensity-modulated proton therapy at Kaohsiung Chang Gung Memorial Hospital were analyzed. The study incorporated 11 clinical and 9 dosimetric parameters. Pearson's correlation was used to eliminate highly correlated variables, followed by feature selection via LASSO to focus on potential RD predictors. Model training involved traditional logistic regression (LR) and advanced ensemble methods such as Random Forest and XGBoost, which were optimized through hyperparameter tuning.

RESULTS

Feature selection identified six key predictors, including smoking history and specific dosimetric parameters. Ensemble machine learning models, particularly XGBoost, demonstrated superior performance, achieving the highest AUC of 0.890. Feature importance was assessed using SHAP (SHapley Additive exPlanations) values, which underscored the relevance of various clinical and dosimetric factors in predicting RD.

CONCLUSION

The study confirms that EML methods, especially XGBoost with its boosting algorithm, provide superior predictive accuracy, enhanced feature selection, and improved data handling compared to traditional LR. While LR offers greater interpretability, the precision and broader applicability of EML make it more suitable for complex medical prediction tasks, such as predicting radiation dermatitis. Given these advantages, EML is highly recommended for further research and application in clinical settings.

Dosimetric parameters predict radiation-induced temporal lobe necrosis in nasopharyngeal carcinoma patients: A systematic review and meta-analysis

This systematic review examines the role of dosimetric parameters in predicting temporal lobe necrosis (TLN) risk in nasopharyngeal carcinoma (NPC) patients treated with three-dimensional conformal RT (3D-CRT), intensity-modulated radiation therapy (IMRT) and volumetric modulated arc therapy (VMAT). TLN is a serious late complication that can adversely affect the quality of life of NPC patients. Understanding the relationship between dosimetric parameters and TLN can guide treatment planning and minimize radiation-related complications. A comprehensive search identified relevant studies published up to July 2023. Studies reporting on dosimetric parameters and TLN in NPC patients undergoing 3D-CRT, IMRT, and VMAT were included. TLN incidence, follow-up duration, and correlation with dosimetric parameters of the temporal lobe were analyzed. The review included 30 studies with median follow-up durations ranging from 28 to 110 months. The crude incidence of TLN varied from 2.3 % to 47.3 % and the average crude incidence of TLN is approximately 14 %. Dmax and D1cc emerged as potential predictors of TLN in 3D-CRT and IMRT-treated NPC patients. Threshold values of >72 Gy for Dmax and >62 Gy for D1cc were associated with increased TLN risk. However, other factors should also be considered, including host characteristics, tumor-specific features and therapeutic factors. In conclusion, this systematic review highlights the significance of dosimetric parameters, particularly Dmax and D1cc, in predicting TLN risk in NPC patients undergoing 3D-CRT, IMRT, and VMAT. The findings provide valuable insights that can help in developing optimal treatment planning strategies and contribute to the development of clinical guidelines in this field.

Using meta-analysis and CNN-NLP to review and classify the medical literature for normal tissue complication probability in head and neck cancer

PURPOSE

The study aims to enhance the efficiency and accuracy of literature reviews on normal tissue complication probability (NTCP) in head and neck cancer patients using radiation therapy. It employs meta-analysis (MA) and natural language processing (NLP).

MATERIAL AND METHODS

The study consists of two parts. First, it employs MA to assess NTCP models for xerostomia, dysphagia, and mucositis after radiation therapy, using Python 3.10.5 for statistical analysis. Second, it integrates NLP with convolutional neural networks (CNN) to optimize literature search, reducing 3256 articles to 12. CNN settings include a batch size of 50, 50-200 epoch range and a 0.001 learning rate.

RESULTS

The study's CNN-NLP model achieved a notable accuracy of 0.94 after 200 epochs with Adamax optimization. MA showed an AUC of 0.67 for early-effect xerostomia and 0.74 for late-effect, indicating moderate to high predictive accuracy but with high variability across studies. Initial CNN accuracy of 66.70% improved to 94.87% post-tuning by optimizer and hyperparameters.

CONCLUSION

The study successfully merges MA and NLP, confirming high predictive accuracy for specific model-feature combinations. It introduces a time-based metric, words per minute (WPM), for efficiency and highlights the utility of MA and NLP in clinical research.

World Workshop on Oral Medicine VIII: Development of a Core Outcome Set for Dry Mouth: A Systematic Review of Outcome Domains for Xerostomia

OBJECTIVE

The purpose of this study was to identify all outcome domains used in clinical studies of xerostomia, that is, subjective sensation of dry mouth. This study is part of the extended project "World Workshop on Oral Medicine Outcomes Initiative for the Direction of Research" to develop a core outcome set for dry mouth.

STUDY DESIGN

A systematic review was performed on MEDLINE, EMBASE, CINAHL, and Cochrane Central Register of Controlled Trials databases. All clinical and observational studies that assessed xerostomia in human participants from 2001 to 2021 were included. Information on outcome domains was extracted and mapped to the Core Outcome Measures in Effectiveness Trials taxonomy. Corresponding outcome measures were summarized.

RESULTS

From a total of 34,922 records retrieved, 688 articles involving 122,151 persons with xerostomia were included. There were 16 unique outcome domains and 166 outcome measures extracted. None of these domains or measures were consistently used across all the studies. The severity of xerostomia and physical functioning were the 2 most frequently assessed domains.

CONCLUSION

There is considerable heterogeneity in outcome domains and measures reported in clinical studies of xerostomia. This highlights the need for harmonization of dry mouth assessment to enhance comparability across studies and facilitate the synthesis of robust evidence for managing patients with xerostomia.

Integrated radiomics, dose-volume histogram criteria and clinical features for early prediction of saliva amount reduction after radiotherapy in nasopharyngeal cancer patients

PURPOSE

Previously, the evaluation of xerostomia depended on subjective grading systems, rather than the accurate saliva amount reduction. Our aim was to quantify acute xerostomia with reduced saliva amount, and apply radiomics, dose-volume histogram (DVH) criteria and clinical features to predict saliva amount reduction by machine learning techniques.

MATERIAL AND METHODS

Computed tomography (CT) of parotid glands, DVH, and clinical data of 52 patients were collected to extract radiomics, DVH criteria and clinical features, respectively. Firstly, radiomics, DVH criteria and clinical features were divided into 3 groups for feature selection, in order to alleviate the masking effect of the number of features in different groups. Secondly, the top features in the 3 groups composed integrated features, and features selection was performed again for integrated features. In this study, feature selection was used as a combination of eXtreme Gradient Boosting (XGBoost) and SHapley Additive exPlanations (SHAP) to alleviate multicollinearity. Finally, 6 machine learning techniques were used for predicting saliva amount reduction. Meanwhile, top radiomics features were modeled using the same machine learning techniques for comparison.

RESULT

17 integrated features (10 radiomics, 4 clinical, 3 DVH criteria) were selected to predict saliva amount reduction, with a mean square error (MSE) of 0.6994 and a R² score of 0.9815. Top 17 and 10 selected radiomics features predicted saliva amount reduction, with MSE of 0.7376, 0.7519, and R² score of 0.9805, 0.9801, respectively.

CONCLUSION

With the same number of features, integrated features (radiomics + DVH criteria + clinical) performed better than radiomics features alone. The important DVH criteria and clinical features mainly included, white blood cells (WBC), parotid_glands_Dmax, Age, parotid_glands_V15, hemoglobin (Hb), BMI and parotid_glands_V45.

Screening Key Genes and Biological Pathways in Nasopharyngeal Carcinoma by Integrated Bioinformatics Analysis

The purpose of this study was to identify the hub genes and biological pathways of nasopharyngeal carcinoma (NPC) through bioinformatics analysis and potential new therapeutic targets. In this study, three datasets were downloaded from the Gene Expression Omnibus (GEO), and differentially expressed genes (DEGs) between NPC and normal tissues were analyzed using the GEO2R online tool. Volcano and heat maps of the DEGs were visualized using the hiplot database. Gene ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses of the upregulated and downregulated DEGs were performed using the DAVID database. Finally, we established a protein-protein interaction (PPI) network using the STRING database and showed the differential expression of hub genes between the normal and tumor tissues. In all, 109,371,221 upregulated DEGs and 139,226,520 downregulated DEGs were obtained in datasets GSE40290, GSE61218, and GSE53819, respectively, and 18 common differential genes, named co-DEGs, were screened in the three datasets. The most abundant biological GO terms of the co-DEGs were inflammatory response et al. The KEGG pathway enrichment analysis showed that co-DEGs mainly participated in the interleukin (IL)-17 signaling pathway et al. Finally, we identified four hub genes using PPI analysis and observed that three of them were highly expressed in tumor tissues. In this study, the hub genes of NPC, such as PTGS2, and pathways such as IL-17 signaling, were identified through bioinformatics analysis, which may be potential new therapeutic targets for NPC.

Cluster model incorporating heterogeneous dose distribution of partial parotid irradiation for radiotherapy induced xerostomia prediction with machine learning methods

PURPOSE

A cluster model incorporating heterogeneous dose distribution within the parotid gland was developed and validated retrospectively for radiotherapy (RT) induced xerostomia prediction with machine learning (ML) techniques.

METHODS

Sixty clusters were obtained at 1 Gy step size with threshold doses ranging from 1 to 60 Gy, for each of the enrolled 155 patients with HNC from three institutions. Feature clusters were selected with the neighborhood component analysis (NCA) and subsequently fed into four supervised ML models for xerostomia prediction comparison: support vector machines (SVM), k-nearest neighbor (kNN), naïve Bayes (NB), and random forest (RF). The predictive performance of each model was evaluated using cross validation resampling with the area-under-the-curves (AUC) of the receiver-operating-characteristic (ROC). The xerostomia predicting capacity using testing data was assessed with accuracy, sensitivity, and specificity for these models and three cluster connectivity choices. Mean dose based logistic regression served as the benchmark for evaluation.

RESULTS

Feature clusters identified by NCA fell in three threshold dose ranges: 5-15Gy, 25-35Gy, and 45-50Gy. Mean dose predictive power was 15% lower than that of the cluster model using the logistic regression classifier. Model validation demonstrated that kNN model outperformed slightly other three models but no substantial difference was observed. Applying the fine-tuned models to testing data yielded that the mean accuracy from SVM, kNN and NB models were between 0.68 and 0.7 while that of RF was ∼0.6. SVM model yielded the best sensitivity (0.76) and kNN model delivered consistent sensitivity and specificity. This is consistent with cross validation. Clusters calculated with three connectivity choices exhibited minimally different predictions.

CONCLUSION

Compared to mean dose, the proposed cluster model has shown its improvement as the xerostomia predictor. When combining with ML techniques, it could provide a clinically useful tool for xerostomia prediction and facilitate decision making during radiotherapy planning for patients with HNC.

A Risk Prediction Model by LASSO for Radiation-Induced Xerostomia in Patients With Nasopharyngeal Carcinoma Treated With Comprehensive Salivary Gland-Sparing Helical Tomotherapy Technique

Objective

This study aimed to develop a least absolute shrinkage and selection operator (LASSO)-based multivariable normal tissue complication probability (NTCP) model to predict radiation-induced xerostomia in patients with nasopharyngeal carcinoma (NPC) treated with comprehensive salivary gland–sparing helical tomotherapy technique.

Methods and Materials

LASSO with the extended bootstrapping technique was used to build multivariable NTCP models to predict factors of patient-reported xerostomia relieved by 50% and 80% compared with the level at the end of radiation therapy within 1 year and 2 years, R50-1year and R80-2years, in 203 patients with NPC. The model assessment was based on 10-fold cross-validation and the area under the receiver operating characteristic curve (AUC).

Results

The prediction model by LASSO with 10-fold cross-validation showed that radiation-induced xerostomia recovery could be predicted by prognostic factors of R50-1year (age, gender, T stage, UICC/AJCC stage, parotid Dmean, oral cavity Dmean, and treatment options) and R80-2years (age, gender, T stage, UICC/AJCC stage, oral cavity Dmean, N stage, and treatment options). These prediction models also demonstrated a good performance by the AUC.

Conclusion

The prediction models of R50-1year and R80-2years by LASSO with 10-fold cross-validation were recommended to validate the NTCP model before comprehensive salivary gland–sparing radiation therapy in patients with NPC.

Evaluating the Risk Factors of Post Inflammatory Hyperpigmentation Complications with Nd-YAG Laser Toning Using LASSO-Based Algorithm

The neodymium-doped yttrium aluminum garnet (Nd-YAG) laser is used for removal of pigmented skin patches and rejuvenation of skin. However, complications such as hyperpigmentation, hypopigmentation, and petechiae can occur after frequent treatments. Therefore, identifying the risk factors for such complications is important. The development of a multivariable logistic regression model with least absolute shrinkage and selection operator (LASSO) is needed to provide valid predictions about the incidence of post inflammatory hyperpigmentation complication probability (PIHCP) among patients treated with Nd-YAG laser toning. A total of 125 female patients undergoing laser toning therapy between January 2014 and January 2016 were examined for post-inflammatory hyperpigmentation (PIH) complications. Factor analysis was performed using 15 potential predictive risk factors of PIH determined by a physician. The LASSO algorithm with cross-validation was used to select the optimal number of predictive risk factors from the potential factors for a multivariate logistic regression PIH complication model. The optimal number of predictive risk factors for the model was five: immediate endpoints of laser (IEL), α-hydroxy acid (AHA) peels, Fitzpatrick skin phototype (FSPT), acne, and melasma. The area under the receiver operating characteristic curve (AUC) was 0.79 (95% CI, 0.70–0.88) in the optimal model. The overall performance of the LASSO-based PIHCP model was satisfactory based on the AUC, Omnibus, Nagelkerke R2, and Hosmer–Lemeshow tests. This predictive risk factor model is useful to further optimize laser toning treatment related to PIH. The LASSO-based PIHCP model could be useful for decision-making.

Developing a multivariable normal tissue complication probability model to predict late rectal bleeding following intensity-modulated radiation therapy

Purpose: To develop a multivariable normal tissue complication probability (NTCP) model to predict moderate to severe late rectal bleeding following intensity-modulated radiation therapy (IMRT).

Methods and materials: Sixty-eight patients with localized prostate cancer treated by IMRT from 2008 to 2011 were enrolled. The median follow-up time was 56 months. According to the criteria of D'Amico risk classifications, there were 9, 20 and 39 patients in low, intermediate and high-risk groups, respectively. Forty-two patients were combined with androgen deprivation therapy. Fifteen patients had suffered from grade 2 or more (grade 2+) late rectal bleeding. The numbers of predictors for a multivariable logistic regression NTCP model were determined by the least absolute shrinkage and selection operator (LASSO).

Results: The most important predictors for late rectal bleeding ranked by LASSO were platelet count, risk group and the relative volume of rectum receiving at least 65 Gy (V₆₅). The NTCP model of grade 2+ rectal bleeding was as follows: S = -17.49 + Platelets (1000/μL) * (-0.025) + Risk group * Corresponding coefficient (low-risk group = 0; intermediate-risk group = 19.07; high-risk group = 20.41) + V₆₅ * 0.045.

Conclusions: A LASSO-based multivariable NTCP model comprising three important predictors (platelet count, risk group and V₆₅) was established to predict the incidence of grade 2+ late rectal bleeding after IMRT.

Developing of predictive models for pneumonitis with forward variable selection and LASSO logistic model for breast cancer patients treated with 3D-CRT

Purpose: To develop a multiple logistic regression model as normal tissue complication probability model by least absolute shrinkage and selection operator (LASSO) technique in breast cancer patients treated with three-dimensional conformal radiation therapy (3D-CRT), we focused on the changes of pulmonary function tests to achieve the optimal predictive parameters for the occurrence of symptomatic radiation pneumonitis (SRP).

Materials and methods: Dosimetric and spirometry data of 60 breast cancer patients were analyzed. Pulmonary function tests were done before RT, after completion of RT, 3, and 6 months after RT. Multiple logistic regression model was used to obtain the effective predictive parameters. Forward selection method was applied in NTCP model to determine the effective risk factors from obtained different parameters.

Results: Symptomatic radiation pneumonitis was observed in five patients. Significant changes in pulmonary parameters have been observed at six months after RT. The parameters of mean lung dose (MLD), bridge separation (BS), mean irradiated lung volume (ILVmean), and the percentage of the ipsilateral lung volume that received dose of 20 Gy (IV20) introduced as risk factors using the LASSO technique for SRP in a multiple normal tissue complication probability model in breast cancer patients treated with 3D-CRT. The BS, central lung distance (CLD) and ILV in tangential field have obtained as 23.5 (20.9-26.0) cm, 2.4 (1.5-3.3) cm, and 12.4 (10.6-14.3) % of lung volume in radiation field in patients without pulmonary complication, respectively.

Conclusion: The results showed that if BS, CLD, and ILV are more than 23 cm, 2 cm, and 12%, respectively, so incidence of SRP in the patients will be considerable. Our multiple NTCP LASSO model for breast cancer patients treated with 3D-CRT showed that in order to have minimum probability of SRP occurrence, parameters of BS, IV20, ILV and especially MLD would be kept in minimum levels. Considering dose-volume histogram, the mean lung dose factor is most important parameter which minimizing it in treatment planning, minimizes the probability of SRP and consequently improves the quality of life in breast cancer patients.

Machine Learning Methods Uncover Radiomorphologic Dose Patterns in Salivary Glands that Predict Xerostomia in Patients with Head and Neck Cancer

Purpose

Patients with head-and-neck cancer (HNC) may experience xerostomia after radiation therapy (RT), which leads to compromised quality of life. The purpose of this study is to explore how the spatial pattern of radiation dose (radiomorphology) in the major salivary glands influences xerostomia in patients with HNC.

Methods and materials

A data-driven approach using spatially explicit dosimetric predictors, voxel dose (ie, actual radiation dose in voxels in parotid glands [PG] and submandibular glands [SMG]) was used to predict whether patients would develop xerostomia 3 months after RT. Using planned radiation dose data and other nondose covariates including baseline xerostomia grade of 427 patients with HNC in our database, the machine learning methods were used to investigate the influence of dose patterns across subvolumes in PG and SMG on xerostomia.

Results

Of the 3 supervised learning methods studied, ridge logistic regression yielded the best predictive performance. Ridge logistic regression was also preferred to evaluate the influence pattern of highly correlated dose on xerostomia, which showed a discriminative pattern of influence of doses in the PG and SMG on xerostomia. Moreover, the superior–anterior portion of the contralateral PG and medial portion of the ipsilateral PG were determined to be the most influential regions regarding dose effect on xerostomia. The area under the receiver operating characteristic curve from a 10-fold cross-validation was 0.70 ± 0.04.

Conclusions

Radiomorphology, combined with machine learning methods, is able to suggest patterns of dose in PG and SMG that are the most influential on xerostomia. The influence pattern identified by this data-driven approach and machine learning methods may help improve RT treatment planning and reduce xerostomia after treatment.

Relationships among patient characteristics, irradiation treatment planning parameters, and treatment toxicity of acute radiation dermatitis after breast hybrid intensity modulation radiation therapy

To evaluate the relationships among patient characteristics, irradiation treatment planning parameters, and treatment toxicity of acute radiation dermatitis (RD) after breast hybrid intensity modulation radiation therapy (IMRT). The study cohort consisted of 95 breast cancer patients treated with hybrid IMRT. RD grade ≥2 (2⁺) toxicity was defined as clinically significant. Patient characteristics and the irradiation treatment planning parameters were used as the initial candidate factors. Prognostic factors were identified using the least absolute shrinkage and selection operator (LASSO)-based normal tissue complication probability (NTCP) model. A univariate cut-off dose NTCP model was developed to find the dose-volume limitation. Fifty-two (54.7%) of ninety-five patients experienced acute RD grade 2⁺ toxicity. The volume of skin receiving a dose >35 Gy (V₃₅) was the most significant dosimetric predictor associated with RD grade 2⁺ toxicity. The NTCP model parameters for V_35Gy were TV₅₀ = 85.7 mL and γ₅₀ = 0.77, where TV₅₀ was defined as the volume corresponding to a 50% incidence of complications, and γ₅₀ was the normalized slope of the volume-response curve. Additional potential predictive patient characteristics were energy and surgery, but the results were not statistically significant.

To ensure a better quality of life and compliance for breast hybrid IMRT patients, the skin volume receiving a dose >35 Gy should be limited to <85.7 mL to keep the incidence of RD grade 2⁺ toxicities below 50%. To avoid RD toxicity, the volume of skin receiving a dose >35 Gy should follow sparing tolerance and the inherent patient characteristics should be considered.

Data Analysis in Radiotherapy Treatments

Radiotherapy is one of the main cancer treatments available today, together with chemotherapy and surgery. Radiotherapy treatments have to be planned for each patient in an individualized manner. The knowledge acquired from one single treatment can be used to improve the treatment planning and outcome of several other patients. In the last years, attention has been drawn to the added value of using data analysis for radiotherapy treatment planning, prediction of treatment outcomes, survival analysis and quality assurance. In this article, existing literature is reviewed.

Propensity-score-matched evaluation of the incidence of radiation pneumonitis and secondary cancer risk for breast cancer patients treated with IMRT/VMAT

Propensity score matching evaluates the treatment incidence of radiation-induced pneumonitis (RP) and secondary cancer risk (SCR) after intensity-modulated radiotherapy (IMRT) and volumetric-modulated arc therapy (VMAT) for breast cancer patients. Of 32 patients treated with IMRT and 58 who received VMAT were propensity matched in a 1:1 ratio. RP and SCR were evaluated as the endpoints of acute and chronic toxicity, respectively. Self-fitted normal tissue complication probability (NTCP) parameter values were used to analyze the risk of RP. SCRs were evaluated using the preferred Schneider's parameterization risk models. The dosimetric parameter of the ipsilateral lung volume receiving 40 Gy (IV₄₀) was selected as the dominant risk factor for the RP NTCP model. The results showed that the risks of RP and NTCP, as well as that of SCR of the ipsilateral lung, were slightly lower than the values in patients treated with VMAT versus IMRT (p ≤ 0.01). However, the organ equivalent dose and excess absolute risk values in the contralateral lung and breast were slightly higher with VMAT than with IMRT (p ≤ 0.05). When compared to IMRT, VMAT is a rational radiotherapy option for breast cancer patients, based on its reduced potential for inducing secondary malignancies and RP complications.

Statistical-learning strategies generate only modestly performing predictive models for urinary symptoms following external beam radiotherapy of the prostate: A comparison of conventional and machine-learning methods

PURPOSE

Given the paucity of available data concerning radiotherapy-induced urinary toxicity, it is important to ensure derivation of the most robust models with superior predictive performance. This work explores multiple statistical-learning strategies for prediction of urinary symptoms following external beam radiotherapy of the prostate.

METHODS

The performance of logistic regression, elastic-net, support-vector machine, random forest, neural network, and multivariate adaptive regression splines (MARS) to predict urinary symptoms was analyzed using data from 754 participants accrued by TROG03.04-RADAR. Predictive features included dose-surface data, comorbidities, and medication-intake. Four symptoms were analyzed: dysuria, haematuria, incontinence, and frequency, each with three definitions (grade ≥ 1, grade ≥ 2 and longitudinal) with event rate between 2.3% and 76.1%. Repeated cross-validations producing matched models were implemented. A synthetic minority oversampling technique was utilized in endpoints with rare events. Parameter optimization was performed on the training data. Area under the receiver operating characteristic curve (AUROC) was used to compare performance using sample size to detect differences of ≥0.05 at the 95% confidence level.

RESULTS

Logistic regression, elastic-net, random forest, MARS, and support-vector machine were the highest-performing statistical-learning strategies in 3, 3, 3, 2, and 1 endpoints, respectively. Logistic regression, MARS, elastic-net, random forest, neural network, and support-vector machine were the best, or were not significantly worse than the best, in 7, 7, 5, 5, 3, and 1 endpoints. The best-performing statistical model was for dysuria grade ≥ 1 with AUROC ± standard deviation of 0.649 ± 0.074 using MARS. For longitudinal frequency and dysuria grade ≥ 1, all strategies produced AUROC>0.6 while all haematuria endpoints and longitudinal incontinence models produced AUROC<0.6.

CONCLUSIONS

Logistic regression and MARS were most likely to be the best-performing strategy for the prediction of urinary symptoms with elastic-net and random forest producing competitive results. The predictive power of the models was modest and endpoint-dependent. New features, including spatial dose maps, may be necessary to achieve better models.

A Nomogram to predict parotid gland overdose in head and neck IMRT

Purposes

To generate a nomogram to predict parotid gland (PG) overdose and to quantify the dosimetric benefit of weekly replanning based on its findings, in the context of intensity-modulated radiotherapy (IMRT) for locally-advanced head and neck carcinoma (LAHNC).

Material and methods

Twenty LAHNC patients treated with radical IMRT underwent weekly computed tomography (CT) scans during IMRT. The cumulated PG dose was estimated by elastic registration. Early predictors of PG overdose (cumulated minus planned doses) were identified, enabling a nomogram to be generated from a linear regression model. Its performance was evaluated using a leave-one-out method. The benefit of weekly replanning was then estimated for the nomogram-identified PG overdose patients.

Results

Clinical target volume 70 (CTV70) and the mean PG dose calculated from the planning and first weekly CTs were early predictors of PG overdose, enabling a nomogram to be generated. A mean PG overdose of 2.5Gy was calculated for 16 patients, 14 identified by the nomogram. All patients with PG overdoses >1.5Gy were identified. Compared to the cumulated delivered dose, weekly replanning of these 14 targeted patients enabled a 3.3Gy decrease in the mean PG dose.

Conclusion

Based on the planning and first week CTs, our nomogram allowed the identification of all patients with PG overdoses >2.5Gy to be identified, who then benefitted from a final 4Gy decrease in mean PG overdose by means of weekly replanning.

Big Data and machine learning in radiation oncology: State of the art and future prospects

Precision medicine relies on an increasing amount of heterogeneous data. Advances in radiation oncology, through the use of CT Scan, dosimetry and imaging performed before each fraction, have generated a considerable flow of data that needs to be integrated. In the same time, Electronic Health Records now provide phenotypic profiles of large cohorts of patients that could be correlated to this information. In this review, we describe methods that could be used to create integrative predictive models in radiation oncology. Potential uses of machine learning methods such as support vector machine, artificial neural networks, and deep learning are also discussed.

LASSO-based NTCP model for radiation-induced temporal lobe injury developing after intensity-modulated radiotherapy of nasopharyngeal carcinoma

We investigated the incidence of temporal lobe injury (TLI) in 132 nasopharyngeal carcinoma (NPC) patients who had undergone intensity-modulated radiotherapy (IMRT) in our hospital between March 2005 and November 2009; and identified significant dosimetric predictors of TLI development. Contrast-enhanced lesions or cysts in the temporal lobes, as detected by magnetic resonance imaging (MRI), were regarded as radiation-induced TLIs. We used the least absolute shrinkage and selection operator (LASSO) method to select D_max (the maximum point dose) and the D_1cc (the top dose delivered to a 1-mL volume) from 15 dose-volume-histogram-associated and four clinically relevant candidate factors; the D_max and the D_1cc were the most significant predictors of TLI development. We drew dose-response curves for D_max and D_1cc. The tolerance dose (TD) for the 5% and 50% probabilities of TLI development were 69.0 ± 1.6 and 82.1 ± 2.4 Gy for D_max and 62.8 ± 2.2 and 80.9 ± 3.4 Gy for D_1cc, respectively. The incidence of TLI in NPC patients after IMRT was higher than expected because the therapeutic window is narrow. High-quality longitudinal studies are needed to gain further insight into the complex spatiotemporal effects of non-uniform irradiation on TLI development in NPC patients.

Machine Learning Approaches for Predicting Radiation Therapy Outcomes: A Clinician's Perspective

Radiation oncology has always been deeply rooted in modeling, from the early days of isoeffect curves to the contemporary Quantitative Analysis of Normal Tissue Effects in the Clinic (QUANTEC) initiative. In recent years, medical modeling for both prognostic and therapeutic purposes has exploded thanks to increasing availability of electronic data and genomics. One promising direction that medical modeling is moving toward is adopting the same machine learning methods used by companies such as Google and Facebook to combat disease. Broadly defined, machine learning is a branch of computer science that deals with making predictions from complex data through statistical models. These methods serve to uncover patterns in data and are actively used in areas such as speech recognition, handwriting recognition, face recognition, "spam" filtering (junk email), and targeted advertising. Although multiple radiation oncology research groups have shown the value of applied machine learning (ML), clinical adoption has been slow due to the high barrier to understanding these complex models by clinicians. Here, we present a review of the use of ML to predict radiation therapy outcomes from the clinician's point of view with the hope that it lowers the "barrier to entry" for those without formal training in ML. We begin by describing 7 principles that one should consider when evaluating (or creating) an ML model in radiation oncology. We next introduce 3 popular ML methods--logistic regression (LR), support vector machine (SVM), and artificial neural network (ANN)--and critique 3 seminal papers in the context of these principles. Although current studies are in exploratory stages, the overall methodology has progressively matured, and the field is ready for larger-scale further investigation.

Patient- and therapy-related factors associated with the incidence of xerostomia in nasopharyngeal carcinoma patients receiving parotid-sparing helical tomotherapy

We investigated the incidence of moderate to severe patient-reported xerostomia among nasopharyngeal carcinoma (NPC) patients treated with helical tomotherapy (HT) and identified patient- and therapy-related factors associated with acute and chronic xerostomia toxicity. The least absolute shrinkage and selection operator (LASSO) normal tissue complication probability (NTCP) models were developed using quality-of-life questionnaire datasets from 67 patients with NPC. For acute toxicity, the dosimetric factors of the mean doses to the ipsilateral submandibular gland (Dis) and the contralateral submandibular gland (Dcs) were selected as the first two significant predictors. For chronic toxicity, four predictive factors were selected: age, mean dose to the oral cavity (Doc), education, and T stage. The substantial sparing data can be used to avoid xerostomia toxicity. We suggest that the tolerance values corresponded to a 20% incidence of complications (TD20) for Dis = 39.0 Gy, Dcs = 38.4 Gy, and Doc = 32.5 Gy, respectively, when mean doses to the parotid glands met the QUANTEC 25 Gy sparing guidelines. To avoid patient-reported xerostomia toxicity, the mean doses to the parotid gland, submandibular gland, and oral cavity have to meet the sparing tolerance, although there is also a need to take inherent patient characteristics into consideration.

Developing Multivariable Normal Tissue Complication Probability Model to Predict the Incidence of Symptomatic Radiation Pneumonitis among Breast Cancer Patients

Purpose

Symptomatic radiation pneumonitis (SRP), which decreases quality of life (QoL), is the most common pulmonary complication in patients receiving breast irradiation. If it occurs, acute SRP usually develops 4–12 weeks after completion of radiotherapy and presents as a dry cough, dyspnea and low-grade fever. If the incidence of SRP is reduced, not only the QoL but also the compliance of breast cancer patients may be improved. Therefore, we investigated the incidence SRP in breast cancer patients after hybrid intensity modulated radiotherapy (IMRT) to find the risk factors, which may have important effects on the risk of radiation-induced complications.

Methods

In total, 93 patients with breast cancer were evaluated. The final endpoint for acute SRP was defined as those who had density changes together with symptoms, as measured using computed tomography. The risk factors for a multivariate normal tissue complication probability model of SRP were determined using the least absolute shrinkage and selection operator (LASSO) technique.

Results

Five risk factors were selected using LASSO: the percentage of the ipsilateral lung volume that received more than 20-Gy (IV20), energy, age, body mass index (BMI) and T stage. Positive associations were demonstrated among the incidence of SRP, IV20, and patient age. Energy, BMI and T stage showed a negative association with the incidence of SRP. Our analyses indicate that the risk of SPR following hybrid IMRT in elderly or low-BMI breast cancer patients is increased once the percentage of the ipsilateral lung volume receiving more than 20-Gy is controlled below a limitation.

Conclusions

We suggest to define a dose-volume percentage constraint of IV20< 37% (or AIV20< 310cc) for the irradiated ipsilateral lung in radiation therapy treatment planning to maintain the incidence of SPR below 20%, and pay attention to the sequelae especially in elderly or low-BMI breast cancer patients. (AIV20: the absolute ipsilateral lung volume that received more than 20 Gy (cc).