Optimization of treatment strategy by using a machine learning model to predict survival time of patients with malignant glioma after radiotherapy

Artificial Neural Network Prediction of Mortality in Cancer Patients Presenting for Radiation Therapy at a Multisite Institution

INTRODUCTION

For many decades, the management of cancer has utilized radiation therapy, which continues to evolve with technology to improve patient outcomes. However, despite the standardization of treatment plans and the establishment of best clinical practices based on prospective, randomized trials and adherence to National Comprehensive Cancer Network (NCCN) guidelines, the outcomes from radiation therapy are highly variable and dependent on a number of factors, including patient demographics, tumor characteristics/histology, and treatment parameters. In this study, we attempt to use available patient data and treatment parameters at the time of radiation therapy to predict future outcomes using artificial intelligence (AI).

METHODS

Six thousand five hundred ninety-five cases of patients who completed radiation treatment were selected retrospectively and used to train artificial neural networks (ANNs) and baseline models (i.e., logistic regression, random forest, support vector machines [SVMs], gradient boosting [XGBoost]) for binary classification of mortality at multiple time points ranging from six months to five years post-treatment. A hyperparameter grid search was used to identify the optimal network architecture for each time point, using sensitivity as the primary outcome metric.

RESULTS

The median age was 75 years (range: 2-102 years). There were 63.8% females and 36.1% males. The results indicate that ANNs were able to successfully perform binary mortality prediction with an accuracy greater than random chance and greater sensitivity than baseline models used. The best-performing algorithm was the ANN, which achieved a sensitivity of 83.00% ± 4.89% for five-year mortality.

CONCLUSION

The neural network was able to achieve higher sensitivity than Logistic Regression, SVM Random Forest, and XGBoost across all output target variables, demonstrating the utility of a neural network model for mortality prediction on the provided dataset.

Virtual clinical trial based on outcome modeling with iteratively redistributed extrapolation data

Virtual clinical trials (VCTs) can potentially simulate clinical trials on a computer, but their application with a limited number of past clinical cases is challenging due to the biased estimation of the statistical population. In this study, we developed ExMixup, a novel training technique based on machine learning, using iteratively redistributed extrapolated data. Information obtained from 100 patients with prostate cancer and 385 patients with oropharyngeal cancer was used to predict the recurrence after radiotherapy. Model performance was evaluated by developing outcome prediction models based on three types of training methods: training with original data (baseline), interpolation data (Mixup), and interpolation + extrapolation data (ExMixup). Two types of VCTs were conducted to predict the treatment response of patients with distinct characteristics compared to the training data obtained from patient cohorts categorized under risk classification or cancer stage. The prediction models developed with ExMixup yielded concordance indices (95% confidence intervals) of 0.751 (0.719-0.818) and 0.752 (0.734-0.785) for VCTs on the prostate and oropharyngeal cancer datasets, respectively, which significantly outperformed the baseline and Mixup models (P < 0.01). The proposed approach could enhance the ability of VCTs to predict treatment results in patients excluded from past clinical trials.

Predictive value of magnetic resonance imaging radiomics-based machine learning for disease progression in patients with high-grade glioma

Background

Accurately predicting the prognosis of patients with high-grade glioma (HGG) is potentially important for treatment. However, the predictive value of images of various magnetic resonance imaging (MRI) sequences for prognosis at different time points is unknown. We established predictive machine learning models of HGG disease progression and recurrence using MRI radiomics and explored the factors influencing prediction accuracy.

Methods

Radiomics features were extracted from T1-weighted (T1WI), contrast-enhanced T1-weighted (CE-T1WI), T2-weighted (T2WI), and fluid-attenuated inversion recovery (FLAIR) images (postoperative radiotherapy planning MRI images) obtained from 162 patients with HGG. The Mann-Whitney U test and least absolute shrinkage and selection operator (LASSO) algorithm were used for feature selection. Machine learning models were used to build prediction models to estimate disease progression or recurrence. The influence of different MRI sequences, regions of interest (ROIs), and prediction time points was also explored. The receiver operating characteristic (ROC) curve was used to evaluate the discriminative performance of each model, and the DeLong test was employed to compare the ROC curves.

Results

Radiomics features from T2WI and FLAIR demonstrated greater predictive value for disease progression compared with T1WI or CE-TIWI. The best predictive models, with areas under the ROC curves (AUCs) of 0.70, 0.68, 0.78, 0.78, and 0.78 for predicting disease progression at the 6th, 9th, 12th, 15th, and 18th month after radiotherapy, respectively, were obtained by combining clinical features with gross tumor volume (GTV) and clinical target volume (CTV) features extracted from T2WI and FLAIR.

Conclusions

Structural MRI obtained before radiotherapy can be used to predict the disease progression or posttreatment recurrence of HGG. When using MRI radiomics to predict long-term outcomes as opposed to short-term outcomes, better predictive results may be obtained.

Machine learning based survival prediction in Glioma using large-scale registry data

Gliomas are the most common central nervous system tumors exhibiting poor clinical outcomes. The ability to estimate prognosis is crucial for both patients and providers in order to select the most appropriate treatment. Machine learning (ML) allows for sophisticated approaches to survival prediction using real world clinical parameters needed to achieve superior predictive accuracy. We employed Cox Proportional hazards (CPH) model, Support Vector Machine (SVM) model, Random Forest (RF) model in a large glioma dataset (3462 patients, diagnosed 2000-2018) to explore the most optimal approach to survival prediction. Features employed were age, sex, surgical resection status, tumor histology and tumor site, administration of radiation therapy (RT) and chemotherapy status. Concordance index (c-index) was employed to assess the accuracy of survival time prediction. All three models performed well with prediction accuracy (CI 0.767, 0.771, 0.57 for CPH, SVM, RF models respectively) with the best performance achieved when incorporating RT and chemotherapy administration status which emerged as key predictive features. Within the subset of glioblastoma patients, similar prediction accuracy was achieved. These findings should prompt stricter clinician oversight over registry data accuracy through quality assurance as we move towards meaningful predictive ability using ML approaches in glioma.

Efficacy and Safety of Temozolomide Combined with Radiotherapy in the Treatment of Malignant Glioma

MG is a clinical common intracranial tumor, with the characteristics of strong invasion. In our study, we aim to explore the efficacy and safety of temozolomide combined with radiotherapy in the treatment of malignant glioma (MG) and its influence on postoperative complications and survival rate of patients. 120 MG patients admitted to our hospital (January 2019-January 2020) were chosen as the research objects and were randomly divided into group A (n = 60) and group B (n = 60). All patients were treated with radiotherapy, and patients in group A were additionally treated with temozolomide. The clinical efficacy, quality of life, incidence of adverse reactions, incidence of postoperative complications, survival rates, and average survival time of the two groups were compared. The objective remission rate (ORR), disease control rate (DCR), survival rates after one year and two years of follow-up, and the number of patients with improved quality of life in group A were markedly higher compared with group B (P < 0.05). The incidence of postoperative complications in group A was remarkably lower compared with group B (P < 0.05). The average survival time of group A was dramatically longer compared with group B (P < 0.001). There was no significant difference in the incidence of adverse reactions between the two groups (P > 0.05), and no new adverse reactions occurred in the patients. Temozolomide combined with radiotherapy can effectively improve the quality of life, treatment effect, and survival rate of MG patients, with a lower incidence of postoperative complications and better tolerance. Our finding indicates that temozolomide combined with radiotherapy has a high clinical application value. In addition, it indicates that this treatment method should be promoted in practice.

Current landscape and potential future applications of artificial intelligence in medical physics and radiotherapy

Artificial intelligence (AI) has seen tremendous growth over the past decade and stands to disrupts the medical industry. In medicine, this has been applied in medical imaging and other digitised medical disciplines, but in more traditional fields like medical physics, the adoption of AI is still at an early stage. Though AI is anticipated to be better than human in certain tasks, with the rapid growth of AI, there is increasing concerns for its usage. The focus of this paper is on the current landscape and potential future applications of artificial intelligence in medical physics and radiotherapy. Topics on AI for image acquisition, image segmentation, treatment delivery, quality assurance and outcome prediction will be explored as well as the interaction between human and AI. This will give insights into how we should approach and use the technology for enhancing the quality of clinical practice.

Towards an Image-Informed Mathematical Model of In Vivo Response to Fractionated Radiation Therapy

Simple Summary

Using medical imaging data and computational models, we develop a modeling framework to provide personalized treatment response forecasts to fractionated radiation therapy for individual tumors. We evaluate this approach in an animal model of brain cancer and forecast changes in tumor cellularity and vasculature.

Abstract

Fractionated radiation therapy is central to the treatment of numerous malignancies, including high-grade gliomas where complete surgical resection is often impractical due to its highly invasive nature. Development of approaches to forecast response to fractionated radiation therapy may provide the ability to optimize or adapt treatment plans for radiotherapy. Towards this end, we have developed a family of 18 biologically-based mathematical models describing the response of both tumor and vasculature to fractionated radiation therapy. Importantly, these models can be personalized for individual tumors via quantitative imaging measurements. To evaluate this family of models, rats (n = 7) with U-87 glioblastomas were imaged with magnetic resonance imaging (MRI) before, during, and after treatment with fractionated radiotherapy (with doses of either 2 Gy/day or 4 Gy/day for up to 10 days). Estimates of tumor and blood volume fractions, provided by diffusion-weighted MRI and dynamic contrast-enhanced MRI, respectively, were used to calibrate tumor-specific model parameters. The Akaike Information Criterion was employed to select the most parsimonious model and determine an ensemble averaged model, and the resulting forecasts were evaluated at the global and local level. At the global level, the selected model’s forecast resulted in less than 16.2% error in tumor volume estimates. At the local (voxel) level, the median Pearson correlation coefficient across all prediction time points ranged from 0.57 to 0.87 for all animals. While the ensemble average forecast resulted in increased error (ranging from 4.0% to 1063%) in tumor volume predictions over the selected model, it increased the voxel wise correlation (by greater than 12.3%) for three of the animals. This study demonstrates the feasibility of calibrating a model of response by serial quantitative MRI data collected during fractionated radiotherapy to predict response at the conclusion of treatment.

A machine learning-based survival prediction model of high grade glioma by integration of clinical and dose-volume histogram parameters

PURPOSE

Glioma is the most common type of primary brain tumor in adults, and it causes significant morbidity and mortality, especially in high-grade glioma (HGG) patients. The accurate prognostic prediction of HGG is vital and helpful for clinicians when developing therapeutic strategies. Therefore, we propose a machine learning-based survival prediction model by analyzing clinical and dose-volume histogram (DVH) parameters, to improve the performance of the risk model in HGG patients.

METHODS

Eight clinical variables and 39 DVH parameters were extracted for each patient, who received radiotherapy for HGG with active follow-up. Ninety-five patients were randomly divided into training and testing cohorts, and we employed random survival forest (RSF), support vector machine (SVM), and Cox proportional hazards (CPHs) models to predict survival. Calibration plots, concordance indexes, and decision curve analyses were used to evaluate the calibration, discrimination, and clinical utility of these three models.

RESULTS

The RSF model showed the best performance among the three models, with concordance indexes of 0.824 and 0.847 in the training and testing sets, respectively, followed by the SVM (0.792/0.823) and CPH (0.821/0.811) models. Specifically, in the RSF model, we identified age, gross tumor volume (GTV), grade, Karnofsky performance status (KPS), isocitrate dehydrogenase (IDH), and D99 as important variables associated with survival. The AUCs of the testing set were 92.4%, 87.7%, and 84.0% for 1-, 2-, and 3-year survival, respectively. According to this model, HGG patients can be divided into high- and low-risk groups.

CONCLUSION

The machine learning-based RSF model integrating both clinical and DVH variables is an improved and useful tool for predicting the survival of HGG patients.

A Comparison Study of Machine Learning (Random Survival Forest) and Classic Statistic (Cox Proportional Hazards) for Predicting Progression in High-Grade Glioma after Proton and Carbon Ion Radiotherapy

Background

Machine learning (ML) algorithms are increasingly explored in glioma prognostication. Random survival forest (RSF) is a common ML approach in analyzing time-to-event survival data. However, it is controversial which method between RSF and traditional cornerstone method Cox proportional hazards (CPH) is better fitted. The purpose of this study was to compare RSF and CPH in predicting tumor progression of high-grade glioma (HGG) after particle beam radiotherapy (PBRT).

Methods

The study enrolled 82 consecutive HGG patients who were treated with PBRT at Shanghai Proton and Heavy Ion Center between 6/2015 and 11/2019. The entire cohort was split into the training and testing set in an 80/20 ratio. Ten variables from patient-related, tumor-related and treatment-related information were utilized for developing CPH and RSF for predicting progression-free survival (PFS). The model performance was compared in concordance index (C-index) for discrimination (accuracy), brier score (BS) for calibration (precision) and variable importance for interpretability.

Results

The CPH model demonstrated a better performance in terms of integrated C-index (62.9%) and BS (0.159) compared to RSF model (C-index = 61.1%, BS = 0.174). In the context of variable importance, CPH model indicated that age (P = 0.024), WHO grade (P = 0.020), IDH gene (P = 0.019), and MGMT promoter status (P = 0.040) were significantly correlated with PFS in the univariate analysis; multivariate analysis showed that age (P = 0.041), surgical completeness (P = 0.084), IDH gene (P = 0.057), and MGMT promoter (P = 0.092) had a significant or trend toward the relation with PFS. RSF showed that merely IDH and age were of positive importance for predicting PFS. A final nomogram was developed to predict tumor progression at the individual level based on CPH model.

Conclusions

In a relatively small dataset with HGG patients treated with PBRT, CPH outperformed RSF for predicting tumor progression. A comprehensive criterion with accuracy, precision, and interpretability is recommended in evaluating ML prognostication approaches for clinical deployment.