Radiomics approaches to predict PD-L1 and PFS in advanced non-small cell lung patients treated with immunotherapy: a multi-institutional study

Abstract

With the increasing use of immune checkpoint inhibitors (ICIs), there is an urgent need to identify biomarkers to stratify responders and non-responders using programmed death-ligand (PD-L1) expression, and to predict patient-specific outcomes such as progression free survival (PFS). The current study is aimed to determine the feasibility of building imaging-based predictive biomarkers for PD-L1 and PFS through systematically evaluating a combination of several machine learning algorithms with different feature selection methods. A retrospective, multicenter study of 385 advanced NSCLC patients amenable to ICIs was undertaken in two academic centers. Radiomic features extracted from pretreatment CT scans were used to build predictive models for PD-L1 and PFS (short-term vs. long-term survivors). We first employed the LASSO methodology followed by five feature selection methods and seven machine learning approaches to build the predictors. From our analyses, we found several combinations of feature selection methods and machine learning algorithms to achieve a similar performance. Logistic regression with ReliefF feature selection (AUC = 0.64, 0.59 in discovery and validation cohorts) and SVM with Anova F-test feature selection (AUC = 0.64, 0.63 in discovery and validation datasets) were the best-performing models to predict PD-L1 and PFS. This study elucidates the application of suitable feature selection approaches and machine learning algorithms to predict clinical endpoints using radiomics features. Through this study, we identified a subset of algorithms that should be considered in future investigations for building robust and clinically relevant predictive models.

Multi-institutional prognostic modeling of survival outcomes in NSCLC patients treated with first-line immunotherapy using radiomics

“…This has propelled research towards building imaging-based prognostic and predictive biomarkers utilizing routine medical images [ 17 ]. Recent studies have leveraged CT-scans of NSCLC patients treated with ICIs and developed imaging biomarkers to predict immunotherapy response and survival outcomes [ 17 ].…”

“…In the validation phase, we trained each of the models on the CHUM dataset with best features, then evaluated them on the IUCPQ dataset with the best features. Finally, we computed the C-index for OS and PFS for the validation phase (similar to our previous works [ 17 , 29 ]). In addition, the hyperparameter tuning of the classifiers was carried out through the process of cross-validation with the help of the GridSearchCV class provided by scikit-learn.…”

“…A recent comparative study by Li et al the authors used 3 machine learning models along with EHR data to build predictive models for clinical endpoints using data from a single institution [ 15 ]. While there have been several promising prognostic modeling studies leveraging radiomics data, the adoption of radiomics-based prognostic models in clinical workflows is still limited [ 16 , 17 ]. There are several factors that affected adopting these models in clinical settings, which include the lack of a clear predictive feature selection and learning strategy, relatively small sample size and single-institution datasets lacking external validation as well as generalizability.…”

A Radiomics-Clinical Model Predicts Overall Survival of Non-Small Cell Lung Cancer Patients Treated with Immunotherapy: A Multicenter Study

“…The advent of radiomics presented a novel non-invasive strategy to build survival risk models as well as to monitor response to various treatments, including ICIs [13,14,37,38]. Despite the growing field of biomarker research in immuno-oncology, none of these prognostic and predictive biomarkers have been translated to routine clinical use in the context of predicting patient outcomes to ICIs in a metastatic setting or to survival.…”

“…Furthermore, these quantitative features are extracted from a region of interest (ROI) of medical scans to build biologically and clinically relevant models for predicting patient-specific endpoints, such as OS, progression-free survival (PFS), response to various therapeutic interventions, etc. [13]. One of the earliest works using radiomics was carried out by Aerts et al, who built a prognostic signature that was validated for both lung and head & neck cancers [14].…”