Radiomics-based Prognosis Analysis for Non-Small Cell Lung Cancer

Abstract

Radiomics characterizes tumor phenotypes by extracting large numbers of quantitative features from radiological images. Radiomic features have been shown to provide prognostic value in predicting clinical outcomes in several studies. However, several challenges including feature redundancy, unbalanced data, and small sample sizes have led to relatively low predictive accuracy. In this study, we explore different strategies for overcoming these challenges and improving predictive performance of radiomics-based prognosis for non-small cell lung cancer (NSCLC). CT images of 112 patients (mean age 75 years) with NSCLC who underwent stereotactic body radiotherapy were used to predict recurrence, death, and recurrence-free survival using a comprehensive radiomics analysis. Different feature selection and predictive modeling techniques were used to determine the optimal configuration of prognosis analysis. To address feature redundancy, comprehensive analysis indicated that Random Forest models and Principal Component Analysis were optimum predictive modeling and feature selection methods, respectively, for achieving high prognosis performance. To address unbalanced data, Synthetic Minority Over-sampling technique was found to significantly increase predictive accuracy. A full analysis of variance showed that data endpoints, feature selection techniques, and classifiers were significant factors in affecting predictive accuracy, suggesting that these factors must be investigated when building radiomics-based predictive models for cancer prognosis.

Repeatability of 18F‐FDG PET radiomic features: A phantom study to explore sensitivity to image reconstruction settings, noise, and delineation method

“…Although qualitative visual image assessment remains important for these purposes, it has a limited capability to objectively quantify tracer uptake. The most widely used semi‐quantitative measures are the maximum, mean, and peak standardized uptake value (SUV max , SUV mean , and SUV peak ) and morphologically based imaging features, such as the metabolic tumor volume or total lesion glycolysis . However, these features ignore the intratumoral 18 F‐FDG spatial distribution .…”

“…However, these features ignore the intratumoral 18 F‐FDG spatial distribution . The rapidly emerging field of ‘radiomics’ computes a large number of quantitative image features to characterize this intratumoral distribution or other tumor phenotypes such as shape …”

Influence of gray level discretization on radiomic feature stability for different CT scanners, tube currents and slice thicknesses: a comprehensive phantom study

“…Extraction of quantitative imaging features, also called radiomics [1], has become an additional source of information for the development of prognostic and predictive models [2][3][4][5][6][7]. The total number of features that can be calculated is almost unlimited, especially if filter-based features (e.g., Laplacian of Gaussian or wavelet) are also taken into consideration.…”

“…Several studies already showed that radiomic feature values are influenced by image acquisition and reconstruction settings, like slice thickness and exposure [2,[9][10][11][12][13]. For instance, Mackin et al [13] scanned a phantom with ten unique inserts using different acquisition parameters on computed tomography (CT) scanners of four manufacturers.…”

Developing a periodontal disease antibody array for the prediction of severe periodontal disease using machine learning classifiers

“…These five models included support vector machine (SVM), random forest (RF), k nearest neighbor (kNN), linear discriminant analysis (LDA), and Classification and Regression Trees (CART). [28][29][30][31] This study represents the first comparison of classification models of periodontitis using GCF.…”