Analysis of Cross-Combinations of Feature Selection and Machine-Learning Classification Methods Based on [18F]F-FDG PET/CT Radiomic Features for Metabolic Response Prediction of Metastatic Breast Cancer Lesions

Abstract

Background: This study aimed to identify optimal combinations between feature selection methods and machine-learning classifiers for predicting the metabolic response of individual metastatic breast cancer lesions, based on clinical variables and radiomic features extracted from pretreatment [18F]F-FDG PET/CT images. Methods: A total of 48 patients with confirmed metastatic breast cancer, who received different treatments, were included. All patients had an [18F]F-FDG PET/CT scan before and after the treatment. From 228 metastatic lesions identified, 127 were categorized as responders (complete or partial metabolic response) and 101 as non-responders (stable or progressive metabolic response), by using the percentage changes in SULpeak (peak standardized uptake values normalized for body lean body mass). The lesion pool was divided into training (n = 182) and testing cohorts (n = 46); for each lesion, 101 image features from both PET and CT were extracted (202 features per lesion). These features, along with clinical and pathological information, allowed the prediction model’s construction by using seven popular feature selection methods in cross-combination with another seven machine-learning (ML) classifiers. The performance of the different models was investigated with the receiver-operating characteristic curve (ROC) analysis, using the area under the curve (AUC) and accuracy (ACC) metrics. Results: The combinations, least absolute shrinkage and selection operator (Lasso) + support vector machines (SVM), or random forest (RF) had the highest AUC in the cross-validation, with 0.93 ± 0.06 and 0.92 ± 0.03, respectively, whereas Lasso + neural network (NN) or SVM, and mutual information (MI) + RF, had the higher AUC and ACC in the validation cohort, with 0.90/0.72, 0.86/0.76, and 87/85, respectively. On average, the models with Lasso and models with SVM had the best mean performance for both AUC and ACC in both training and validation cohorts. Conclusions: Image features obtained from a pretreatment [18F]F-FDG PET/CT along with clinical vaiables could predict the metabolic response of metastatic breast cancer lesions, by their incorporation into predictive models, whose performance depends on the selected combination between feature selection and ML classifier methods.

Identification of Resting-State Network Functional Connectivity and Brain Structural Signatures in Fibromyalgia Using a Machine Learning Approach

“…The cross-combination strategy could filter down irrelevant features and select the optimal ML model for classification [ 39 , 40 ]. Regarding feature selection methods, our study showed that the RFE method exhibited high performance when combined with all classifiers in building baseline ML models.…”

“…To find good performance baseline ML models, we used the cross-combination strategy in which each feature selection method mentioned above was combined with several classifiers [ 39 , 40 ]. We investigated six classification algorithms: support vector machine (SVM), Logistic Regression (LR), k-nearest neighbors (KNN), random forest (RF), linear discriminative analysis (LDA), and Gaussian Naïve Bayes (GBN), which were commonly used in machine learning of MRI data [ 41 , 42 , 43 , 44 , 45 ].…”

Nuclear medicine radiomics in digestive system tumors: Concept, applications, challenges, and future perspectives

“…[45,[83][84][85][86][87] The application of ML algorithms, including but not limited to random forest (RF), support vector machine (SVM), Gaussian naive Bayes (GNB), artificial neural network (ANN), KNN, adaptive boosting (AdaBoost), logistic regression, K-means clustering, principal component analysis (PCA), and hierarchical clustering, has demonstrated potential to enhance both predictive and diagnostic accuracy. [87][88][89][90] The RF algorithm generates an ensemble of decision trees, collectively known as a forest, which can effectively tackle classification or regression tasks. Additionally, an internal optimization technique is utilized to determine an importance score for each feature, reflecting their respective contributions to the optimization process.…”

“…[ 45,83–87 ] The application of ML algorithms, including but not limited to random forest (RF), support vector machine (SVM), Gaussian naive Bayes (GNB), artificial neural network (ANN), KNN, adaptive boosting (AdaBoost), logistic regression, K ‐means clustering, principal component analysis (PCA), and hierarchical clustering, has demonstrated potential to enhance both predictive and diagnostic accuracy. [ 87–90 ]…”

Optimal 18F-FDG PET/CT radiomics model development for predicting EGFR mutation status and prognosis in lung adenocarcinoma: a multicentric study

“…Recently, the value of PET/CT-based HRFs in predicting EGFR mutation status, EGFR subtypes and prognosis had been well demonstrated ( 12 – 18 ), which involved different feature selection methods and machine learning algorithms. These different radiomics pipelines may lead to various predictive performance ( 19 ). Feature selection methods also have a great influence on clinical predictive models ( 20 ).…”