Assessing Glioblastoma Treatment Response Using Machine Learning Approach Based on Magnetic Resonance Images Radiomics: An Exploratory Study DOI Creative Commons

Amirreza Sadeghinasab,

Jafar Fatahiasl, Marziyeh Tahmasbi

et al.

Health Science Reports, Journal Year: 2024, Volume and Issue: 8(1)

Published: Dec. 30, 2024

ABSTRACT Background and Objectives Assessing treatment response in glioblastoma multiforme (GBM) tumors necessitates developing more objective quantitative approaches. A machine learning‐based approach is presented this exploratory study for GBM patients' assessment based on radiomics extracted from magnetic resonance (MR) images. Methods MR images 77 patients were acquired at two post‐surgery stages preprocessed. From these images, 107 the segmented tumoral cavities. The most informative features training learning (ML) classifiers identified using Spearman correlation analysis of retained by forward sequential LASSO algorithms. Applied models included support vector (SVM), random forest (RF), K‐nearest neighbors (KNN), AdaBoost, categorical boosting (CatBoost), light gradient (LightGBM), extreme (XGBoost), Naïve Bayes (NB) logistic regression (LR). Ten‐fold cross‐validation was used to validate models. Statistical conducted SPSS version 27; p ‐value < 0.05 considered significant. Results classifier demonstrated highest performance among trained models, achieving an AUC (area under receiver operating characteristic curve) 0.86 ± 0.13 when seven selected algorithm 0.84 0.14 five chosen algorithm. second‐best observed with KNN classifier, which achieved 0.80 0.17 Conclusion Findings that MRI‐based could be as distinctive train ML assessment. Trained serve aiding tools expedite besides qualitative evaluations.

Language: Английский

A Machine Learning-Based Radiomics Model for the Differential Diagnosis of Benign and Malignant Thyroid Nodules in F-18 FDG PET/CT: External Validation in the Different Scanner DOI Open Access

J.W. Lee,

Jinny Lee, Bong‐Il Song

et al.

Cancers, Journal Year: 2025, Volume and Issue: 17(2), P. 331 - 331

Published: Jan. 20, 2025

Background/Objectives: Accurate diagnosis is essential to avoid unnecessary procedures for thyroid incidentalomas (TIs). Advances in radiomics and machine learning applied medical imaging offer promise assessing nodules. This study utilized analysis on F-18 FDG PET/CT improve preoperative differential of TIs. Methods: A total 152 patient cases were retrospectively analyzed split into training validation sets (7:3) using stratification randomization. Results: The least absolute shrinkage selection operator (LASSO) algorithm identified nine features from 960 candidates construct a signature predictive malignancy. Performance the score was evaluated receiver operating characteristic (ROC) area under curve (AUC). In set, achieved an AUC 0.794 (95% CI: 0.703–0.885, p < 0.001). Validation performed internal external datasets, yielding AUCs 0.702 0.547–0.858, = 0.011) 0.668 0.500–0.838, 0.043), respectively. Conclusions: These results demonstrate that selected effectively differentiate malignant Overall, model shows potential as valuable tool cancer patients with TIs, supporting improved decision-making.

Language: Английский

Citations

0
Predicting overall survival in glioblastoma patients using machine learning: an analysis of treatment efficacy and patient prognosis DOI Creative Commons

Răzvan Onciul,

Felix-Mircea Brehar,

Adrian Dumitru

et al.

Frontiers in Oncology, Journal Year: 2025, Volume and Issue: 15

Published: April 9, 2025

Glioblastoma (GBM), the most aggressive primary brain tumor, poses a significant challenge in predicting patient survival due to its heterogeneity and resistance treatment. Accurate prediction is essential for optimizing treatment strategies improving clinical outcomes. This study utilized metadata from 135 GBM patients, including demographic, clinical, molecular variables such as age, Karnofsky Performance Status (KPS), MGMT promoter methylation, EGFR amplification. Six machine learning models-XGBoost, Random Forests, Support Vector Machines, Artificial Neural Networks, Extra Trees Regressor, K- Nearest Neighbors-were employed classify patients into predefined categories. Data preprocessing included label encoding categorical MinMax scaling numerical features. Model performance was assessed using ROC-AUC accuracy metrics, with hyperparameters optimized through grid search. XGBoost demonstrated highest predictive accuracy, achieving mean of 0.90 an 0.78. Ensemble models outperformed simpler classifiers, emphasizing value metadata. The identified key prognostic markers, methylation KPS, contributors prediction. application offers robust approach survival. highlights potential ML enhance decision-making contribute personalized strategies, focus on reliability, interpretability.

Language: Английский

Citations

0
Survival prediction of glioblastoma patients using machine learning and deep learning: a systematic review DOI Creative Commons

Roya Poursaeed,

Mohsen Mohammadzadeh, Ali Asghar Safaei

et al.

BMC Cancer, Journal Year: 2024, Volume and Issue: 24(1)

Published: Dec. 27, 2024

Glioblastoma Multiforme (GBM), classified as a grade IV glioma by the World Health Organization (WHO), is prevalent and notably aggressive form of brain tumor derived from glial cells. It stands one most severe forms primary cancer in humans. The median survival time GBM patients only 12–15 months, making it lethal type tumor. Every year, about 200,000 people worldwide succumb to this disease. also highly heterogeneous, meaning that its characteristics behavior vary widely among different patients. This leads outcomes times for each individual. Predicting accurately can have multiple benefits. enable optimal personalized treatment planning based on patient's condition prognosis. support their families cope with possible make informed decisions care quality life. Furthermore, assist researchers scientists discover relevant biomarkers, features, mechanisms disease design more effective therapies. Artificial intelligence methods, such machine learning deep learning, been applied prediction various fields, breast cancer, lung gastric cervical liver prostate covid 19. systematic review summarizes current state-of-the-art methods predicting glioblastoma using types input data, clinical molecular markers, imaging radiomics omics data or combination them. Following PRISMA guidelines, we searched databases 2015 2024, reviewing 107 articles meeting our criteria. We analyzed sources, performance metrics studies. found random forest was popular method, common data.

Language: Английский

Citations

2
Assessing Glioblastoma Treatment Response Using Machine Learning Approach Based on Magnetic Resonance Images Radiomics: An Exploratory Study DOI Creative Commons

Amirreza Sadeghinasab,

Jafar Fatahiasl, Marziyeh Tahmasbi

et al.

Health Science Reports, Journal Year: 2024, Volume and Issue: 8(1)

Published: Dec. 30, 2024

ABSTRACT Background and Objectives Assessing treatment response in glioblastoma multiforme (GBM) tumors necessitates developing more objective quantitative approaches. A machine learning‐based approach is presented this exploratory study for GBM patients' assessment based on radiomics extracted from magnetic resonance (MR) images. Methods MR images 77 patients were acquired at two post‐surgery stages preprocessed. From these images, 107 the segmented tumoral cavities. The most informative features training learning (ML) classifiers identified using Spearman correlation analysis of retained by forward sequential LASSO algorithms. Applied models included support vector (SVM), random forest (RF), K‐nearest neighbors (KNN), AdaBoost, categorical boosting (CatBoost), light gradient (LightGBM), extreme (XGBoost), Naïve Bayes (NB) logistic regression (LR). Ten‐fold cross‐validation was used to validate models. Statistical conducted SPSS version 27; p ‐value < 0.05 considered significant. Results classifier demonstrated highest performance among trained models, achieving an AUC (area under receiver operating characteristic curve) 0.86 ± 0.13 when seven selected algorithm 0.84 0.14 five chosen algorithm. second‐best observed with KNN classifier, which achieved 0.80 0.17 Conclusion Findings that MRI‐based could be as distinctive train ML assessment. Trained serve aiding tools expedite besides qualitative evaluations.

Language: Английский

Citations

0