Survival Rate Prediction in Glioblastoma Patients Using Radiomics Extracted from Post-Contrast Magnetic Resonance Images: Comparison of Multiple Machine Learning Models.
Saved in:
| Title: | Survival Rate Prediction in Glioblastoma Patients Using Radiomics Extracted from Post-Contrast Magnetic Resonance Images: Comparison of Multiple Machine Learning Models. |
|---|---|
| Authors: | Sadeghinasab, Amirreza, Jafari, Mostafa, Tahmasbi, Marziyeh, Fatahiasl, Jafar |
| Source: | Reports of Radiotherapy & Oncology; 2024 Supplement, Vol. 10, p52-53, 2p |
| Subject Terms: | MACHINE learning, MAGNETIC resonance imaging, IMAGE analysis, BRAIN tumors, OVERALL survival |
| Abstract: | Introduction: Glioblastoma multiforme (GBM) is one of the most aggressive primary malignant brain tumors. The standard treatment for GBM combines surgery, chemotherapy, and radiotherapy. Magnetic resonance imaging (MRI) serves as the primary imaging modality for GBM tumors, accurately depicting the tumor's center and margins. MR sequences including T1, T2, T2FLAIR, diffusion - weighted imaging (DWI), and post - contrast T1, are mainly used for GBM studies. In addition to subjective visual analysis of the images by radiologists, radiomics (shape, first, second, or higher - order statistics, and texture features) extracted from images, can also be applied to evaluate GBM patients' treatment outcomes. These extracted radiomics can serve as input data for machine learning (ML) models, contributing to predicting GBM patients' survival rates. This study aims to review various ML algorithms utilized for predicting GBM patients' survival rates. Methods: PubMed, Science Direct, Web of Science, and Google Scholar databases were explored up to November 2023, using different combinations of the keywords: "Glioblastoma multiforme ", "GBM", "survival prediction", "Magnetic resonance imaging", "radiomics", "artificial intelligence", "machine learning" and "deep learning". Finally, seven more recent and relevant records were included in the study. Results: Based on the results of the reviewed records, SVM (support vector machine), RF (random forest), LASSO (least absolute shrinkage and selection operator), XGBOOST, CATBOOST, LIGHTGBM, Nearest neighbor, Neural network, Multilayer perceptron (MLP), and Naïve Bayes models have been applied for GBM patients' survival rate prediction. The highest reported AUC (area under the curve) value was related to a Nearest neighbor - based model. The reported average AUC values for SVM and RF models were equal to 0.793 and 0.889, respectively. Furthermore, the RFE - SVM combined model presented the highest level of accuracy. The LASSO model also showed an appropriate performance in predicting GBM patients' survival rates. Moreover, Haralick features were the most significant applied radiomics with ML models. Conclusions: The high AUC and accuracy values in different studies for ML models predicting GBM patients' survival rates and treatment outcomes indicate the potential efficacy of such models as supportive tools for clinicians. These ML - based approaches using MRI - based extracted quantitative features, precisely predict GBM patient survival. [ABSTRACT FROM AUTHOR] |
| Copyright of Reports of Radiotherapy & Oncology is the property of Brieflands and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Complementary Index |
Nájsť tento článok vo Web of Science | Abstract: | Introduction: Glioblastoma multiforme (GBM) is one of the most aggressive primary malignant brain tumors. The standard treatment for GBM combines surgery, chemotherapy, and radiotherapy. Magnetic resonance imaging (MRI) serves as the primary imaging modality for GBM tumors, accurately depicting the tumor's center and margins. MR sequences including T1, T2, T2FLAIR, diffusion - weighted imaging (DWI), and post - contrast T1, are mainly used for GBM studies. In addition to subjective visual analysis of the images by radiologists, radiomics (shape, first, second, or higher - order statistics, and texture features) extracted from images, can also be applied to evaluate GBM patients' treatment outcomes. These extracted radiomics can serve as input data for machine learning (ML) models, contributing to predicting GBM patients' survival rates. This study aims to review various ML algorithms utilized for predicting GBM patients' survival rates. Methods: PubMed, Science Direct, Web of Science, and Google Scholar databases were explored up to November 2023, using different combinations of the keywords: "Glioblastoma multiforme ", "GBM", "survival prediction", "Magnetic resonance imaging", "radiomics", "artificial intelligence", "machine learning" and "deep learning". Finally, seven more recent and relevant records were included in the study. Results: Based on the results of the reviewed records, SVM (support vector machine), RF (random forest), LASSO (least absolute shrinkage and selection operator), XGBOOST, CATBOOST, LIGHTGBM, Nearest neighbor, Neural network, Multilayer perceptron (MLP), and Naïve Bayes models have been applied for GBM patients' survival rate prediction. The highest reported AUC (area under the curve) value was related to a Nearest neighbor - based model. The reported average AUC values for SVM and RF models were equal to 0.793 and 0.889, respectively. Furthermore, the RFE - SVM combined model presented the highest level of accuracy. The LASSO model also showed an appropriate performance in predicting GBM patients' survival rates. Moreover, Haralick features were the most significant applied radiomics with ML models. Conclusions: The high AUC and accuracy values in different studies for ML models predicting GBM patients' survival rates and treatment outcomes indicate the potential efficacy of such models as supportive tools for clinicians. These ML - based approaches using MRI - based extracted quantitative features, precisely predict GBM patient survival. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 23453192 |