Radiomics Analysis to Enhance Precise Identification of Epidermal Growth Factor Receptor Mutation Based on Positron Emission Tomography Images of Lung Cancer Patients

Abstract

How to recognize precisely epidermal growth factor receptor (EGFR) mutation in lung cancer patients owns great clinical requirement. In this study, 1575 radiomics features were extracted from PET images of 75 lung cancer patients based on contrast agents such as 18F-MPG and 18F-FDG. The Mann-Whitney U test was used for single factor analysis, the Least Absolute Shrinkage and Selection Operator (Lasso) Regression was used for feature screening, then the radiomics classification models were established by using support vector machines and ten-fold cross-validation, and were used to identify EGFR mutation in primary lung cancers and metastasis lung cancers, accuracy based on 18F-MPG PET images are respectively 90% for primary lung cancers, and 89.66% for metastasis lung cancers, accuracy based on 18F-FDG PET images are respectively 76% for primary lung cancers and 82.75% for metastasis lung cancers. The area under the curves (AUC) based on 18F-MPG PET images are respectively 0.94877 for primary lung cancers, and 0.91775 for metastasis lung cancers, AUC based on 18F-FDG PET images are respectively 0.87374 for primary lung cancers, and 0.82251 for metastasis lung cancers. In conclusion, both 18F-MPG PET images and 18F-FDG PET images combined with established classification models can identify EGFR mutation, but 18F-MPG PET images have more precisely than 18F-FDG PET images, own clinical translational prospects.

Application of artificial intelligence in cancer diagnosis and tumor nanomedicine

“…Their research adds valuable insights into subtype-specific diagnosis and classification. In 2021, Li et al 123 extracted radiomic features from PET images to identify EGFR mutations in lung cancer, achieving high accuracy. Chaunzwa et al 124 proposed a radiomic method for predicting tumor histology in non-small cell lung cancer (NSCLC) from standard of care CT data, enhancing diagnostic precision.…”

The predictive value of [18F]FDG PET/CT radiomics combined with clinical features for EGFR mutation status in different clinical staging of lung adenocarcinoma

“…Compared to SUV max , radiomics features can better reflect the spatial distribution of tumors and more comprehensively evaluate tumor heterogeneity. In recent years, using machine learning methods with high prediction efficiency and strong feasibility to assess radiomic features and predict EGFR mutation status has become a research "hot spot" [5,[19][20][21][22][23][24][25][26][27][28][29][30][31][32]. However, most of these studies had small sample sizes, with a total sample number of no more than 200 cases [19][20][21][22][23][24][25][26], and the TNM stages of the enrolled patients varied greatly [21,27,31].…”

Prediction of subsequent osteoporotic vertebral compression fracture on CT radiography via deep learning

“…Now, big data analysis based on deep learning has been applied in the diagnosis of various complex and difficult diseases 14–16 and accurate parameter analysis 17–19 . The radiomics classification model, which was established based on radiomics features from PET images via support vector machines, could accurately identify EGFR mutation in primary lung cancers and metastasis lung cancers 20 . Also, amino acid biomarkers in saliva can also be used to detect gastric cancer through artificial intelligence analysis at very high accuracy 21 .…”

“…[17][18][19] The radiomics classification model, which was established based on radiomics features from PET images via support vector machines, could accurately identify EGFR mutation in primary lung cancers and metastasis lung cancers. 20 Also, amino acid biomarkers in saliva can also be used to detect gastric cancer through artificial intelligence analysis at very high accuracy. 21 A deep learning-based method could achieve fully automatic identification of osteoporosis, osteopenia, and normal BMD in CT images.…”