Radiomics is a rapidly evolving field that extracts quantitative metrics from medical images to analyze tissue and lesion characteristics. This article introduces the basic radiomics workflow, including feature calculation, selection, dimensionality reduction, and data processing. It discusses potential clinical applications in nuclear medicine, such as PET-based prediction of treatment response and survival. The article also covers current limitations, such as sensitivity to acquisition parameter variations, and common pitfalls in radiomics studies. Radiomic features are categorized into statistical, model-based, transform-based, and shape-based features. Statistical features include histogram-based and texture-based features, while model-based features involve analyzing spatial frequency of intensity changes. Transform-based features use methods like Fourier, Gabor, and Haar wavelet transforms to analyze gray-level patterns. Shape-based features describe geometric properties of regions of interest (ROIs). Acquisition parameters and feature standardization are critical for radiomics, as image-derived metrics like SUVs and radiomic features are sensitive to image acquisition settings, reconstruction algorithms, and image processing. Studies have shown that spatial resolution has the strongest effect on radiomic features, followed by scan duration, segmentation method, and reconstruction method. To minimize feature variations, small voxels, narrow Gaussian postfiltering, and point-spread function modeling are recommended. Feature harmonization is a postprocessing technique to remove batch effects on radiomic features. Feature selection and dimensionality reduction are essential to identify relevant features for clinical problem solving. Dimensionality reduction techniques like principal-component analysis are popular but may mix variables and complicate subsequent tracing of predictors. Once radiomic features are selected, they are used to predict target variables, such as disease presence or treatment response. Machine learning models, including support vector machines, random forests, and neural networks, are used to analyze radiomic features. These models can learn the relationship between high-dimensional inputs (radiomic features) and target variables based on training examples. However, overfitting and underfitting are common pitfalls in radiomics studies, and careful validation is necessary to ensure model generalizability. Radiomics has shown promise in clinical applications, such as predicting survival in non–small cell lung cancer patients and differentiating tumor grades in glioma patients. Radiogenomics, which links imaging data to biological data, has also been explored. Radiomics can provide complementary information about tumor heterogeneity, potentially improving survival prediction and patient stratification. In conclusion, radiomics is a sophisticated image analysis technique with potential in precision medicine. However, standardized image acquisition and reconstruction protocols are vital for its development and application.Radiomics is a rapidly evolving field that extracts quantitative metrics from medical images to analyze tissue and lesion characteristics. This article introduces the basic radiomics workflow, including feature calculation, selection, dimensionality reduction, and data processing. It discusses potential clinical applications in nuclear medicine, such as PET-based prediction of treatment response and survival. The article also covers current limitations, such as sensitivity to acquisition parameter variations, and common pitfalls in radiomics studies. Radiomic features are categorized into statistical, model-based, transform-based, and shape-based features. Statistical features include histogram-based and texture-based features, while model-based features involve analyzing spatial frequency of intensity changes. Transform-based features use methods like Fourier, Gabor, and Haar wavelet transforms to analyze gray-level patterns. Shape-based features describe geometric properties of regions of interest (ROIs). Acquisition parameters and feature standardization are critical for radiomics, as image-derived metrics like SUVs and radiomic features are sensitive to image acquisition settings, reconstruction algorithms, and image processing. Studies have shown that spatial resolution has the strongest effect on radiomic features, followed by scan duration, segmentation method, and reconstruction method. To minimize feature variations, small voxels, narrow Gaussian postfiltering, and point-spread function modeling are recommended. Feature harmonization is a postprocessing technique to remove batch effects on radiomic features. Feature selection and dimensionality reduction are essential to identify relevant features for clinical problem solving. Dimensionality reduction techniques like principal-component analysis are popular but may mix variables and complicate subsequent tracing of predictors. Once radiomic features are selected, they are used to predict target variables, such as disease presence or treatment response. Machine learning models, including support vector machines, random forests, and neural networks, are used to analyze radiomic features. These models can learn the relationship between high-dimensional inputs (radiomic features) and target variables based on training examples. However, overfitting and underfitting are common pitfalls in radiomics studies, and careful validation is necessary to ensure model generalizability. Radiomics has shown promise in clinical applications, such as predicting survival in non–small cell lung cancer patients and differentiating tumor grades in glioma patients. Radiogenomics, which links imaging data to biological data, has also been explored. Radiomics can provide complementary information about tumor heterogeneity, potentially improving survival prediction and patient stratification. In conclusion, radiomics is a sophisticated image analysis technique with potential in precision medicine. However, standardized image acquisition and reconstruction protocols are vital for its development and application.