Get Started with Radiomics

Radiomics is an emerging field in medical imaging that enables quantitative analysis of image data for applications such as oncology, personalized medicine, and clinical outcome prediction.

The image biomarker standardisation initiative (IBSI) provides standardized nomenclature and definitions for radiomics features, a standard procedure for medical image preprocessing, and reporting guidelines, among other standardization tools. This standardization of radiomics features ensures the reproducibility and enables the validation of radiomics studies.

What Is Radiomics?

Radiomics is a technique in medical imaging for extracting mineable data from medical images to decode the radiographic phenotype. By computing standardized features related to shape, intensity, and texture, radiomics provides a non-invasive approach for identifying imaging biomarkers that complement traditional diagnostic methods. Radiomics features include shape descriptors such as volume and surface area, first-order intensity statistics like mean and entropy, and texture measures derived from gray-level co-occurrence and run-length matrices. Together, these features capture both macroscopic and microscopic patterns within the region of interest (ROI), enabling you to quantitatively characterize tissue heterogeneity.

In modern healthcare and medical research, radiomics has applications in oncology for tumor characterization, treatment response prediction, and prognosis modeling. Beyond oncology, radiomics aids in personalized medicine, biomarker discovery, and outcome prediction across multiple imaging modalities, including CT, MRI, PET, and ultrasound. This versatility makes radiomics a powerful tool for bridging imaging and computational analysis.

Radiomics is particularly useful for studying tumor heterogeneity, which refers to variations in tissue characteristics within a tumor that are often invisible to the human eye. Quantifying these variations can improve risk assessment and treatment planning. You can also integrate radiomics features into machine learning and deep learning pipelines for tasks such as classification, prediction of clinical outcomes, and biomarker discovery, making radiomics a key component in data-driven healthcare and precision medicine.

Standardization in Radiomics

Standardization is essential for radiomics to deliver reliable and reproducible results. The image biomarker standardisation initiative (IBSI) provides guidelines for feature definitions, image preprocessing, and reporting practices, ensuring consistency across studies. These guidelines address critical challenges such as feature harmonization and reproducibility, which are necessary to reduce the variability introduced by different scanners, acquisition protocols, and preprocessing steps.

Despite its uses, radiomics faces challenges such as standardizing features across scanners and managing variability in image acquisition. Adhering to IBSI standards helps overcome these challenges and ensures that radiomics computations produce consistent and validated results suitable for clinical and research applications.

Typical Radiomics Application Process

The typical process of a radiomics application involves these steps.

Import Medical Image

Radiomics begins with acquiring and importing medical images into your workspace. These images can come from various modalities such as CT, MRI, PET, or ultrasound. For accurate feature computation, ensure that importing the image preserves the image metadata and voxel dimensions for accurate feature computation.

Import the medical image into the workspace as a 2-D matrix, medicalImage object, 3-D array, or medicalVolume object. You can compute radiomics features from medical images of any modality, such as MRI, CT, or ultrasound. For more information, see Medical Imaging Modalities. Medical Imaging Toolbox™ supports DICOM, NIfTI, NRRD, and other file formats. For more information, see Import and Spatial Referencing.

Preprocess Medical Image

Preprocessing improves image quality and consistency by removing artifacts and normalizing intensity values. Common steps include background removal, denoising, registration, augmentation, and intensity normalization. For more information, see Medical Image Preprocessing and Medical Image Registration.

Identify Region of Interest (ROI)

To identify an ROI, you must segment your medical image. Accurate segmentation of your ROI is important because further processing requires you to use radiomics features from the entire ROI. Segment the ROI from the medical image using manual or automated segmentation methods.

If you have already identified the ROI, import the mask of the ROI as a 2-D matrix, medicalImage object, 3-D array, or medicalVolume object. If you have not identified the ROI, Medical Imaging Toolbox provides the Medical Image Labeler app, the Medical Segment Anything Model (MedSAM), and various functions for performing medical image segmentation. For more information, see Segmentation and Get Started with MedSAM in Medical Image Labeler.

Preprocess Medical Image for Radiomics

Before feature extraction, IBSI recommends additional preprocessing steps such as resampling to a uniform voxel size, resegmentation to refine ROI boundaries, and discretization of intensity values. Use the radiomics object to preprocess your medical image as required by IBSI standards.

Radiomics Feature Computation

Radiomics features quantify the shape, intensity, and texture characteristics of an ROI. These features form the basis of statistical analysis and predictive modeling.

Use the shapeFeatures, intensityFeatures, and textureFeatures object functions of the radiomics object to compute IBSI-standardized radiomics features related to the shape, intensity, and texture of the region of interest, respectively. Use the selectFeatures object function of the radiomics object to compute selected features. The radiomics features include shape descriptors such as volume and surface area, first-order intensity statistics like mean and entropy, and texture measures derived from gray-level co-occurrence and run-length matrices. For more information on all the IBSI-standardized features that you can compute, see IBSI Standard and Radiomics Function Feature Correspondences.

Postprocessing

You can apply statistical methods to the computed radiomics features to identify associations between medical imaging features and patient biology, or apply machine learning or deep learning models to predict clinical outcomes. For an example of clinical prediction using radiomics, see Classify Breast Tumors from Ultrasound Images Using Radiomics Features.