Radiomics is the practice of turning medical images such as CT, MRI, PET, ultrasound, or digital pathology slides into quantitative features that can be analyzed as biomarkers. Instead of relying only on what the human eye can describe, radiomics measures texture, shape, intensity, spatial pattern, and other image-derived properties that may correlate with diagnosis, prognosis, or treatment response.
Why It Matters
Radiomics matters because imaging is often already part of routine care. If the images can be converted into reproducible measurements, they may provide a noninvasive way to estimate tumor biology, inflammatory activity, progression risk, or likely therapy response without adding a separate invasive test.
That is why radiomics is often discussed as a form of imaging biomarker discovery. It can help bridge the gap between a scan that looks informative and a measurement that can actually be tested, compared, and validated across cohorts.
Where AI Fits
AI helps radiomics by handling segmentation, feature extraction, multimodal fusion, and prediction from large image datasets. It also helps link image features to outcomes, pathology, or molecular markers, which is why radiomics often overlaps with computer vision and multimodal learning.
At the same time, radiomics is sensitive to scanner settings, acquisition protocols, annotation quality, and cohort shift. Good radiomics work therefore depends on calibration, strong ground truth, and explicit handling of uncertainty.
Related Yenra articles: Biomarker Discovery in Healthcare, Arthritis Progression Modeling, Cancer Treatment Planning, Patient Outcome Prediction, Precision Oncology and Targeted Therapies, and Personalized Medicine.
Related concepts: Computer Vision, Multimodal Learning, Calibration, Ground Truth, and Uncertainty.