AI and biomarkers

AI links imaging patterns to molecular biology, revealing non-invasive “imaging biomarkers” that guide decisions without extra biopsies.

A biomarker is a measurable biological signal — a gene, a protein or a pattern — that tells doctors about a tumour. Tissue analysis is the gold standard, but tissue requires a biopsy. AI learns to find biomarkers in images and blood signals instead.

Radiomics, for example, extracts hundreds of quantitative features from a scan — texture, shape, intensity distribution — and links them to tumour genetics. The promise: a “virtual biopsy”, rich biological information from images the patient is already receiving.

Biomarkers AI can reveal

  • Tumour aggressiveness from imaging texture
  • Receptor status predicted from scans
  • Molecular subtype classification
  • Response prediction before treatment begins

Virtual biopsies are a research frontier. Today they complement — they do not replace — tissue diagnosis.

In brief

Radiomics extracts hundreds of features from a single scan.
AI can link imaging patterns to tumour genetics.
Virtual biopsies could reduce the need for invasive sampling.

Keep learning