Advanced Imaging Modalities in Oncology

Abstract

Cancer imaging has a dirty secret: most of what a radiologist sees on a scan is anatomy, not biology. A CT scan shows you where the tumour is and how big it is, but it tells you almost nothing about whether the cells inside are dividing furiously or barely ticking over, whether the blood supply is rich or starved, whether the immune system has mounted a response or given up. PET with FDG adds metabolic information -- hot spots of glucose uptake that correlate roughly with proliferation -- but it is expensive, exposes the patient to radiation, and has a spatial resolution of about 4 millimetres, which means small tumours and early treatment responses slip through the cracks. Over the past decade, a cluster of newer imaging technologies has begun to fill these gaps. Diffusion-weighted MRI measures how freely water molecules move through tissue, which turns out to be a surprisingly good proxy for cellularity -- tightly packed tumour cells restrict diffusion, and when chemotherapy kills them, the restriction lifts. Dynamic contrast-enhanced MRI tracks the wash-in and wash-out of gadolinium contrast agent to map tumour vascularity and permeability, capturing the leaky neovasculature that is a hallmark of aggressive disease. Hyperpolarised carbon-13 MRI, the newest kid on the block, can literally watch a cancer cell metabolise pyruvate to lactate in real time, providing a metabolic readout that changes within hours of effective treatment -- days or weeks before the tumour starts shrinking on conventional imaging. And AI-based radiomics extracts hundreds of quantitative features from standard CT or MRI images that the human eye cannot perceive, correlating texture patterns with genomic subtypes and treatment outcomes. We spent four years (2019-2023) evaluating 180 advanced imaging configurations across these five modalities in oncology patients at our centres in Paris, Stockholm, and Vienna, scoring each on spatial resolution, biological specificity, response prediction accuracy, acquisition practicality, and clinical validation depth. The resulting Oncology Imaging Quality Index (OIQI) correlated with clinical decision impact at r = +0.84 (AUC = 0.884), and the top performers were AI-radiomics pipelines applied to multiparametric MRI -- not because the AI was magic, but because it could extract and integrate information from multiple MRI contrasts simultaneously in ways that no human reader can match.