Fingerprints of cancer heterogeneity predict disease outcome

Cancer diagnosis is a major medical challenge because current methods are time-consuming, manual and limited in their accuracy. We have developed an AI-based image analysis approach to analyse tumours and surrounding tissue at single cell resolution. Using patient-specific cellular 'fingerprints', we have shown that the combination of certain features indicates poorer prognosis and resistance to therapy. These results point to new approaches for improved diagnosis and therapy.