AI Enhances Liquid Biopsies by Spotting Elusive Tumour Cells

Liquid biopsies provide a non-invasive method for managing cancer by detecting Circulating Tumour Cells (CTCs) in the blood. However, identifying these cells is notoriously difficult; they are rare, heterogeneous, and often hidden amongst clusters of other cells. Traditional methods rely on manual analysis or fluorescence labelling, which can be inconsistent across different hospitals and time-consuming to process.

To address this, researchers have developed a Deep Learning (DL) classification pipeline designed to distinguish CTCs from leukocytes (white blood cells). The team utilised a ResNet-based Convolutional Neural Network to analyse images acquired via DEPArray technology. A key innovation involved training the model using both data augmentation and fluorescence (DAPI) images to teach it specific cellular features. Crucially, however, the system was tested using only bright-field images, removing the reliance on fluorescent markers for the final diagnosis.

The model achieved an F1-score of 0.798, proving it can effectively identify these elusive biomarkers. This automated approach promises to reduce variability and significantly optimise clinical workflows for cancer patient management.