Closing gaps in early cancer detection
Closing gaps in early cancer detection lead image
Early cancer diagnosis can mean the difference between life and death. But standard detection procedures rely primarily on tissue biopsy and histological examinations, which are highly invasive, slow, subjective – and often ineffective.
An emerging alternative, liquid biopsy is based on single cell analyses and helps detect circulating tumor cells (CTCs) that detach from the primary tumor and appear in the bloodstream during early cancer stages. However, CTCs occur in only a very small fraction of cells, and accurately identifying them remains challenging. The most common techniques use CTC markers, which are cancer-specific and require a-priori information about type of tumor that might be unavailable.
Borelli et al. demonstrated the use of a holographic flow cytometer aided by artificial intelligence to discriminate ovarian cancer cells from monocytes in a marker-free mode.
“Holographic microscopy is unique in providing quantitative, label-free images of in vivo cells with high-throughput,” said author Vittorio Bianco.
The researchers let samples flow through a ‘lab-on-a-chip’ – a compact device embedding microfluidic circuits for precise control of tiny amounts of fluid.
“When cells flow under our microscope, we can access a rich content of biophysical information that is then used to feed an AI,” said Bianco.
“The AI must first be trained with known samples, but that is a one-time effort,” said author Pietro Ferraro. “Then, it can recognize CTCs in the cell stream with accuracies above 90 percent.”
Indeed, the study shows the setup outperformed other approaches, with gains in accuracy and reductions of false negatives.
“In this context, any accuracy improvement, although small, is highly relevant given the rarity of CTCs and the importance of their correct detection,” said author Natan T. Shaked.
Source: “AI-aided holographic flow cytometry for label-free identification of ovarian cancer cells in the presence of unbalanced datasets,” by F. Borrelli, J. Behal, A. Cohen, L. Miccio, P. Memmolo, I. Kurelac, A. Capozzoli, C. Curcio, A. Liseno, V. Bianco, N. T. Shaked, and P. Ferraro, APL Bioengineering (2023). The article can be accessed at https://doi.org/10.1063/5.0153413 .
