Using machine learning and optical imaging to identify malignant tumor cells

While many types of cancer can be deadly as local tumors, they become significantly more dangerous when they metastasize, sending cells around the body to form new tumors elsewhere. Locating and identifying circulating tumor cells (CTCs) is the first step toward treating any type of metastasized cancer.

Pirone et al. described a hybrid approach to identifying CTCs in blood samples using label-free optical imaging, microfluidics, and machine learning. Their approach can identify and phenotype neuroblastoma cells amid white blood cells in a fast, effective, and non-invasive manner.

Identifying CTCs and distinguishing them from white blood cells is often challenging for liquid biopsy methods.

“Most technologies are label-dependent, and the limitations are due to the lack of universal and specific cell-surface markers,” said author Lisa Miccio. “Furthermore, methods based on filtering by dimensions are not effective because CTCs are comparable in size with white blood cells.”

Instead, the authors employed a three-pronged approach to identify CTCs. First, they sent the samples through a microfluidic channel to induce rotation in the cells. Then, they created a three-dimensional reconstruction of each cell by combining optical images using digital holography. Lastly, machine learning algorithms trained on these images rapidly and accurately identified CTCs based on morphological features.

“Our upcoming research endeavors involve the practical application of this technology on patient-derived samples, thereby demonstrating the robustness of the method in a real-world clinical setting,” said author Mario Capasso. “Our primary objective in these new research directions is to refine the methodology for classifying neuroblastoma cell populations as either resistant or non-resistant to cancer therapies.”

Source: “Phenotyping neuroblastoma cells through intelligent scrutiny of stain free biomarkers in holographic flow cytometry,” by Daniele Pirone, Annalaura Montella, Daniele Sirico, Martina Mugnano, Danila Del Giudice, Ivana Kurelac, Matilde Tirelli, Achille Iolascon, Vittorio Bianco, Pasquale Memmolo, Mario Capasso, Lisa Miccio, and Pietro Ferraro, APL Bioengineering (2023). The article can be accessed at https://doi.org/10.1063/5.0159399 .