Identifying breast cancer subtypes at the single-cell level

Breast cancer is the leading cause of cancer death in women, and it is predicted that, by 2050, new cases will increase by 38% and deaths will increase by 68%.

Swift and early detection is key for preventing fatal outcomes, so Lin et al. developed a machine learning-enhanced dynamic light scattering (DLS) technique to automatically detect breast cancer subtypes.

“Dynamic light scattering detects how light scattered by a sample changes over time,” said author Xiangyuan Ma. “In our system, a laser beam illuminates an individual cell, and the scattered light is recorded as a time-series image. Because intracellular components, such as organelles and other subcellular structures, are constantly moving in viable cells, the scattering pattern fluctuates over time.”

The key to this identification is movement within the cell, since different breast cancer subtypes can be identified by intercellular movement — something that can be identified with DLS. Raw DLS images are often noisy and can be difficult to analyze manually. The researchers’ deep-learning algorithm automatically separates the spatial data — the scattering pattern of the cell — and the temporal data — the internal motion of the cell.

While other single-cell diagnostics exist, this technique is faster and less invasive, removing the staining or labeling steps.

“It does not require fluorescent markers, antibodies, or destructive staining procedures; instead, it uses the cell’s intrinsic light-scattering dynamics to capture biophysical information,” Ma said. “Therefore, we consider it a complementary approach for rapid and automated single-cell analysis with minimal sample preparation.”

In the future, the researchers plan to diversify their samples and integrate DLS analyzers with microfluidic platforms to automate the process.

Source: “A deep-learning-enhanced dynamic light scattering analyzer for fast, label-free and automatic classification of breast cancer subtypes,” by Meiai Lin, Yixiong Zheng, Cong Xiao, Lijun Yang, Haodong Lv, and Xiangyuan Ma, APL Photonics (2026). The article can be accessed at https://doi.org/10.1063/5.0335530 .