Machine learning system can identify cancer cells based on how they scatter light
Cytological tests are a common method of screening for cancer cells in stained cell samples. Using a microscope, pathologists examine cells collected from bodily fluids, looking for telltale signs of malignancy, like enlarged nuclei or abnormal cell shapes. Owing to their minimally invasive nature, these tests are widely used for early cancer screening and diagnosis.
Researchers from Nara Institute of Science and Technology (NAIST) and Kindai University Faculty of Medicine in Japan have developed a machine learning system that can accurately identify cancer cells based on how they scatter light, according to an Aug. 3, 2026 study in Scientific Reports. Conventional cytology tests for early cancer screening rely on the expertise of pathologists, who examine stained cell samples under a microscope to detect signs of malignancy.
However, cancerous cells can sometimes appear identical to normal cells at the resolution of conventional microscopes, requiring further differentiation. Professor Yoichiroh Hosokawa, Ryohei Yasukuni, and Akihiko Ito proposed using dark-field microscopy to capture the light scattered by cells instead of light absorbed while passing through them.
They examined cytology specimens containing cancerous mesothelioma cells and reactive mesothelial cells, which often appear similar, and recorded their light scattering patterns in the visible spectrum (420-720 nm). Machine learning algorithms, including principal component analysis and support vector machine classification, were then applied to the data to differentiate between cell types.
The resulting system achieved about 91% accuracy in distinguishing mesothelioma cells from reactive mesothelial cells during patient-based validation. The study demonstrates that light scattering spectra contain valuable submicron-scale information that can help distinguish between different cancer cell types, even when visual inspection or traditional image analysis is insufficient.
The researchers envision integrating this spectroscopic approach with conventional cytology to complement pathologists' skill and judgment, potentially improving diagnostic accuracy. Further work is underway to optimize the optical settings and machine learning methods for enhanced performance.
Written by urgent.news from Medical Xpress's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
