A team has developed plasmonic biosensors based on X-shaped resonators capable of detecting cancer cells, combining the optical response of the device with machine learning algorithms that predict the system's transmittance. The approach couples the sensitivity of plasmonic nanostructures to changes in their environment with models able to anticipate the sensor's spectral behaviour.

The X-shaped resonators support surface plasmon modes whose transmittance shifts when cells are deposited on the structure. This dependence on the local refractive index is the basis of detection, but interpreting the resulting spectra quantitatively has until now required costly fitting.

Machine learning is used to predict transmittance from the device and sample characteristics, bypassing part of the conventional spectral analysis. The trained model reproduces the optical response of the biosensor and makes it possible to identify the signal associated with the presence of tumour cells.

The combination of plasmonic nanophotonics and machine learning points toward faster, more compact diagnostic platforms, although clinical validation and comparison with established methods remain outstanding steps.