Researchers have recently achieved a significant breakthrough in all-optical cell classification using a single-layer diffractive neural network (DNN). This in-silico demonstration successfully differentiated between breast cells, lung cancer cells, and white blood cells. Consequently, this technology offers a glimpse into a future where medical diagnostics operate at the speed of light with minimal energy consumption.

The team virtually implemented the DNN using a spatial light modulator (SLM). Unlike traditional electronic artificial intelligence that processes data through silicon chips, this system uses optical diffraction to perform calculations. Specifically, the researchers trained the network using backpropagation on phase and amplitude images obtained from optofluidic time-stretch quantitative phase imaging. Therefore, the computation occurs physically as light passes through the modulator, eliminating the latency found in digital systems.

Advantages of All-optical Cell Classification

The optimized DNN achieved a remarkable 96.1% accuracy, nearly matching the performance of conventional convolutional neural networks. Furthermore, this method is exceptionally energy-efficient because the processing is passive. It removes the need for power-hungry digital-to-analog conversions usually required in medical imaging. Because of these benefits, the system could revolutionize high-throughput screening in clinical laboratories and pathology departments.

While this remains a simulation study, the results highlight the massive potential for SLM-based DNNs in real-world biomedical image processing. As data volumes in oncology and hematology continue to grow, light-based computing may provide the necessary speed to keep up with diagnostic demands. Future developments may lead to real-time, label-free cell sorting during routine clinical procedures.

Frequently Asked Questions

How does all-optical cell classification differ from traditional AI?

Traditional AI relies on electronic transistors and digital signal processing, which can be slow and energy-intensive. All-optical classification uses light diffraction through physical layers to process information. This allows for near-instantaneous results at the speed of light.

Which cell types were successfully identified in this study?

The study demonstrated that the diffractive neural network could accurately differentiate between lung cancer cells, breast cells, and white blood cells. This versatility suggests it could be adapted for a wide range of hematologic and oncologic diagnostic applications.

Disclaimer: This content is for informational and educational purposes only. It does not constitute medical advice or a substitute for professional healthcare consultation. Refer to the latest local and national guidelines for clinical practice.

References

Sagami N et al. All-optical classification of real biomedical cell images using a diffractive neural network: a simulation study. Opt Lett. 2026 Apr 01. doi: 10.1364/OL.593231. PMID: 41920665.

Lin X, et al. All-optical machine learning with diffractive deep neural networks. Science. 2018;361(6406):1004-1008.

Rahman S, et al. Deep learning-based quantitative phase imaging for cancer cell classification. J Biomed Opt. 2023;28(1):016002.