Machine learning assisted cancer cell detection using strip waveguide Bragg gratings

An early cancer diagnosis has become necessary in the medical domain as it can facilitate a timely and appropriate treatment, providing for the clinical management of patients. This work proposes a rapid and inexpensive label-free bio-sensing platform based on Waveguide Bragg Gratings for cancer detection. The refractive index values of cancer cells are different from normal cells, which provide a different reflection spectrum when the sample is placed on the functionalized surface of the waveguide grating structure. We combine simulations from the effective index method, finite element method, and Transfer Matrix method of the Grating structure to optimize the sensing structure and employ a neural network model to predict whether a sample contains cancerous cells. The machine learning model provides excellent accuracy for high-resolution interrogator data, and the accuracy falls below 95% only when the resolution of the interrogator is taken as 0.9 nm, thereby eliminating the need for an expensive high-resolution optical Interrogator system. This, combined with the mass-scale production capability of the waveguide Bragg gratings using standard CMOS technology, paves the way towards combining nano-scale optical biosensing platforms and machine learning-based optimization techniques for early and low-cost label-free cancer detections.