Neuromorphic Photonic Spiking Neural Network for Medical Data Classification Based on VCSEL-SA

Abstract

Neuromorphic photonic computinghas emerged as a transformative approach in information processing and artificial intelligence, leveraging its inherent advantages of high bandwidth and low latency. In this study, we present an energy-efficient photonic spiking neural network (PSNN) that utilizes a threshold- based encoding scheme to convert medical data into optical stimulus pulses. These pulses are processed through vertical cavity surface-emitting lasers with an embedded saturable absorber (VCSELs-SA), effectively emulating biological synaptic transmission for classification tasks. The network is trained using the Remote Supervised Method (ReSuMe), a supervised learning algorithm that optimizes synaptic weights to align the output spike and the target spike. Combined with Spike Timing-Dependent Plasticity (STDP), the system demonstrates exceptional classification performance on benchmark medical datasets, achieving training set accuracies of 95.6% and 97.9% on the Pima Indian Diabetes and Wisconsin Breast Cancer datasets, respectively. These results not only validate the efficacy and reliability of our proposed PSNN but also underscore its potential for advanced medical data analysis. To the best of our knowledge, this study represents the first application of PSNN for medical dataset classification. Furthermore, this work expands the application scope of PSNN, highlighting their promise for broader adoption in the medical field as photonic computing technology continues to evolve.