A Skin Cancer Detector Using a Spiking Neural Network Based on All-Digital Resonate-and-Fire Neurons

Diseases of the skin are becoming the most popular health issues due to the changing weather. One of the deadliest types is skin cancer. Skin cancer generally metastasizes slowly to other parts of the body; early detection is the best step in treatment. There were many published studies to develop early detection methods. The combination of visual computing and Artificial Neural Networks (ANN) is the most common method to help a patient quickly detect diseases at home. Symmetry, color, size, and shape of the lesion are some of the features that are used to diagnose skin cancer. This study did a brief comparison among some ANN methods for detecting skin cancers. Most of these ANNs are based on Multiple Layer Peceptron (MLP) networks, which have high power consumption and are being slowly replaced by Spiking Neural Networks (SNN). By that, the study proposed a design for a simple SNN using 288 AII-Digital Resonate-And-Fire (ADRAF) neurons that allows for unsupervised on-chip training. The design was simulated on Verilog and implemented on a 28-nm FPGA device to evaluate its resource usage.