Edge Detector-Based Hybrid Artificial Neural Network Models for Urinary Bladder Cancer Diagnosis

Abstract

In this chapter, a methodology for the utilization of edge-detector based hybrid artificial neural network (ANN) models for urinary bladder cancer is presented. The diagnosis of the bladder cancer can often be complex and it requires invasive diagnostic methods such as biopsy and histopatological evaluation. ANN utilization can provide faster and less invasive diagnosis. The methodology of ANN utilization for urinary bladder cancer diagnosis is based on images obtained with confocal laser endomicroscope during cystoscopy. Such approach can be challenging from the standpoint of computational resources, due to ANN model complexity. Higher computational resources are often inaccessible, especially in clinical practice. Here lies a motive for simplification of ANN models for urinary bladder cancer diagnosis. For these reasons, edge detector-based hybrid models are introduced due to their simpler architectures. From obtained results, it can be noticed that the highest performances are achieved with Laplacian-based convolutional neural network (CNN) model. On the other hand, such approach requires more complex CNN architectures in comparison to gradient-based hybrid CNN models. If Sobel edge detector is utilized, similar classification performances are achieved with less complex CNN model.