Hybrid Multi-Layer Perceptron and Enhanced Edge-Free Active Contour Model Lung Nodule Detection Algorithm

Abstract

A computer-assisted method is usually used to identify pneumoconiosis by segmenting and classifying the initial area of interest. The accuracy of current techniques for segmenting and classifying lung nodules is often suboptimal. Therefore, a hybrid Multi-Layer Perceptron (MLP) and enhanced edge-free active contour model (CV) algorithm are proposed to solve this problem. First, lung nodules are identified by the MLP to determine their locations and delineate their initial boundaries. CV is then used to automate segmentation and detect lung nodules quickly and accurately. Based on the LUNA16 dataset, the algorithm is trained and evaluated for more than 100 epochs, with 17 epochs achieving perfect accuracy. It is worth noting that only running the third epoch achieves 100% accuracy, proving its efficiency and effectiveness. The proposed method has greater accuracy than the comparing me.