Pulmonary Image Recognition and Respiratory Feature Analysis Under Neural Network and Genetic Algorithm

Abstract

It was to study the recognition performance of the fusion of neural network and genetic algorithm for pulmonary images, and to realize the diagnosis of pulmonary diseases by recognizing the respiratory sound signals. Pulmonary computerized tomography (CT) images were selected as the data base, and the genetic algorithm was applied to achieve fast global optimal search. On the combination of neural network and genetic algorithm, an improved genetic intelligent algorithm model was put forward. The simulation experiments were performed to compare the performances such as the algorithmic rate, accuracy, and sensitivity, so as to verify the superiority of the model. Then, the proposed algorithm was used to verify its effectiveness by collecting the respiratory sound signals of related diseases. The genetic algorithm could not only obtain the global optimal solution, but also greatly shorten the calculation time. With the pulmonary CT images, the complete segmentation of the pulmonary airways and the recognition of pulmonary images could be achieved. The algorithm could effectively recognize respiratory sound signals of health people and patients with chronic obstructive pulmonary diseases (COPD) and pneumonia. Its accuracy reached 0.943, with a precision of 0.921 and a recall rate of 0.931. It allowed to achieve the goal of diagnosing pulmonary diseases by respiratory sound signals. The fusion of neural network and genetic algorithm could realize pulmonary image recognition, and the diagnosis of pulmonary diseases could also be diagnosed through the feature analysis of respiratory sound signals.