Recognizing and segmenting particles inside the respiratory system using texture descriptor and deep learning approaches

Abstract

The study of aerosol deposition and bronchial tube flows in the human respiratory system can assistance improve an understanding of the damaging or beneficial effects of the inhalation of lung aerosols. In this study, we propose a strategy for segmenting the decomposition of particles inside the respiratory system. Firstly, a texture descriptor method is used to represent more unique features for obtaining the border of each particle more accurately. Next, the original image and the encoded image are applied to a Convolutional Neural Network model to generate the edge map of the input image. Lastly, a circle fitting approach to compare each object with a lot of potential circles is employed to find the best match and recognize the object. A comparison of the results obtained in this study with some texture descriptor approaches is demonstrated the good performance of our model.