Automatic screening of lung diseases by 3D active contour method for inhomogeneous motion estimation in CT image pairs

Abstract

Lung diseases are now the third leading cause of death worldwide because of the many risk factors we are exposed to daily, such as air pollution, tobacco use, viruses (such as COVID-19), and bacteria. This work introduces a new approach of the 3D Active Contour Model (3D ACM) to estimate an inhomogeneous motion of lungs, which can be used to analyze lung disease patterns using a hierarchical predictive model. The biophysical model of lungs consists of End Expiratory (EE) and End Inspiratory (EI) models, generated by high-resolution computed tomography images (HRCT). A proposed technique uses the 3D ACM to estimate the velocity vector by using the corresponding points on the parametric surface model of the EE model to the EI model. The external energy from the EI models is the external force that pushes the 3D parametric surface to reach the boundary. The external forces, such as the balloon force and Gradient Vector Flow (GVF), were adjusted adaptively based on the Zratio which was calculated from the ratio of the maximum value of EI to EE on the Z axis. Next, the feature representation is studied and evaluated based on the lung structure, separated into five lobes. The stepwise regression, Support Vector Machine (SVM), and Artificial Neural Network (ANN) techniques are applied to classify the lung diseases into normal, obstructive lung, and restrictive lung diseases. In conclusion, the inhomogeneous motion pattern of lungs integrated with medical-based knowledge can be used to analyze lung diseases by differentiating normal and abnormal motion patterns and separating restrictive and obstructive lung diseases.