Radiomics-Based Chest Disease Classification Using Machine Learning and Medical Image Processing

Abstract

Chest diseases are a major killer throughout the world and the early and accurate diagnosis is fundamental to the treatment provided. Chest medical images, especially computed tomography (CT) images, have greatly advanced the field of computer aided diagnosis in recent years by computerized image analysis including radiomics and machine learning. In this paper, a chest disease classification framework based on radiomics is proposed to classify multiple chest diseases through image preprocessing, feature extraction, feature selection and supervised machine learning algorithms. Medical image processing techniques are used to process the imaging in order to improve its quality, while radiomic features are utilized to provide a detailed texture, shape and intensity characteristic analysis. The experimental results also prove that the proposed framework has the potential of enhancing the classification accuracy and helping radiologists make more efficient and accurate clinical decisions in less time.