An Ensemble of Deep Semantic Representation for Medical X-ray Image Classification
Mohammad Reza Zare, Mehdi Mehtarizadeh · 2021
An efficient medical image classification system has gained high interest in the scientific community. This paper presents a classification algorithm that aims to gain a high accuracy rate by addressing some of the typical challenges involved in classification of large medical datasets. In this paper, the convolutional neural networks (CNNs) are employed together with probabilistic latent semantic analysis (PLSA) which are capable of mining hidden semantics of images. This high-level semantic representation of the images is then fed into a discriminative support vector machine (SVM) to build a classification model. An ensemble of machine learning models is also employed to utilize the capability of classification models created from different sets of data. The evaluation is based on a medical image dataset consisting of 11,000 X-ray images from 116 distinct categories. The classification accuracy rate obtained by the proposed classification model is 94.5 %. The results show that the proposed classification model outperformed the methods in the literature evaluated on the same benchmark dataset.