Applying Multi-CNNs model for detecting abnormal problem on chest x-ray images

Image diagnosis is the significant problem in medicine. Nowadays, with modern facilities that allow doctors to diagnose early and accurately disease, limiting unnecessary treatment procedures. By that way, the image diagnosis is at the forefront of the processing diagnosis and treatment of the disease. Heart and lung failure accounts for more than 500,000 deaths annually in the United States and is most commonly screened for using plain film chest X-Ray (CXR). With the growing number of patients, the doctors must overwork, so he cannot counsel and direct take care of his patient. So, a computer system that supports image classification is needed. In this paper, we propose a deep learning model to detect abnormal sentisy in chest x-ray images. The proposed model uses multiple Convolutional Neural Network to decide input image, this is called Multi-CNNs. Input data is the digital chest X-ray image dataset that was collected from 6/2017 to 3/2018 at An Binh Hospital, HCM, VN (AB-CXR-Database). Each component of the Multi-CNN is a convolutional neural network that is developed base on ConvnetJS library. The output of the proposed model is Normal/Abnormal density. In this paper, we also propose a method for synthesizing the results of the components of the model which we are called Fusion rules. The experimental results 96% in our x-rays image dataset showed the feasibility of a proposed Multi-CNNs model.