Dual-Ray Net: Automatic Diagnosis of Thoracic Diseases Using Frontal and Lateral Chest X-rays

Abstract

Computer-aided diagnosis (CAD) is an important work which can improve the working efficiency of physicians. With the availability of large-scale data sets, several methods have been proposed to classify pathology on chest X-ray images. However, most methods report performance based on a frontal chest radiograph, ignoring the effect of the lateral chest radiography on the diagnosis. This paper puts forward a kind of model, Dual-Ray Net, of a deep convolutional neural network which can deal with the front and lateral chest radiography at the same time by referring the method of using lateral chest radiography to assist diagnose during the diagnosis used by radiologists. Firstly, we evaluated the performance of parameter migration to small data after pre-training for large datasets. The data sets for pre-training are chest X-ray 14 and ImageNet respectively. The results showed that pre-training with chest X-ray 14 performed better than with the generic dataset ImageNet. Secondly, We evaluated the performance of the Frontal and lateral chest radiographs in different modes of input model for the diagnosis of assisted chest disease. Finally, by comparing different feature fusion methods of addition and concatenation, we found that the fusion effect of concatenation is better, which average AUC reached 0.778. The comparison results show that whether it is a public or a non-public dataset, our Dual-Ray Net (concatenation) architecture shows improved performance in recognizing findings in CXR images when compared to applying separate baseline frontal and lateral classes.

School of Mechanical Engineering, Shanghai Jiao Tong University, 200240, Shanghai, China

“…The approaches can be classified into two types: model-based and input-based. In model-based approaches, one idea is to design models by imitating the radiological practice [34,21,55,26] or change the model structure based on diagnostic patterns [42,20,15]. While in input-based approaches, the knowledge is treated as external inputs to calculate features [63,62,52] or to guide the final loss [12,27].…”

MedKLIP: Medical Knowledge Enhanced Language-Image Pre-Training

“…Depending the approaches of using medical knowledge, They can be classified into modelbased and input-based two types. In model-based types, the authors may imitate the radiological practice to design the model [20, 23, 31, 49] or change the model structure based on diagnostic patterns [11, 18, 38]. In input-based types, the knowledge is viewed as an extra input to calculate features [46, 53, 54] or to guide the final loss [9, 24].…”

Pneumonia X‐ray detection with anchor‐free detection framework and data augmentation

“…Huang Xin and others put forward a kind of model, Dual‐Ray Net, of a deep convolutional neural network that can deal with the front and lateral chest radiography simultaneously by referring to the method of using lateral chest radiography to assist in the diagnosis used by radiologists. By comparing different feature fusion methods of addition and concatenation, they found that concatenation's fusion effect is better, with an average AUC reaching 0.778 19 . Currently, COVID‐19 is a contagious infection that has severe effects on the global economy and our daily life 20 .…”