Attention-Based Transfer Learning for Efficient Pneumonia Detection in Chest X-ray Images

So-Mi Cha,Bonggyun Ko,Seung-Seok Lee

doi:10.3390/app11031242

So-Mi Cha, Bonggyun Ko + Show 1 more

Open Access

https://doi.org/10.3390/app11031242

Copy DOI

Journal: Applied Sciences	Publication Date: Jan 29, 2021
Citations: 18	License type: CC BY 4.0

Affiliation: Chonnam National University

Abstract

Pneumonia is a form of acute respiratory infection commonly caused by germs, viruses, and fungi, and can prove fatal at any age. Chest X-rays is the most common technique for diagnosing pneumonia. There have been several attempts to apply transfer learning based on a Convolutional Neural Network to build a stable model in computer-aided diagnosis. Recently, with the appearance of an attention mechanism that automatically focuses on the critical part of the image that is crucial for the diagnosis of disease, it is possible to increase the performance of previous models. The goal of this study is to improve the accuracy of a computer-aided diagnostic approach that medical professionals can easily use as an auxiliary tool. In this paper, we proposed the attention-based transfer learning framework for efficient pneumonia detection in chest X-ray images. We collected features from three-types of pre-trained models, ResNet152, DenseNet121, ResNet18 as a role of feature extractor. We redefined the classifier for a new task and applied the attention mechanism as a feature selector. As a result, the proposed approach achieved accuracy, F-score, Area Under the Curve(AUC), precision and recall of 96.63%, 0.973, 96.03%, 96.23% and 98.46%, respectively.

Highlights

Pneumonia is a disease that causes inflammatory reactions and hardening of the lung tissue in areas below the respiratory tract along the breathing path as a result of disease-causing bacteria [1]
The primary goal of our approach is to improve the accuracy of pneumonia diagnosis using chest X-ray images
Feature vectors were extracted using the backbone structures of pre-trained ResNet152, DenseNet121, ResNet18 to confirm the performance of the classification of pneumonia and normal

Summary

Introduction

Pneumonia is a disease that causes inflammatory reactions and hardening of the lung tissue in areas below the respiratory tract along the breathing path as a result of disease-causing bacteria [1]. It is common to start treatment with a diagnosis of pneumonia from chest photos [3], usually accompanied by coughing, sputum, and fever. Medical experts perform additional computed tomography (CT) when a possibility of pneumonia exists based on the chest X-ray results. In chest X-ray images, blurry parts of the bronchial tubes can be expressed depending on the value of each pixel. At this time, blurry parts indicate suspected pneumonia due to inflammation or the discharge of inflammation. Values corresponding to each pixel are classified as normal or pneumonia through a layer of weights inside the model. It is not easy to diagnose pneumonia through X-ray images without the help of experts.

Objectives

Methods

Results

Discussion

Conclusion