My-CBAM-ResNet-Enhanced-COVID-19-Chest-X-ray-Classifier

KO | EN

gitlite — search

#javascript #react #ai #llm #go #android #machine-learning #deep-learning #linux #nodejs #java #vue #pytorch #open-source

My-CBAM-ResNet-Enhanced-COVID-19-Chest-X-ray-Classifier

(★ 8)

An enhanced deep learning model built upon CBAM and residual connections to outperform ResNet-18 in classifying COVID-19 from chest X-ray images.

My-CBAM-ResNet-Enhanced-COVID-19-Chest-X-ray-Classifier 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# An enhanced deep learning model built upon CBAM and residual connections to outperform ResNet-18 in classifying COVID-19 from chest X-ray images.

## ABSTRACT
This project focuses on building a deep learning model based on the Residual Network (ResNet) architecture combined with the Convolutional Block Attention Module (CBAM) to detect COVID-19 from chest X-ray images. By utilizing this architecture, the model is expected to improve detection performance by focusing on more relevant features in the images. The goal of this project is to clearly demonstrate the superiority of this model over the ResNet18 model in classifying chest X-ray images.

## Overview
The COVID-19 pandemic, caused by the SARS-CoV-2 virus, has had a profound and long-lasting impact on global healthcare systems and economies. According to a report by the World Health Organization (WHO) on August 9, 2023, more than 760 million confirmed cases and 6.9 million deaths have been recorded worldwide, although the actual figures are likely higher. This pandemic has created an urgent need for rapid, accurate, and scalable diagnostic tools to effectively control transmission and support timely treatment.

While PCR testing is an effective diagnostic method, it requires advanced infrastructure and highly trained personnel, which limits its accessibility in resource-constrained areas. On the other hand, diagnosing COVID-19 through chest X-ray imaging heavily relies on the expertise of radiologists, which can pose challenges in timely and accurate detection. Therefore, the development of deep learning models that can automatically detect COVID-19 from chest X-ray images has become essential to support healthcare professionals, reduce diagnostic time, and improve accuracy.
 
 
## Project goals
This project focuses on developing a deep learning model based on the Residual Network architecture enhanced with the Convolutional Block Attention Module (CBAM) to improve the performance of COVID-19 diagnosis using raw chest X-ray images, without relying on lung segmentation masks or infection region masks. CBAM allows the model to focus on the most relevant features in the image, thereby enhancing classification accuracy. The proposed model’s effectiveness is evaluated through comparison with the standard ResNet-18 architecture, highlighting the advantages of integrating attention mechanisms in deep learning for medical image analysis.

By improving diagnostic precision and reliability, this model can assist healthcare professionals in quickly and effectively identifying COVID-19 cases and contribute to better preparedness for future public health crises.
 
 
## Result
The deep learning model integrated with the Convolutional Block Attention Module (CBAM) and ResNet-18 have been tested for classifying chest X-ray images. The results demonstrate that the CBAM model outperforms ResNet-18 in detecting COVID-19.

In addition to training the CBAM model, I also trained and tested ResNet-18 simultaneously. In the code I uploaded to GitHub, there is no code for ResNet-18, but you just need to change the selected model for training and testing, as well as the number of color channels in the dataset file, test, and train that I uploaded, to be compatible with ResNet-18 if you also want to compare performance.

Validation loss curve: the blue line corresponds to the custom-designed model, while the orange line represents the baseline ResNet-18 model.

| **Loss visualization**  |
|-------------------------|
| ![img_4.png](https://raw.githubusercontent.com/truonghuyhoang608/My-CBAM-ResNet-Enhanced-COVID-19-Chest-X-ray-Classifier/master/img_4.png) |

Validation accuracy curve: the blue line corresponds to the custom-designed model, while the orange line represents the baseline ResNet-18 model.

| **Accuracy visualization** |
|----------------------------|
| ![img_5.png](https://raw.githubusercontent.com/truonghuyhoang608/My-CBAM-ResNet-Enhanced-COVID-19-Chest-X-ray-Classifier/master/img_5.png)    |

Confusion Matrix:

| ** Custom-designed model** | **ResNet-18** |
|---------------------------------------------|--|
| ![img.png](https://raw.githubusercontent.com/truonghuyhoang608/My-CBAM-ResNet-Enhanced-COVID-19-Chest-X-ray-Classifier/master/img.png)                         | ![img_1.png](https://raw.githubusercontent.com/truonghuyhoang608/My-CBAM-ResNet-Enhanced-COVID-19-Chest-X-ray-Classifier/master/img_1.png) |

Classification Report:

| **Custom-designed model** | **ResNet-18** |
|--------------------------------------------|--|
| ![img_2.png](https://raw.githubusercontent.com/truonghuyhoang608/My-CBAM-ResNet-Enhanced-COVID-19-Chest-X-ray-Classifier/master/img_2.png) | ![img_3.png](https://raw.githubusercontent.com/truonghuyhoang608/My-CBAM-ResNet-Enhanced-COVID-19-Chest-X-ray-Classifier/master/img_3.png) |

## Data Source
This study uses the COVID-QU-Ex dataset developed by Qatar University, which contains 33,920 chest X-ray (CXR) images divided into three categories: COVID-19 (11,956 images), non-COVID pneumonia (11,263 images), and normal (10,701 images). All images are provided with lung segmentation masks, and a subset also includes infection masks, supporting the analysis of COVID-19 severity. 
 
Dataset: 
<a href="https://www.kaggle.com/datasets/anasmohammedtahir/covidqu/data">kaggle Dataset</a>

## Conclusion
This study demonstrates that integrating the Convolutional Block Attention Module (CBAM) into a residual deep learning architecture significantly enhances the performance of chest X-ray (CXR) image classification, particularly for COVID-19 diagnosis. The custom-designed model outperformed the baseline ResNet-18 in terms of accuracy and robustness across all evaluation metrics. By enabling the model to focus more effectively on critical regions of the lungs, CBAM contributes to better feature representation and more precise predictions. These results highlight the potential of attention mechanisms in improving deep learning models for medical image analysis. The proposed architecture can serve as a solid foundation for future work in automated diagnostic systems using radiographic imaging.

My-CBAM-ResNet-Enhanced-COVID-19-Chest-X-ray-Classifier

My-CBAM-ResNet-Enhanced-COVID-19-Chest-X-ray-Classifier 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

My-CBAM-ResNet-Enhanced-COVID-19-Chest-X-ray-Classifier

My-CBAM-ResNet-Enhanced-COVID-19-Chest-X-ray-Classifier 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README