[Paper Review] Classification of COVID-19 on chest X-Ray images using Deep Learning model with Histogram Equalization and Lungs Segmentation

Aman Swaraj, Karan Verma|arXiv (Cornell University)|Dec 5, 2021

COVID-19 diagnosis using AI4 citations

TL;DR

This study proposes a deep learning framework for classifying COVID-19 from chest X-ray images using histogram equalization and U-Net-based lung segmentation to enhance image quality and focus on relevant regions. The method employs VGG-16 for feature extraction, SMOTE for dataset balancing, and SVM for classification, achieving 98% accuracy on a 2,470-image dataset with three classes: healthy, pneumonia, and COVID-19.

ABSTRACT

Background and Objective: Artificial intelligence (AI) methods coupled with biomedical analysis has a critical role during pandemics as it helps to release the overwhelming pressure from healthcare systems and physicians. As the ongoing COVID-19 crisis worsens in countries having dense populations and inadequate testing kits like Brazil and India, radiological imaging can act as an important diagnostic tool to accurately classify covid-19 patients and prescribe the necessary treatment in due time. With this motivation, we present our study based on deep learning architecture for detecting covid-19 infected lungs using chest X-rays. Dataset: We collected a total of 2470 images for three different class labels, namely, healthy lungs, ordinary pneumonia, and covid-19 infected pneumonia, out of which 470 X-ray images belong to the covid-19 category. Methods: We first pre-process all the images using histogram equalization techniques and segment them using U-net architecture. VGG-16 network is then used for feature extraction from the pre-processed images which is further sampled by SMOTE oversampling technique to achieve a balanced dataset. Finally, the class-balanced features are classified using a support vector machine (SVM) classifier with 10-fold cross-validation and the accuracy is evaluated. Result and Conclusion: Our novel approach combining well-known pre-processing techniques, feature extraction methods, and dataset balancing method, lead us to an outstanding rate of recognition of 98% for COVID-19 images over a dataset of 2470 X-ray images. Our model is therefore fit to be utilized in healthcare facilities for screening purposes.

Motivation & Objective

• To address the diagnostic burden in regions with limited testing capacity during the COVID-19 pandemic by developing an AI-based screening tool.
• To improve the performance of COVID-19 detection in chest X-rays through image preprocessing and region-of-interest focusing.
• To overcome class imbalance in medical imaging datasets using SMOTE oversampling for more robust model training.
• To evaluate the effectiveness of combining histogram equalization, U-Net segmentation, and transfer learning with VGG-16 for classification.
• To provide a reliable, high-accuracy model suitable for deployment in healthcare facilities for early detection of COVID-19.

Proposed method

• Preprocessing chest X-ray images using histogram equalization to enhance contrast and visibility of lung structures.
• Applying U-Net architecture to segment lung regions from the preprocessed X-ray images, isolating the area of interest.
• Using pre-trained VGG-16 for feature extraction from the segmented lung regions to capture hierarchical patterns.
• Implementing SMOTE oversampling to balance the dataset across three classes: healthy, pneumonia, and COVID-19.
• Training a support vector machine (SVM) classifier on the balanced, high-dimensional features with 10-fold cross-validation.
• Evaluating model performance using 10-fold cross-validation to ensure robustness and generalization.

Experimental results

Research questions

• RQ1 Can histogram equalization and lung segmentation improve the accuracy of COVID-19 detection in chest X-ray images?
• RQ2 How effective is the combination of VGG-16 feature extraction and SMOTE-based data balancing for classifying low-resource medical images?
• RQ3 To what extent does the proposed pipeline outperform baseline methods in distinguishing between healthy, pneumonia, and COVID-19 cases?
• RQ4 Does the use of U-Net segmentation reduce noise and irrelevant structures, thereby improving classifier performance?
• RQ5 Can the proposed model achieve high accuracy on a small, imbalanced dataset typical of real-world medical imaging scenarios?

Key findings

• The proposed method achieved a classification accuracy of 98% for detecting COVID-19 in chest X-ray images.
• The integration of histogram equalization and U-Net segmentation significantly improved image quality and region-of-interest focus.
• SMOTE oversampling effectively mitigated class imbalance, particularly for the minority COVID-19 class with 470 images.
• The use of VGG-16 for feature extraction enabled robust learning of discriminative patterns from segmented lung regions.
• The 10-fold cross-validation confirmed consistent performance across all folds, indicating model stability and generalization.
• The model demonstrates strong potential for deployment in clinical settings to support early and accurate detection of COVID-19.