Exploring the Effect of Image Enhancement Techniques on COVID-19 Detection using Chest X-rays Images (2012.02238v1)

Abstract: The use of computer-aided diagnosis in the reliable and fast detection of coronavirus disease (COVID-19) has become a necessity to prevent the spread of the virus during the pandemic to ease the burden on the medical infrastructure. Chest X-ray (CXR) imaging has several advantages over other imaging techniques as it is cheap, easily accessible, fast and portable. This paper explores the effect of various popular image enhancement techniques and states the effect of each of them on the detection performance. We have compiled the largest X-ray dataset called COVQU-20, consisting of 18,479 normal, non-COVID lung opacity and COVID-19 CXR images. To the best of our knowledge, this is the largest public COVID positive database. Ground glass opacity is the common symptom reported in COVID-19 pneumonia patients and so a mixture of 3616 COVID-19, 6012 non-COVID lung opacity, and 8851 normal chest X-ray images were used to create this dataset. Five different image enhancement techniques: histogram equalization, contrast limited adaptive histogram equalization, image complement, gamma correction, and Balance Contrast Enhancement Technique were used to improve COVID-19 detection accuracy. Six different Convolutional Neural Networks (CNNs) were investigated in this study. Gamma correction technique outperforms other enhancement techniques in detecting COVID-19 from standard and segmented lung CXR images. The accuracy, precision, sensitivity, f1-score, and specificity in the detection of COVID-19 with gamma correction on CXR images were 96.29%, 96.28%, 96.29%, 96.28% and 96.27% respectively. The accuracy, precision, sensitivity, F1-score, and specificity were 95.11 %, 94.55 %, 94.56 %, 94.53 % and 95.59 % respectively for segmented lung images. The proposed approach with very high and comparable performance will boost the fast and robust COVID-19 detection using chest X-ray images.

• The paper demonstrates that gamma correction significantly outperforms other techniques in enhancing CNN-based COVID-19 detection, achieving over 96% accuracy on non-segmented images.
• The methodology combines lung segmentation via a modified U-Net with six CNN architectures, using a substantial dataset of 18,479 CXR images for robust evaluation.
• The findings imply that improving image quality with enhancement techniques can accelerate clinical diagnosis and relieve pressure on healthcare systems.

Effect of Image Enhancement Techniques on COVID-19 Detection using Chest X-rays

The paper "Exploring the Effect of Image Enhancement Techniques on COVID-19 Detection using Chest X-rays" investigates the impact of various image enhancement methods on the automatic detection of COVID-19 utilizing chest X-rays (CXRs). The paper is essential for improving diagnostic accuracy and speed, thereby supporting the medical community in managing COVID-19 cases effectively.

Overview

The research employs a substantial dataset, termed COVQU-20, comprising 18,479 CXR images categorized into normal, non-COVID lung opacity, and COVID-19 positive cases. This paper distinguishes itself by being the first to thoroughly analyze the effect of image enhancement on COVID-19 detection, leveraging such a large dataset.

Image Enhancement Techniques and Methodology

The investigation encompassed five image enhancement techniques:

1. Histogram Equalization (HE)
2. Contrast Limited Adaptive Histogram Equalization (CLAHE)
3. Image Inversion/Complement
4. Gamma Correction
5. Balance Contrast Enhancement Technique (BCET)

Six different Convolutional Neural Networks (CNNs)—ResNet18, ResNet50, ResNet101, InceptionV3, DenseNet201, and CheXNet—were evaluated to detect COVID-19 from non-segmented and segmented CXR images. Segmentation was carried out using a modified U-Net architecture.

The paper was structured in three primary phases:

1. Training and testing the U-Net model for lung segmentation.
2. Training and evaluating the six CNNs on plain CXR images.
3. Training and assessing the CNNs on segmented lung images.

Numerical Results

The results reveal that the gamma correction technique consistently outperformed other enhancement methods across different CNN models. Specifically, CheXNet combined with gamma correction achieved the highest performance when utilized on non-segmented images, boasting:

• Accuracy: 96.29%
• Precision: 96.28%
• Sensitivity: 96.29%
• F1-Score: 96.28%
• Specificity: 97.27%

For segmented lung images, DenseNet201 with gamma correction provided superior performance:

• Accuracy: 95.11%
• Precision: 94.55%
• Sensitivity: 94.56%
• F1-Score: 94.53%
• Specificity: 95.59%

Implications

The findings emphasize the importance of image enhancement in improving the diagnostic accuracy of COVID-19 from CXRs. Specifically, gamma correction significantly enhances the discriminative power of CNN models. The comparison between non-segmented and segmented lung images demonstrates that while segmentation marginally reduces performance metrics, it notably increases the reliability of predictions by ensuring that the model focuses on relevant regions of interest, as verified by Score-CAM visualizations.

Future Work

Future research may pivot towards exploring additional image enhancement techniques and other deep learning architectures. The promising results warrant more in-depth studies on real-world deployment and integration into clinical workflows. Potential advancements could include optimizing image enhancement parameters and incorporating multi-modal data (e.g., combining CXRs with clinical data) to further augment diagnostic accuracy.

In summary, this paper provides a systematic and thorough evaluation of image enhancement techniques in COVID-19 detection, underscoring gamma correction's efficacy. The findings hold substantial implications for enhancing AI-based diagnostic tools, potentially expediting the screening process in clinical settings and alleviating the burden on medical infrastructures.