[Paper Review] COVIDLite: A depth-wise separable deep neural network with white balance and CLAHE for detection of COVID-19

Manu Siddhartha, Avik Santra|arXiv (Cornell University)|Jun 18, 2020

COVID-19 diagnosis using AIMedicine36 citations

TL;DR

COVIDLite combines white balance and CLAHE preprocessing with a depth-wise separable CNN to detect COVID-19 from chest X-rays, achieving high accuracy and lightweight deployment suitability.

ABSTRACT

Background and Objective:Currently, the whole world is facing a pandemic disease, novel Coronavirus also known as COVID-19, which spread in more than 200 countries with around 3.3 million active cases and 4.4 lakh deaths approximately. Due to rapid increase in number of cases and limited supply of testing kits, availability of alternative diagnostic method is necessary for containing the spread of COVID-19 cases at an early stage and reducing the death count. For making available an alternative diagnostic method, we proposed a deep neural network based diagnostic method which can be easily integrated with mobile devices for detection of COVID-19 and viral pneumonia using Chest X-rays (CXR) images. Methods:In this study, we have proposed a method named COVIDLite, which is a combination of white balance followed by Contrast Limited Adaptive Histogram Equalization (CLAHE) and depth-wise separable convolutional neural network (DSCNN). In this method, white balance followed by CLAHE is used as an image preprocessing step for enhancing the visibility of CXR images and DSCNN trained using sparse cross entropy is used for image classification with lesser parameters and significantly lighter in size, i.e., 8.4 MB without quantization. Results:The proposed COVIDLite method resulted in improved performance in comparison to vanilla DSCNN with no pre-processing. The proposed method achieved higher accuracy of 99.58% for binary classification, whereas 96.43% for multiclass classification and out-performed various state-of-the-art methods. Conclusion:Our proposed method, COVIDLite achieved exceptional results on various performance metrics. With detailed model interpretations, COVIDLite can assist radiologists in detecting COVID-19 patients from CXR images and can reduce the diagnosis time significantly.

Motivation & Objective

• Motivate rapid and accessible COVID-19 screening via chest X-rays when RT-PCR resources are limited.
• Develop a lightweight CNN architecture suitable for mobile/web deployment.
• Enhance CXR image quality through targeted preprocessing to improve CNN feature extraction.
• Provide model interpretability to assist radiologists via Grad-CAM, LIME, and saliency maps.

Proposed method

• Apply white balance followed by CLAHE as a preprocessing pipeline to enhance CXR visibility.
• Employ a depth-wise separable CNN (DSCNN) to reduce parameters and compute compared with standard CNNs.
• Train the proposed COVIDLite network with 18 layers (2 standard conv + 12 DSC layers + dense blocks) using 3x3 kernels.
• Normalize and resize images to 224x224x3 before feeding into the network.
• Use sparse categorical cross-entropy loss with softmax activation for multi-class classification.
• Validate with 5-fold cross-validation and report standard metrics (accuracy, precision, recall, F1, AUC, Cohen's kappa).
• Assess model interpretability with Grad-CAM, LIME, and saliency maps to highlight pneumonia-affected regions.

Experimental results

Research questions

• RQ1 How does white balance + CLAHE preprocessing affect COVID-19 detection performance from CXR images?
• RQ2 Can a lightweight depth-wise separable CNN achieve competitive accuracy for COVID-19 classification (2-class and 3-class) on CXR data?
• RQ3 What are the performance trade-offs (accuracy, precision, recall, AUC, kappa) between preprocessing-enabled and preprocessing-disabled pipelines?
• RQ4 To what extent can Grad-CAM, LIME, and saliency maps provide clinically meaningful explanations for the model’s decisions?

Key findings

• Binary classification accuracy with preprocessing: 98.75% (avg across 5 folds).
• Three-class classification accuracy with preprocessing: 99.58% (avg across 5 folds).
• Class-wise performance (with preprocessing) yields Precision 99.00% (Normal) / 100.00% (COVID-19) / 96.00% (Viral Pneumonia); Sensitivity 99.00% / 99.58% / 96.00%; Specificity 98.25% / 99.34% / 97.79%; F1-Score 99.00% / 99.79% / 95.00%; AUC 1.0 / 1.0 / 0.99.
• AUC for binary: average AUC = 1.0 for COVID-19 and Normal classes; for multi-class, class-wise AUCs are 1.0 (COVID-19) and 0.99–1.0 for others.
• The model size is lightweight (8.4 MB without quantization) and suitable for mobile/web deployment; preprocessing improves performance and narrows 95% CIs.