Coronavirus disease identification using Multi-subband feature analysis in DWT domain.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Signal Processing

Background:

• Early and accurate diagnosis of Coronavirus disease (COVID-19) is crucial for effective patient management and public health.
• Current diagnostic methods like RT-PCR and Antigen tests can be time-consuming and may yield inaccurate results.
• Radiological scans, particularly Chest X-rays (CXRs), offer valuable insights into lung infection characteristics.

Purpose of the Study:

• To develop an automated machine learning approach for COVID-19 identification using CXR images.
• To evaluate the efficacy of multi-subband feature extraction via 2D Discrete Wavelet Transform (DWT) for COVID-19 detection.
• To compare the performance of different wavelet families in classifying COVID-19 from CXR images.

Main Methods:

• Utilized 2D Discrete Wavelet Transform (DWT) to decompose CXR images into multi-frequency subbands.
• Concatenated low and high-frequency components into a single feature vector for analysis.
• Trained a Support Vector Machine (SVM) classifier using extracted features and experimented with various wavelet families (Haar, Daubechies, Symlets, Biorthogonal).

Main Results:

• The Haar wavelet-based feature extraction demonstrated superior performance compared to other wavelet families.
• The developed machine learning model achieved a classification accuracy of 100% for COVID-19 detection.
• This automated approach shows promise as an alternative to traditional diagnostic methods.

Conclusions:

• Automated analysis of CXR images using DWT and machine learning, specifically the Haar wavelet, provides a highly accurate method for COVID-19 diagnosis.
• This technique can potentially expedite the diagnostic process and improve accuracy, aiding in differentiating COVID-19 from other lung infections.
• The findings suggest a viable alternative or complementary tool for early COVID-19 identification.