Machine learning-based automatic detection of novel coronavirus (COVID-19) disease.

Area of Science:

• Medical Imaging Analysis
• Machine Learning in Healthcare
• Computational Virology

Background:

• The COVID-19 pandemic necessitates rapid and accurate diagnostic tools.
• Current RT-PCR testing faces limitations including high turnaround times and false negatives.
• Medical imaging like Chest Radiography (CXR) and Computed Tomography (CT) offer potential for early detection.

Purpose of the Study:

• To develop and validate an automated machine learning-based algorithm for COVID-19 detection and grading.
• To leverage image processing techniques for enhanced accuracy in identifying COVID-19 from medical images.
• To provide a faster, more reliable alternative to existing COVID-19 diagnostic methods.

Main Methods:

• Image preprocessing including normalization and noise reduction.
• Image segmentation using fuzzy c-means clustering.
• Feature extraction (statistical, textural, HOG, DWT) and selection via Principal Component Analysis.
• Classification using k-NN, SRC, ANN, and SVM algorithms.
• Validation using k-fold cross-validation.

Main Results:

• The Support Vector Machine (SVM) classifier achieved the highest accuracy at 99.14% for COVID-19 detection.
• The proposed algorithm demonstrated improved recognition rates compared to existing literature.
• Feature combination and selection enhanced performance, with classification completed in 14.34 seconds.

Conclusions:

• The developed machine learning and image processing approach shows significant potential for rapid and accurate COVID-19 diagnosis.
• SVM classifier proved most effective, offering promising results comparable to current literature.
• The algorithm can assist radiologists in early diagnosis and differentiation of COVID-19 from other respiratory conditions.