COVID-19 anomaly detection and classification method based on supervised machine learning of chest X-ray images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer-Aided Diagnosis

Background:

• Coronavirus Disease 2019 (COVID-19) is a global pandemic requiring rapid detection.
• Early diagnosis of COVID-19 is crucial for patient recovery and limiting disease spread.
• X-ray imaging presents a viable modality for COVID-19 screening.

Purpose of the Study:

• To propose and evaluate an automated image processing method for COVID-19 classification and early detection using X-ray images.
• To compare the performance of different feature extraction and classification models for COVID-19 diagnosis.

Main Methods:

• Image preprocessing techniques including noise removal, thresholding, and morphological operations.
• Region of Interest (ROI) detection and segmentation.
• Feature extraction using Local Binary Pattern (LBP), Histogram of Gradient (HOG), and Haralick texture features.
• Classification using K-Nearest Neighbor (KNN) and Support Vector Machine (SVM) algorithms, forming six distinct models.
• Model evaluation using 5-fold cross-validation on 5,000 X-ray images.

Main Results:

• Six models (LBP-KNN, HOG-KNN, Haralick-KNN, LBP-SVM, HOG-SVM, Haralick-SVM) were evaluated.
• Diagnosis accuracy ranged from 89.2% to 98.66%.
• The LBP-KNN model demonstrated superior performance with 98.66% average accuracy, 97.76% sensitivity, 100% specificity, and 100% precision.

Conclusions:

• The proposed automated image processing method provides an effective, end-to-end solution for COVID-19 detection from X-ray images.
• The LBP-KNN model shows significant potential for accurate and reliable COVID-19 diagnosis.
• The method eliminates the need for manual feature extraction and selection, streamlining the diagnostic process.