Semi-Supervised KPCA-Based Monitoring Techniques for Detecting COVID-19 Infection through Blood Tests.

Area of Science:

• Biomedical data analysis
• Machine learning for disease detection
• Infectious disease diagnostics

Background:

• Accurate and early detection of COVID-19 is crucial for public health management.
• Traditional diagnostic methods may have limitations in accessibility or require specific sample types.
• Developing semi-supervised approaches can leverage unlabeled data for improved diagnostic capabilities.

Purpose of the Study:

• To introduce a novel semi-supervised anomaly detection method for identifying COVID-19 infections using blood test data.
• To differentiate between healthy individuals and those infected with COVID-19.
• To evaluate the performance of the proposed method against existing semi-supervised models.

Main Methods:

• Combining kernel principal component analysis (KPCA) for nonlinear pattern identification with one-class support vector machine (OCSVM) for anomaly detection.
• Utilizing a semi-supervised learning approach that trains on data from healthy cases only.
• Validating the method on two independent blood test datasets from Brazil and Italy.

Main Results:

• The proposed KPCA-OCSVM method demonstrated superior discrimination performance compared to other semi-supervised models (iForest, LOF, EE, ICA, PCA-OCSVM).
• Achieved an area under the receiver operating characteristic curve (AUC) of 0.99 on two COVID-19 blood test datasets.
• Indicated high accuracy in distinguishing between positive and negative COVID-19 samples.

Conclusions:

• The KPCA-OCSVM approach is a promising, highly accurate method for detecting COVID-19 infections.
• This semi-supervised technique offers a viable solution for COVID-19 detection without the need for labeled training data.
• Blood test data combined with advanced machine learning offers a powerful tool for infectious disease surveillance.