Developing Deep LSTMs With Later Temporal Attention for Predicting COVID-19 Severity, Clinical Outcome, and Antibody...

Area of Science:

• Medical diagnostics
• Artificial intelligence in healthcare
• Immunology

Background:

• COVID-19 presents highly variable clinical courses and immunological responses.
• Understanding disease progression and patient status requires identifying key associated factors.
• Dynamic changes in Spike protein antibodies are vital for assessing immune response.

Purpose of the Study:

• To explore a temporal attention (TA) deep learning mechanism for predicting COVID-19 severity, clinical outcomes, and Spike antibody levels.
• To utilize time-series serological indicators for enhanced disease monitoring.
• To develop a computer-aided diagnostic system for COVID-19.

Main Methods:

• Feature selection techniques were used to identify relevant serological indicators.
• Deep Long Short-Term Memory (LSTM) models were employed to capture dynamic changes in disease markers.
• A Temporal Attention Long Short-Term Memory (TA-LSTM) model was proposed, emphasizing later blood test results.

Main Results:

• Key risk factors associated with COVID-19 were identified.
• LSTM models achieved high accuracy in predicting disease severity.
• TA-LSTM demonstrated superior performance in predicting clinical outcomes.
• LSTM models showed the best performance for Spike antibody level prediction.

Conclusions:

• The proposed deep learning models effectively predict COVID-19 disease severity, clinical outcomes, and antibody levels.
• These models offer a valuable tool for computer-aided medical diagnostics using time-series serological data.