LHSPred

LHSPred predicts lung health severity and pneumonia risk from chest computed tomography (CT) scans and clinical features to support assessment and monitoring of pulmonary diseases including COVID-19.

Key Features:

• Automated CT Scan Evaluation: Computes a CT severity score by evaluating lung lobe involvement due to lesions on chest CT scans.
• Pneumonia Risk Prediction from Blood Tests and Age: Predicts pneumonia risk using features derived from blood examinations and patient age.
• Post-COVID Lung Health Monitoring: Provides predictive assessments for monitoring lung health in recovered COVID-19 patients.

Scientific Applications:

• CT severity quantification: Quantifies lung involvement for pulmonary disease assessment such as pneumonia and COVID-19.
• Risk stratification without CT: Estimates pneumonia risk using blood-derived features and age when CT imaging is unavailable.
• Longitudinal monitoring: Tracks changes in predicted lung health severity over time for post-infection follow-up.
• Model development and validation: Supports research on machine learning models for pulmonary severity prediction using reported COVID-19 case data.

Methodology:

Employs machine learning techniques—Support Vector Regression (SVR) and Multi-Layer Perceptron Regression (MLPR)—trained on reported COVID-19 patient cases from the literature; SVR computes CT severity scores from lung lobe lesion involvement (PCC=0.77, MAE=2.239), and MLPR predicts pneumonia risk from blood-test features and age (PCC=0.77, MAE=2.309).