An enhanced machine learning-based prognostic prediction model for patients with AECOPD on invasive mechanical...

Area of Science:

• Pulmonary Medicine
• Artificial Intelligence
• Biomedical Informatics

Background:

• Chronic obstructive pulmonary disease (COPD) is a leading cause of global morbidity and mortality.
• Acute exacerbations of COPD (AECOPDs) significantly worsen patient prognosis and frequently necessitate mechanical ventilation.
• Identifying prognostic factors for AECOPD patients on mechanical ventilation is critical for improving clinical outcomes.

Purpose of the Study:

• To enhance the RIME algorithm for improved feature selection and solution space exploration in AECOPD prognosis prediction.
• To develop and validate a novel algorithm, DDRIME, for predicting outcomes in AECOPD patients requiring mechanical ventilation.
• To identify key clinical indicators associated with prognosis in mechanically ventilated AECOPD patients.

Main Methods:

• The study introduces DDRIME, an enhanced version of the RIME algorithm incorporating a dispersed foraging mechanism and a differential crossover operator.
• DDRIME's feature selection capabilities were evaluated against the original RIME algorithm using 83 functions and 12 public datasets.
• Patient data from AECOPD cases requiring invasive mechanical ventilation was analyzed to identify prognostic factors.

Main Results:

• DDRIME demonstrated superior performance in feature selection compared to most other algorithms tested.
• The bDDRIME_KNN variant achieved high accuracy in predicting AECOPD patient outcomes.
• Chronic heart failure (CHF), D-dimer (D-D) levels, fungal infection (FI), and pectoral muscle area (PMA) were identified as key indicators with >0.98 accuracy in predicting prognosis.

Conclusions:

• DDRIME represents a significant advancement over the RIME algorithm for feature selection and optimization.
• The bDDRIME_KNN model shows strong potential for accurately predicting outcomes in AECOPD patients undergoing mechanical ventilation.
• The identified key indicators (CHF, D-D, FI, PMA) provide valuable insights for prognostic assessment and clinical management of these high-risk patients.