Related Experiment Video
Updated: Aug 13, 2026
Detection of Invasive Pulmonary Aspergillosis in Haematological Malignancy Patients by using Lateral-flow Technology
Published on: March 22, 2012
Development and Validation of a Machine Learning-Based Bedside Score (IPA-2T2C2DW) for Invasive Pulmonary
Linna Huang1,2, Ruyi Rong1,2, Xiaoyi Zhou1,3
1National Center for Respiratory Medicine; State Key Laboratory of Respiratory Health and Multimorbidity; National Clinical Research Center for Respiratory Diseases; Institute of Respiratory Medicine, Chinese Academy of Medical Sciences & Peking Union Medical College; Department of Pulmonary and Critical Care Medicine, Center of Respiratory Medicine, China-Japan Friendship Hospital, Beijing, P.R. China.
A new machine learning model and bedside score (IPA-2T2C2DW) can rapidly identify intensive care unit (ICU) patients at high risk for invasive pulmonary aspergillosis (IPA). This tool aids in guiding antifungal therapy and monitoring intensity for better patient outcomes.
Area of Science:
- Medical Informatics
- Infectious Diseases
- Pulmonology
Background:
- Invasive pulmonary aspergillosis (IPA) poses a significant threat to immunocompromised patients with pneumonia and respiratory failure in the ICU.
- Early and accurate diagnosis of IPA is crucial for timely initiation of antifungal therapy and improved patient survival.
- Current diagnostic methods may be limited by time constraints and data availability upon ICU admission.
Purpose of the Study:
- To develop and validate machine learning (ML)-based diagnostic models for IPA using data available within 24 hours of ICU admission.
- To create three model versions suitable for different healthcare settings.
- To derive a bedside score for rapid IPA risk stratification.
Main Methods:
- A prospective, multi-center cohort study enrolled immunocompromised patients with pneumonia and respiratory failure admitted to the ICU.
- Data from a derivation cohort (China-Japan Friendship Hospital) and an external validation cohort (15 additional centers) were used.
- Six ML algorithms were tested, and nine independent predictors were identified, including CD4+ T-cell count, CT findings, temperature, WBC, and corticosteroid dose.
Main Results:
- Logistic regression demonstrated strong performance in internal validation (AUC 0.915).
- In external validation, CatBoost (AUC 0.844) and logistic regression (AUC 0.863 for simplified model) showed robust discrimination.
- The derived IPA-2T2C2DW score effectively stratified patients into low (8.2% IPA), intermediate (28.6% IPA), and high-risk (62.4% IPA) groups.
Conclusions:
- The IPA-2T2C2DW score provides a rapid, bedside tool for IPA risk stratification in diverse ICU settings.
- This score can guide the intensity of monitoring and empirical antifungal treatment decisions.
- Further prospective validation in unselected cohorts is recommended prior to widespread clinical implementation.
