Related Experiment Video
Updated: Jul 5, 2025
Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
MI-DenseCFNet: deep learning-based multimodal diagnosis models for Aureus and Aspergillus pneumonia
Tong Liu1, Zheng-Hua Zhang2, Qi-Hao Zhou3
1The Second Department of Respiratory and Critical Care Medicine, The First Affiliated Hospital of Kunming Medical University, No. 295, Xichang Road, Wuhua District, Kunming, Yunnan, 650032, People's Republic of China.
A new diagnostic model, MI-DenseCFNet, effectively distinguishes Staphylococcus aureus pneumonia (SAP) from Aspergillus pneumonia (ASP) with high accuracy. This AI tool aids clinicians by identifying key features, improving diagnosis in primary hospitals.
Area of Science:
- Medical imaging analysis
- Artificial intelligence in diagnostics
- Pneumonia classification
Background:
- Distinguishing between Staphylococcus aureus pneumonia (SAP) and Aspergillus pneumonia (ASP) is clinically challenging.
- Accurate and efficient diagnosis is crucial for appropriate treatment and patient outcomes.
Purpose of the Study:
- To develop and validate a multimodal diagnostic model (MI-DenseCFNet) integrating deep learning and clinical features for SAP and ASP differentiation.
- To identify significant clinical features contributing to the diagnosis of SAP and ASP using a random forest model.
Main Methods:
- A deep learning model (DenseNet) was fused with clinical data to create the MI-DenseCFNet.
- Thoracic CT images and clinical data from 60 patients with confirmed SAP or ASP were analyzed.
- A random forest model was employed to screen for significant clinical features.
Main Results:
- The MI-DenseCFNet achieved an AUC of 0.92 on the internal validation set and 0.83 on the external validation set.
- The model demonstrated diagnostic accuracies superior to junior and mid-ranking radiologists (78% vs. 75% vs. 60%).
- Eleven significant clinical features were identified by the random forest model.
Conclusions:
- The MI-DenseCFNet model offers effective and efficient diagnosis of SAP and ASP, outperforming junior radiologists.
- The identified clinical features provide valuable insights for clinicians in diagnosing these pneumonia types.
- This AI-driven approach can support diagnostic capabilities in primary healthcare settings, potentially reducing antibiotic misuse.
