Related Experiment Video
Updated: Jun 30, 2026
Lung CT Segmentation to Identify Consolidations and Ground Glass Areas for Quantitative Assesment of SARS-CoV Pneumonia
Published on: December 19, 2020
Enhancing Interstitial Lung Disease Diagnoses Through Multimodal AI Integration of Histopathological and CT Image
Kris Lami1, Mutsumi Ozasa1,2, Xiangqian Che3
1Department of Pathology Informatics, Nagasaki University Graduate School of Biomedical Sciences, Nagasaki, Japan.
A new multimodal artificial intelligence (AI) algorithm integrating computed tomography (CT) and histopathology images significantly improves the diagnosis of usual interstitial pneumonia (UIP) in interstitial lung diseases (ILDs). This AI tool enhances diagnostic accuracy and consistency for pathologists, especially when expert knowledge is limited.
Area of Science:
- Pulmonary Medicine
- Artificial Intelligence in Medicine
- Medical Imaging Analysis
Background:
- Interstitial lung diseases (ILDs) diagnosis requires integrating clinical, radiological, and histopathological data.
- Accurate differentiation of usual interstitial pneumonia (UIP) within ILDs is diagnostically challenging.
- Multidisciplinary approaches are essential for high diagnostic accuracy in ILDs.
Purpose of the Study:
- To develop a multimodal artificial intelligence (AI) algorithm for improved UIP diagnosis.
- To combine computed tomography (CT) and histopathological images for enhanced diagnostic performance.
- To increase the accuracy and consistency of UIP diagnosis using AI.
Main Methods:
- Collected a dataset of CT and pathological images from 324 ILD patients (2009-2021).
- Trained a CT component to identify 28 radiological features; utilized a previously developed pathological counterpart.
- Tested the multimodal AI model on 114 samples, comparing its performance against expert and general pathologists.
Main Results:
- The multimodal AI achieved an AUC of 0.92 for distinguishing UIP from non-UIP.
- General pathologists showed improved diagnostic agreement (κ=0.737 post-model vs. 0.273 pre-model).
- Diagnostic consensus with expert pathologists increased (κ scores improved from 0.278-0.53 to 0.474-0.602).
Conclusions:
- The multimodal AI algorithm significantly enhances diagnostic accuracy and consistency for UIP identification.
- The AI tool boosts diagnostic confidence among general pathologists.
- This AI approach supports accurate UIP diagnosis, even in settings with limited specialized expertise.
