© FUJIFILM Corporation

FUJIFILM

Deep-learning-based 3D content-based image retrieval system on chest HRCT: Performance assessment for interstitial lung diseases and usual interstitial pneumonia

2025.09.11Publications

Diffuse parenchymal lung diseases (DPLDs), such as interstitial lung diseases (ILDs), are common in general hospitals but are often difficult to diagnose due to the wide variety of disease types and CT imaging findings. Even among experienced radiologists, high interobserver variability in identifying CT findings has been reported. To address this issue, in this study [1], we developed a prototype 3D-CBIR system with fully automated database registration and retrieval (Figure 1), utilizing artificial intelligence-based quantitative CT image analysis software (AIQCT). The prototype system was applied to the case database, and its search performance and clinical usefulness for differential diagnosis were evaluated.

Figure 1. Flowchart of image processing in our prototype system.[1]
To evaluate search performance, we used a database of 2058 cases and assessed image similarity between query and retrieved cases using a 5-point visual score (5: Similar, 4: Somewhat similar, 3: Neither, 2: Somewhat dissimilar, 1: Dissimilar). For search performance, the mean score of visual similarity between 70 queries and their top 5 retrieved cases was 4.37 ± 0.83.  To assess clinical usefulness, we evaluated the concordance of labels (ILD/non-ILD, with/without UIP) between query and retrieved cases, using a database of 301 cases across 57 diseases. For clinical usefulness, label concordance between 25 queries and their top 5 retrieved cases was assessed across 4 labels. For ILD, the mean concordance of labels was 0.94 ± 0.15, while for non-ILD, it was 0.64 ± 0.31. For cases with UIP, the mean concordance of labels was 0.86 ± 0.17, while for cases without UIP, it was 0.83 ± 0.24.  The CBIR system showed high accuracy for identifying cases with/without UIP, suggesting its potential to support UIP differentiation in clinical practice. 

[1] Oosawa A, Kurosaki A, Miyamoto A, Hanada S, Nei Y, Nakahama H, et al. (2025) Deep-learning-based 3D content-based image retrieval system on chest HRCT: Performance assessment for interstitial lung diseases and usual interstitial pneumonia. European Journal of Radiology Open 15: 100670

DOI: https://doi.org/10.1016/j.ejro.2025.100670

CAUTION：This is Fujifilm Global Website. Fujifilm makes no representation that products on this website are commercially available in all countries. Approved uses of products vary by country and region. Specifications and appearance of products are subject to change without notice.

Index