Dataset Information

Content-Based Image Retrieval of Chest CT with Convolutional Neural Network for Diffuse Interstitial Lung Disease: Performance Assessment in Three Major Idiopathic Interstitial Pneumonias.

ABSTRACT:

Objective

To assess the performance of content-based image retrieval (CBIR) of chest CT for diffuse interstitial lung disease (DILD).

Materials and methods

The database was comprised by 246 pairs of chest CTs (initial and follow-up CTs within two years) from 246 patients with usual interstitial pneumonia (UIP, n = 100), nonspecific interstitial pneumonia (NSIP, n = 101), and cryptogenic organic pneumonia (COP, n = 45). Sixty cases (30-UIP, 20-NSIP, and 10-COP) were selected as the queries. The CBIR retrieved five similar CTs as a query from the database by comparing six image patterns (honeycombing, reticular opacity, emphysema, ground-glass opacity, consolidation and normal lung) of DILD, which were automatically quantified and classified by a convolutional neural network. We assessed the rates of retrieving the same pairs of query CTs, and the number of CTs with the same disease class as query CTs in top 1-5 retrievals. Chest radiologists evaluated the similarity between retrieved CTs and queries using a 5-scale grading system (5-almost identical; 4-same disease; 3-likelihood of same disease is half; 2-likely different; and 1-different disease).

Results

The rate of retrieving the same pairs of query CTs in top 1 retrieval was 61.7% (37/60) and in top 1-5 retrievals was 81.7% (49/60). The CBIR retrieved the same pairs of query CTs more in UIP compared to NSIP and COP (p = 0.008 and 0.002). On average, it retrieved 4.17 of five similar CTs from the same disease class. Radiologists rated 71.3% to 73.0% of the retrieved CTs with a similarity score of 4 or 5.

Conclusion

The proposed CBIR system showed good performance for retrieving chest CTs showing similar patterns for DILD.

SUBMITTER: Hwang HJ

PROVIDER: S-EPMC7817627 | biostudies-literature | 2021 Feb

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Content-Based Image Retrieval of Chest CT with Convolutional Neural Network for Diffuse Interstitial Lung Disease: Performance Assessment in Three Major Idiopathic Interstitial Pneumonias.

Hwang Hye Jeon HJ Seo Joon Beom JB Lee Sang Min SM Kim Eun Young EY Park Beomhee B Bae Hyun Jin HJ Kim Namkug N

Korean journal of radiology 20201021 2

<h4>Objective</h4>To assess the performance of content-based image retrieval (CBIR) of chest CT for diffuse interstitial lung disease (DILD).<h4>Materials and methods</h4>The database was comprised by 246 pairs of chest CTs (initial and follow-up CTs within two years) from 246 patients with usual interstitial pneumonia (UIP, n = 100), nonspecific interstitial pneumonia (NSIP, n = 101), and cryptogenic organic pneumonia (COP, n = 45). Sixty cases (30-UIP, 20-NSIP, and 10-COP) were selected as the ...[more]

PMID: 33169547

Similar Datasets

Project description:BackgroundPleuroparenchymal fibroelastosis (PPFE) is characterised by predominant upper lobe pleural and subpleural lung parenchymal fibrosis. Radiological PPFE-like lesion has been associated with various types of interstitial lung diseases. However, the prevalence and clinical significance of radiological PPFE-like lesion in patients with idiopathic interstitial pneumonias (IIPs) are not fully understood. We aimed to determine the prevalence and clinical impact on survival of radiological PPFE-like lesion in patients with IIPs.MethodsA post-hoc analysis was conducted using data from the Japanese nationwide cloud-based database of patients with IIPs. All the patients in the database were diagnosed as having IIPs by multidisciplinary discussion. Patients diagnosed with idiopathic PPFE were excluded. Clinical data and chest computed tomography (CT) image of 419 patients with IIPs were analysed. The presence of radiological PPFE-like lesion was independently evaluated by two chest radiologists blind to the clinical data.ResultsOf the 419 patients with IIPs, radiological PPFE-like lesions were detected in 101 (24.1%) patients, mainly in idiopathic pulmonary fibrosis (IPF) and unclassifiable IIPs, but less in idiopathic nonspecific interstitial pneumonia. Prognostic analyses revealed that radiological PPFE-like lesion was significantly associated with poor outcome in patients with IIPs, which was independent of age, IPF diagnosis and %FVC. In survival analyses, the patients with radiological PPFE-like lesions had poor survival compared with those without (log-rank, p < 0.0001). Subgroup analyses demonstrated that radiological PPFE-like lesion was significantly associated with poor survival both in patients with IPF and those with unclassifiable IIPs.ConclusionRadiological PPFE-like lesion is a condition that could exist in IIPs, mainly in IPF and unclassifiable IIPs. Importantly, the radiological PPFE-like lesion is a non-invasive marker to predict poor outcome in patients with IIPs, which should be carefully considered in clinical practice.

Project description:BACKGROUND:Idiopathic interstitial pneumonias (IIPs) are a group of heterogeneous, somewhat unpredictable diseases characterized by progressive scarring of the interstitium. Since lung function is a key determinant of survival, we reasoned that the transcriptional profile in IIP lung tissue would be associated with measures of lung function, and could enhance prognostic approaches to IIPs. RESULTS:Using gene expression profiling of 167 lung tissue specimens with IIP diagnosis and 50 control lungs, we identified genes whose expression is associated with changes in lung function (% predicted FVC and % predicted DLCO) modeled as categorical (severe vs mild disease) or continuous variables while adjusting for smoking status and IIP subtype; false discovery rate (FDR) approach was used to correct for multiple comparisons. This analysis identified 58 transcripts that are associated with mild vs severe disease (categorical analysis), including those with established role in fibrosis (ADAMTS4, ADAMTS9, AGER, HIF-1?, SERPINA3, SERPINE2, and SELE) as well as novel IIP candidate genes such as rhotekin 2 (RTKN2) and peptidase inhibitor 15 (PI15). Protein-protein interactome analysis of 553 genes whose expression is significantly associated with lung function when modeled as continuous variables demonstrates that more severe presentation of IIPs is characterized by an increase in cell cycle progression and apoptosis, increased hypoxia, and dampened innate immune response. Our findings were validated in an independent cohort of 131 IIPs and 40 controls at the mRNA level and for one gene (RTKN2) at the protein level by immunohistochemistry in a subset of samples. CONCLUSIONS:We identified commonalities and differences in gene expression among different subtypes of IIPs. Disease progression, as characterized by lower measures of FVC and DLCO, results in marked changes in expression of novel and established genes and pathways involved in IIPs. These genes and pathways represent strong candidates for biomarker studies and potential therapeutic targets for IIP severity.

Dataset Information

Content-Based Image Retrieval of Chest CT with Convolutional Neural Network for Diffuse Interstitial Lung Disease: Performance Assessment in Three Major Idiopathic Interstitial Pneumonias.

Objective

Materials and methods

Results

Conclusion

Publications

Content-Based Image Retrieval of Chest CT with Convolutional Neural Network for Diffuse Interstitial Lung Disease: Performance Assessment in Three Major Idiopathic Interstitial Pneumonias.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets