Blog
Feedback
schliessen

Filtern

 

Bibliotheken

Logo der Bibliothek

Logo FID Pharmazie PubPharm Discovery System Universitätsbibliothek Braunschweig Institut für Informationssysteme

Relational Modeling for Robust and Efficient Pulmonary Lobe Segmentation in CT Scans

Pulmonary lobe segmentation in computed tomography scans is essential for regional assessment of pulmonary diseases. Recent works based on convolution neural networks have achieved good performance for this task. However, they are still limited in capturing structured relationships due to the nature of convolution. The shape of the pulmonary lobes affect each other and their borders relate to the appearance of other structures, such as vessels, airways, and the pleural wall. We argue that such structural relationships play a critical role in the accurate delineation of pulmonary lobes when the lungs are affected by diseases such as COVID-19 or COPD. In this paper, we propose a relational approach (RTSU-Net) that leverages structured relationships by introducing a novel non-local neural network module. The proposed module learns both visual and geometric relationships among all convolution features to produce self-attention weights. With a limited amount of training data available from COVID-19 subjects, we initially train and validate RTSU-Net on a cohort of 5000 subjects from the COPDGene study (4000 for training and 1000 for evaluation). Using models pre-trained on COPDGene, we apply transfer learning to retrain and evaluate RTSU-Net on 470 COVID-19 suspects (370 for retraining and 100 for evaluation). Experimental results show that RTSU-Net outperforms three baselines and performs robustly on cases with severe lung infection due to COVID-19

Year of Publication: 2020
Contained in: IEEE transactions on medical imaging Vol. 39, No. 8 (2020), p. 2664-2675
All journal articles: Search for all articles in this journal
Language: English
Contributors: Xie, Weiyi | Author
Jacobs, Colin
Charbonnier, Jean-Paul
van Ginneken, Bram
Full text access:
Electronic availability is being checked...
Links: Full Text (dx.doi.org)
Keywords: Journal Article
Research Support, N.I.H., Extramural
ISSN: 1558-254X
Note: Copyright: From MEDLINE®/PubMed®, a database of the U.S. National Library of Medicine
Notes: Date Revised 03.08.2020
published: Print
Citation Status In-Process
Copyright: From MEDLINE®/PubMed®, a database of the U.S. National Library of Medicine
PMID:
    32730216
Physical Description: Online-Ressource
ID (e.g. DOI, URN): 10.1109/TMI.2020.2995108
more publication details ...

Associated Publications

  • Associated records are being queried...
more (+)
Internes Format
LEADER 02847nma a2200397 c 4500
001 NLM313983852
003 DE-601
005 20200803214350.0
007 cr uuu---uuuuu
008 200731s2020 000 0 eng d
024 7 |a 10.1109/TMI.2020.2995108  |2 doi 
028 5 2 |a pubmed20n1289.xml 
035 |a (DE-599)NLM32730216 
040 |b ger  |c GBVCP 
041 0 |a eng 
100 1 |a Xie, Weiyi 
245 1 0 |a Relational Modeling for Robust and Efficient Pulmonary Lobe Segmentation in CT Scans  |h Elektronische Ressource 
300 |a Online-Ressource 
500 |a Date Revised 03.08.2020 
500 |a published: Print 
500 |a Citation Status In-Process 
500 |a Copyright: From MEDLINE®/PubMed®, a database of the U.S. National Library of Medicine 
520 |a Pulmonary lobe segmentation in computed tomography scans is essential for regional assessment of pulmonary diseases. Recent works based on convolution neural networks have achieved good performance for this task. However, they are still limited in capturing structured relationships due to the nature of convolution. The shape of the pulmonary lobes affect each other and their borders relate to the appearance of other structures, such as vessels, airways, and the pleural wall. We argue that such structural relationships play a critical role in the accurate delineation of pulmonary lobes when the lungs are affected by diseases such as COVID-19 or COPD. In this paper, we propose a relational approach (RTSU-Net) that leverages structured relationships by introducing a novel non-local neural network module. The proposed module learns both visual and geometric relationships among all convolution features to produce self-attention weights. With a limited amount of training data available from COVID-19 subjects, we initially train and validate RTSU-Net on a cohort of 5000 subjects from the COPDGene study (4000 for training and 1000 for evaluation). Using models pre-trained on COPDGene, we apply transfer learning to retrain and evaluate RTSU-Net on 470 COVID-19 suspects (370 for retraining and 100 for evaluation). Experimental results show that RTSU-Net outperforms three baselines and performs robustly on cases with severe lung infection due to COVID-19 
611 2 7 |a Journal Article  |2 gnd 
611 2 7 |a Research Support, N.I.H., Extramural  |2 gnd 
689 0 0 |A f  |a Journal Article 
689 0 1 |A f  |a Research Support, N.I.H., Extramural 
689 0 |5 DE-601 
700 1 |a Jacobs, Colin 
700 1 |a Charbonnier, Jean-Paul 
700 1 |a van Ginneken, Bram 
773 0 8 |i in  |t IEEE transactions on medical imaging  |g Vol. 39, No. 8 (2020), p. 2664-2675  |q 39:8<2664-2675  |w (DE-601)NLM082857202  |x 1558-254X 
856 4 1 |u http://dx.doi.org/10.1109/TMI.2020.2995108  |3 Volltext 
912 |a GBV_NLM 
951 |a AR 
952 |d 39  |j 2020  |e 8  |b 26  |c 08  |h 2664-2675