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.
Errataetall: | |
---|---|
Media Type: |
Electronic Article |
Year of Publication: |
2020 |
---|
Contained In: |
IEEE transactions on medical imaging - Vol. 39, No. 8 (2020), p. 2664-2675 |
---|
Language: |
English |
---|
Contributors: |
Xie, Weiyi |
---|
Links: |
---|
Notes: |
Date Completed 14.08.2020 Date Revised 18.12.2020 published: Print UpdateOf: ArXiv. 2020 Apr 16;:. - PMID 32550251 Citation Status MEDLINE Copyright: From MEDLINE®/PubMed®, a database of the U.S. National Library of Medicine |
---|
Physical Description: |
Online-Ressource |
---|
doi: |
10.1109/TMI.2020.2995108 |
---|---|
PMID: |
32730216 |
PPN (Catalogue-ID): |
NLM313983852 |
---|
LEADER | 03611nma a2200565 c 4500 | ||
---|---|---|---|
001 | NLM313983852 | ||
003 | DE-601 | ||
005 | 20201218220513.0 | ||
007 | cr uuu---uuuuu | ||
008 | 200731s2020 000 0 eng d | ||
024 | 7 | |a 10.1109/TMI.2020.2995108 |2 doi | |
028 | 5 | 2 | |a pubmed21n1071.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 Completed 14.08.2020 | ||
500 | |a Date Revised 18.12.2020 | ||
500 | |a published: Print | ||
500 | |a UpdateOf: ArXiv. 2020 Apr 16;:. - PMID 32550251 | ||
500 | |a Citation Status MEDLINE | ||
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 |
653 | 2 | |a Algorithms |6 D000465 | |
653 | 2 | |a Betacoronavirus |6 D000073640 | |
653 | 2 | |a COVID-19 |6 D000086382 | |
653 | 2 | |a Coronavirus Infections |6 D018352 |a *diagnostic imaging |6 Q000000981 | |
653 | 2 | |a Humans |6 D006801 | |
653 | 2 | |a Lung |6 D008168 |a *diagnostic imaging |6 Q000000981 | |
653 | 2 | |a *Neural Networks, Computer |6 D016571 | |
653 | 2 | |a Pandemics |6 D058873 | |
653 | 2 | |a Pneumonia, Viral |6 D011024 |a *diagnostic imaging |6 Q000000981 | |
653 | 2 | |a Pulmonary Disease, Chronic Obstructive |6 D029424 |a diagnostic imaging |6 Q000000981 | |
653 | 2 | |a SARS-CoV-2 |6 D000086402 | |
653 | 2 | |a Tomography, X-Ray Computed |6 D014057 |a *methods |6 Q000379 | |
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 30 |c 08 |h 2664-2675 |