Dense regression activation maps for lesion segmentation in CT scans of COVID-19 patients
Copyright © 2023 The Author(s). Published by Elsevier B.V. All rights reserved..
Automatic lesion segmentation on thoracic CT enables rapid quantitative analysis of lung involvement in COVID-19 infections. However, obtaining a large amount of voxel-level annotations for training segmentation networks is prohibitively expensive. Therefore, we propose a weakly-supervised segmentation method based on dense regression activation maps (dRAMs). Most weakly-supervised segmentation approaches exploit class activation maps (CAMs) to localize objects. However, because CAMs were trained for classification, they do not align precisely with the object segmentations. Instead, we produce high-resolution activation maps using dense features from a segmentation network that was trained to estimate a per-lobe lesion percentage. In this way, the network can exploit knowledge regarding the required lesion volume. In addition, we propose an attention neural network module to refine dRAMs, optimized together with the main regression task. We evaluated our algorithm on 90 subjects. Results show our method achieved 70.2% Dice coefficient, substantially outperforming the CAM-based baseline at 48.6%. We published our source code at https://github.com/DIAGNijmegen/bodyct-dram.
Medienart: |
E-Artikel |
---|
Erscheinungsjahr: |
2023 |
---|---|
Erschienen: |
2023 |
Enthalten in: |
Zur Gesamtaufnahme - volume:86 |
---|---|
Enthalten in: |
Medical image analysis - 86(2023) vom: 01. Mai, Seite 102771 |
Sprache: |
Englisch |
---|
Beteiligte Personen: |
Xie, Weiyi [VerfasserIn] |
---|
Links: |
---|
Themen: |
COVID-19 |
---|
Anmerkungen: |
Date Completed 21.04.2023 Date Revised 27.04.2023 published: Print-Electronic Citation Status MEDLINE |
---|
doi: |
10.1016/j.media.2023.102771 |
---|
funding: |
|
---|---|
Förderinstitution / Projekttitel: |
|
PPN (Katalog-ID): |
NLM353548200 |
---|
LEADER | 01000naa a22002652 4500 | ||
---|---|---|---|
001 | NLM353548200 | ||
003 | DE-627 | ||
005 | 20231226060245.0 | ||
007 | cr uuu---uuuuu | ||
008 | 231226s2023 xx |||||o 00| ||eng c | ||
024 | 7 | |a 10.1016/j.media.2023.102771 |2 doi | |
028 | 5 | 2 | |a pubmed24n1178.xml |
035 | |a (DE-627)NLM353548200 | ||
035 | |a (NLM)36848720 | ||
035 | |a (PII)S1361-8415(23)00032-4 | ||
040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
041 | |a eng | ||
100 | 1 | |a Xie, Weiyi |e verfasserin |4 aut | |
245 | 1 | 0 | |a Dense regression activation maps for lesion segmentation in CT scans of COVID-19 patients |
264 | 1 | |c 2023 | |
336 | |a Text |b txt |2 rdacontent | ||
337 | |a ƒaComputermedien |b c |2 rdamedia | ||
338 | |a ƒa Online-Ressource |b cr |2 rdacarrier | ||
500 | |a Date Completed 21.04.2023 | ||
500 | |a Date Revised 27.04.2023 | ||
500 | |a published: Print-Electronic | ||
500 | |a Citation Status MEDLINE | ||
520 | |a Copyright © 2023 The Author(s). Published by Elsevier B.V. All rights reserved. | ||
520 | |a Automatic lesion segmentation on thoracic CT enables rapid quantitative analysis of lung involvement in COVID-19 infections. However, obtaining a large amount of voxel-level annotations for training segmentation networks is prohibitively expensive. Therefore, we propose a weakly-supervised segmentation method based on dense regression activation maps (dRAMs). Most weakly-supervised segmentation approaches exploit class activation maps (CAMs) to localize objects. However, because CAMs were trained for classification, they do not align precisely with the object segmentations. Instead, we produce high-resolution activation maps using dense features from a segmentation network that was trained to estimate a per-lobe lesion percentage. In this way, the network can exploit knowledge regarding the required lesion volume. In addition, we propose an attention neural network module to refine dRAMs, optimized together with the main regression task. We evaluated our algorithm on 90 subjects. Results show our method achieved 70.2% Dice coefficient, substantially outperforming the CAM-based baseline at 48.6%. We published our source code at https://github.com/DIAGNijmegen/bodyct-dram | ||
650 | 4 | |a Journal Article | |
650 | 4 | |a Research Support, Non-U.S. Gov't | |
650 | 4 | |a COVID-19 | |
650 | 4 | |a Class activation map | |
650 | 4 | |a Computed tomography | |
650 | 4 | |a Dense regression activation map | |
650 | 4 | |a Medical imaging | |
650 | 4 | |a Weakly-supervised semantic segmentation | |
700 | 1 | |a Jacobs, Colin |e verfasserin |4 aut | |
700 | 1 | |a Charbonnier, Jean-Paul |e verfasserin |4 aut | |
700 | 1 | |a van Ginneken, Bram |e verfasserin |4 aut | |
773 | 0 | 8 | |i Enthalten in |t Medical image analysis |d 1996 |g 86(2023) vom: 01. Mai, Seite 102771 |w (DE-627)NLM096527188 |x 1361-8423 |7 nnns |
773 | 1 | 8 | |g volume:86 |g year:2023 |g day:01 |g month:05 |g pages:102771 |
856 | 4 | 0 | |u http://dx.doi.org/10.1016/j.media.2023.102771 |3 Volltext |
912 | |a GBV_USEFLAG_A | ||
912 | |a GBV_NLM | ||
951 | |a AR | ||
952 | |d 86 |j 2023 |b 01 |c 05 |h 102771 |