Thyroid ultrasound diagnosis improvement via multi-view self-supervised learning and two-stage pre-training
Copyright © 2024 Elsevier Ltd. All rights reserved..
Thyroid nodule classification and segmentation in ultrasound images are crucial for computer-aided diagnosis; however, they face limitations owing to insufficient labeled data. In this study, we proposed a multi-view contrastive self-supervised method to improve thyroid nodule classification and segmentation performance with limited manual labels. Our method aligns the transverse and longitudinal views of the same nodule, thereby enabling the model to focus more on the nodule area. We designed an adaptive loss function that eliminates the limitations of the paired data. Additionally, we adopted a two-stage pre-training to exploit the pre-training on ImageNet and thyroid ultrasound images. Extensive experiments were conducted on a large-scale dataset collected from multiple centers. The results showed that the proposed method significantly improves nodule classification and segmentation performance with limited manual labels and outperforms state-of-the-art self-supervised methods. The two-stage pre-training also significantly exceeded ImageNet pre-training.
Medienart: |
E-Artikel |
---|
Erscheinungsjahr: |
2024 |
---|---|
Erschienen: |
2024 |
Enthalten in: |
Zur Gesamtaufnahme - volume:171 |
---|---|
Enthalten in: |
Computers in biology and medicine - 171(2024) vom: 15. März, Seite 108087 |
Sprache: |
Englisch |
---|
Beteiligte Personen: |
Wang, Jian [VerfasserIn] |
---|
Links: |
---|
Themen: |
Journal Article |
---|
Anmerkungen: |
Date Completed 21.03.2024 Date Revised 21.03.2024 published: Print-Electronic Citation Status MEDLINE |
---|
doi: |
10.1016/j.compbiomed.2024.108087 |
---|
funding: |
|
---|---|
Förderinstitution / Projekttitel: |
|
PPN (Katalog-ID): |
NLM368547299 |
---|
LEADER | 01000caa a22002652 4500 | ||
---|---|---|---|
001 | NLM368547299 | ||
003 | DE-627 | ||
005 | 20240322000128.0 | ||
007 | cr uuu---uuuuu | ||
008 | 240217s2024 xx |||||o 00| ||eng c | ||
024 | 7 | |a 10.1016/j.compbiomed.2024.108087 |2 doi | |
028 | 5 | 2 | |a pubmed24n1338.xml |
035 | |a (DE-627)NLM368547299 | ||
035 | |a (NLM)38364658 | ||
035 | |a (PII)S0010-4825(24)00171-9 | ||
040 | |a DE-627 |b ger |c DE-627 |e rakwb | ||
041 | |a eng | ||
100 | 1 | |a Wang, Jian |e verfasserin |4 aut | |
245 | 1 | 0 | |a Thyroid ultrasound diagnosis improvement via multi-view self-supervised learning and two-stage pre-training |
264 | 1 | |c 2024 | |
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 21.03.2024 | ||
500 | |a Date Revised 21.03.2024 | ||
500 | |a published: Print-Electronic | ||
500 | |a Citation Status MEDLINE | ||
520 | |a Copyright © 2024 Elsevier Ltd. All rights reserved. | ||
520 | |a Thyroid nodule classification and segmentation in ultrasound images are crucial for computer-aided diagnosis; however, they face limitations owing to insufficient labeled data. In this study, we proposed a multi-view contrastive self-supervised method to improve thyroid nodule classification and segmentation performance with limited manual labels. Our method aligns the transverse and longitudinal views of the same nodule, thereby enabling the model to focus more on the nodule area. We designed an adaptive loss function that eliminates the limitations of the paired data. Additionally, we adopted a two-stage pre-training to exploit the pre-training on ImageNet and thyroid ultrasound images. Extensive experiments were conducted on a large-scale dataset collected from multiple centers. The results showed that the proposed method significantly improves nodule classification and segmentation performance with limited manual labels and outperforms state-of-the-art self-supervised methods. The two-stage pre-training also significantly exceeded ImageNet pre-training | ||
650 | 4 | |a Journal Article | |
650 | 4 | |a Multi-view learning | |
650 | 4 | |a Nodule classification | |
650 | 4 | |a Nodule segmentation | |
650 | 4 | |a Self-supervised learning | |
650 | 4 | |a Thyroid ultrasound image | |
650 | 4 | |a Two-stage pre-training | |
700 | 1 | |a Yang, Xin |e verfasserin |4 aut | |
700 | 1 | |a Jia, Xiaohong |e verfasserin |4 aut | |
700 | 1 | |a Xue, Wufeng |e verfasserin |4 aut | |
700 | 1 | |a Chen, Rusi |e verfasserin |4 aut | |
700 | 1 | |a Chen, Yanlin |e verfasserin |4 aut | |
700 | 1 | |a Zhu, Xiliang |e verfasserin |4 aut | |
700 | 1 | |a Liu, Lian |e verfasserin |4 aut | |
700 | 1 | |a Cao, Yan |e verfasserin |4 aut | |
700 | 1 | |a Zhou, Jianqiao |e verfasserin |4 aut | |
700 | 1 | |a Ni, Dong |e verfasserin |4 aut | |
700 | 1 | |a Gu, Ning |e verfasserin |4 aut | |
773 | 0 | 8 | |i Enthalten in |t Computers in biology and medicine |d 1970 |g 171(2024) vom: 15. März, Seite 108087 |w (DE-627)NLM000382272 |x 1879-0534 |7 nnns |
773 | 1 | 8 | |g volume:171 |g year:2024 |g day:15 |g month:03 |g pages:108087 |
856 | 4 | 0 | |u http://dx.doi.org/10.1016/j.compbiomed.2024.108087 |3 Volltext |
912 | |a GBV_USEFLAG_A | ||
912 | |a GBV_NLM | ||
951 | |a AR | ||
952 | |d 171 |j 2024 |b 15 |c 03 |h 108087 |