Deep learning enables automated MRI-based estimation of uterine volume also in patients with uterine fibroids undergoing high-intensity focused ultrasound therapy.
Deep learning
Leiomyoma
Magnetic resonance imaging
Uterus
Journal
Insights into imaging
ISSN: 1869-4101
Titre abrégé: Insights Imaging
Pays: Germany
ID NLM: 101532453
Informations de publication
Date de publication:
05 Jan 2023
05 Jan 2023
Historique:
received:
02
09
2022
accepted:
02
12
2022
entrez:
4
1
2023
pubmed:
5
1
2023
medline:
5
1
2023
Statut:
epublish
Résumé
High-intensity focused ultrasound (HIFU) is used for the treatment of symptomatic leiomyomas. We aim to automate uterine volumetry for tracking changes after therapy with a 3D deep learning approach. A 3D nnU-Net model in the default setting and in a modified version including convolutional block attention modules (CBAMs) was developed on 3D T2-weighted MRI scans. Uterine segmentation was performed in 44 patients with routine pelvic MRI (standard group) and 56 patients with uterine fibroids undergoing ultrasound-guided HIFU therapy (HIFU group). Here, preHIFU scans (n = 56), postHIFU imaging maximum one day after HIFU (n = 54), and the last available follow-up examination (n = 53, days after HIFU: 420 ± 377) were included. The training was performed on 80% of the data with fivefold cross-validation. The remaining data were used as a hold-out test set. Ground truth was generated by a board-certified radiologist and a radiology resident. For the assessment of inter-reader agreement, all preHIFU examinations were segmented independently by both. High segmentation performance was already observed for the default 3D nnU-Net (mean Dice score = 0.95 ± 0.05) on the validation sets. Since the CBAM nnU-Net showed no significant benefit, the less complex default model was applied to the hold-out test set, which resulted in accurate uterus segmentation (Dice scores: standard group 0.92 ± 0.07; HIFU group 0.96 ± 0.02), which was comparable to the agreement between the two readers. This study presents a method for automatic uterus segmentation which allows a fast and consistent assessment of uterine volume. Therefore, this method could be used in the clinical setting for objective assessment of therapeutic response to HIFU therapy.
Sections du résumé
BACKGROUND
BACKGROUND
High-intensity focused ultrasound (HIFU) is used for the treatment of symptomatic leiomyomas. We aim to automate uterine volumetry for tracking changes after therapy with a 3D deep learning approach.
METHODS
METHODS
A 3D nnU-Net model in the default setting and in a modified version including convolutional block attention modules (CBAMs) was developed on 3D T2-weighted MRI scans. Uterine segmentation was performed in 44 patients with routine pelvic MRI (standard group) and 56 patients with uterine fibroids undergoing ultrasound-guided HIFU therapy (HIFU group). Here, preHIFU scans (n = 56), postHIFU imaging maximum one day after HIFU (n = 54), and the last available follow-up examination (n = 53, days after HIFU: 420 ± 377) were included. The training was performed on 80% of the data with fivefold cross-validation. The remaining data were used as a hold-out test set. Ground truth was generated by a board-certified radiologist and a radiology resident. For the assessment of inter-reader agreement, all preHIFU examinations were segmented independently by both.
RESULTS
RESULTS
High segmentation performance was already observed for the default 3D nnU-Net (mean Dice score = 0.95 ± 0.05) on the validation sets. Since the CBAM nnU-Net showed no significant benefit, the less complex default model was applied to the hold-out test set, which resulted in accurate uterus segmentation (Dice scores: standard group 0.92 ± 0.07; HIFU group 0.96 ± 0.02), which was comparable to the agreement between the two readers.
CONCLUSIONS
CONCLUSIONS
This study presents a method for automatic uterus segmentation which allows a fast and consistent assessment of uterine volume. Therefore, this method could be used in the clinical setting for objective assessment of therapeutic response to HIFU therapy.
Identifiants
pubmed: 36600120
doi: 10.1186/s13244-022-01342-0
pii: 10.1186/s13244-022-01342-0
pmc: PMC9813298
doi:
Types de publication
Journal Article
Langues
eng
Pagination
1Informations de copyright
© 2022. The Author(s).
Références
Int J Hyperthermia. 2021 Sep;38(2):30-38
pubmed: 34420447
Semin Reprod Med. 2017 Nov;35(6):473-480
pubmed: 29100234
Magn Reson Imaging. 2012 Nov;30(9):1323-41
pubmed: 22770690
Int J Hyperthermia. 2021 Sep;38(2):89-95
pubmed: 34420439
Sci Rep. 2022 May 18;12(1):8297
pubmed: 35585118
AJR Am J Roentgenol. 2004 Dec;183(6):1713-9
pubmed: 15547216
Dtsch Arztebl Int. 2014 Dec 22;111(51-52):877-83
pubmed: 25597366
Sci Rep. 2021 Nov 23;11(1):22768
pubmed: 34815488
Ultrason Sonochem. 2004 May;11(3-4):149-54
pubmed: 15081972
Nat Biomed Eng. 2018 Oct;2(10):719-731
pubmed: 31015651
Nat Rev Cancer. 2018 Aug;18(8):500-510
pubmed: 29777175
Hum Reprod Update. 2016 Nov;22(6):665-686
pubmed: 27466209
PLoS One. 2016 Feb 26;11(2):e0149631
pubmed: 26919185
Rofo. 2019 Mar;191(3):216-227
pubmed: 30703824
Med Biol Eng Comput. 2016 Jul;54(7):1071-84
pubmed: 26530047
BJOG. 2017 Sep;124(10):1501-1512
pubmed: 28296146
Int J Hyperthermia. 2021 Sep;38(2):65-74
pubmed: 34420445
Int J Womens Health. 2017 Sep 05;9:607-617
pubmed: 28919823
Nat Methods. 2021 Feb;18(2):203-211
pubmed: 33288961
Int J Hyperthermia. 2018;35(1):463-470
pubmed: 30204024
Ultraschall Med. 2019 Oct;40(5):625-637
pubmed: 29665583
Eur Radiol. 2022 May;32(5):3142-3151
pubmed: 34595539
Eur Radiol. 2021 Nov;31(11):8807-8815
pubmed: 33974149
IEEE Trans Med Imaging. 2020 Nov;39(11):3309-3320
pubmed: 32356741
Radiology. 2021 Jan;298(1):155-163
pubmed: 33141003
Acad Radiol. 2011 Aug;18(8):970-6
pubmed: 21718955
Annu Int Conf IEEE Eng Med Biol Soc. 2021 Nov;2021:4015-4018
pubmed: 34892111
J Med Imaging (Bellingham). 2021 Sep;8(5):054001
pubmed: 34589556
Front Psychol. 2021 Sep 28;12:710982
pubmed: 34650476