Acute ischemic stroke lesion segmentation in non-contrast CT images using 3D convolutional neural networks | Sciact - Система учета научной деятельности ИМ СО РАН

Acute ischemic stroke lesion segmentation in non-contrast CT images using 3D convolutional neural networks Научная публикация

Журнал

Компьютерная оптика (Computer Optics)
ISSN: 0134-2452

Вых. Данные

Год: 2023, Том: 47, Номер: 5, Страницы: 770-777 Страниц : 8 DOI: 10.18287/2412-6179-CO-1233

Ключевые слова

ischemic stroke, brain, non-contrast CT, segmentation, CNN, 3D U-Net

Авторы

Dobshik A.V. ¹ , Verbitskiy S.K. ¹ , Pestunov I.A. ^1,2 , Sherman K.M. ³ , Sinyavskiy Yu.N. ² , Tulupov A.A. ³ , Berikov V.B. ^1,4

Организации

1	Novosibirsk State University
2	Federal Research Center for Information and Computational Technologies
3	International Tomography Center SB RAS
4	Sobolev Institute of Mathematics SB RAS

Информация о финансировании (2)

1	Институт математики им. С.Л. Соболева СО РАН	FWNF-2022-0015
2	Российский фонд фундаментальных исследований	19-29-01175

In this paper, an automatic algorithm aimed at volumetric segmentation of acute ischemic stroke lesion in non-contrast computed tomography brain 3D images is proposed. Our deeplearning approach is based on the popular 3D U-Net convolutional neural network architecture, which was modified by adding the squeeze-and-excitation blocks and residual connections. Robust pre-processing methods were implemented to improve the segmentation accuracy. Moreover, a special patches sampling strategy was used to address the large size of medical images and class imbalance and to stabilize neural network training. All experiments were performed using fivefold cross-validation on the dataset containing non-contrast computed tomography volumetric brain scans of 81 patients diagnosed with acute ischemic stroke. Two radiology experts manually segmented images independently and then verified the labeling results for inconsistencies. The quantitative results of the proposed algorithm and obtained segmentation were measured by the Dice similarity coefficient, sensitivity, specificity and precision metrics. The suggested pipeline provides a Dice improvement of 12.0 %, sensitivity of 10.2 % and precision 10.0 % over the baseline and achieves an average Dice of 62.8  3.3 %, sensitivity of 69.9  3.9 %, specificity of 99.7  0.2 % and precision of 61.9  3.6 %, showing promising segmentation results.

Dobshik A.V. , Verbitskiy S.K. , Pestunov I.A. , Sherman K.M. , Sinyavskiy Y.N. , Tulupov A.A. , Berikov V.B.
Acute ischemic stroke lesion segmentation in non-contrast CT images using 3D convolutional neural networks
Компьютерная оптика (Computer Optics). 2023. V.47. N5. P.770-777. DOI: 10.18287/2412-6179-CO-1233 WOS Scopus РИНЦ OpenAlex

Поступила в редакцию:	1 окт. 2022 г.
Принята к публикации:	4 апр. 2023 г.
Опубликована в печати:	16 окт. 2023 г.
Опубликована online:	16 окт. 2023 г.

Web of science:	WOS:001109072500010
Scopus:	2-s2.0-85175071232
РИНЦ:	54493475
OpenAlex:	W4387910697

БД	Цитирований
РИНЦ	6
OpenAlex	7
Scopus	5