Standard

Deep Learning Methods for Right Ventricle Segmentation in Radionuclide Imaging. / Ларочкин, Петр Викторович; Котина, Елена Дмитриевна; Гирдюк, Дмитрий Викторович; Остроумов, Евгений Николаевич.

In: Cybernetics and Physics, Vol. 14, No. 1, 28.06.2025, p. 62-66.

Research output: Contribution to journal › Article › peer-review

Harvard

Ларочкин, ПВ, Котина, ЕД, Гирдюк, ДВ & Остроумов, ЕН 2025, 'Deep Learning Methods for Right Ventricle Segmentation in Radionuclide Imaging', Cybernetics and Physics, vol. 14, no. 1, pp. 62-66. https://doi.org/10.35470/2226-4116-2024-14-1-62-66

APA

Ларочкин, П. В., Котина, Е. Д., Гирдюк, Д. В., & Остроумов, Е. Н. (2025). Deep Learning Methods for Right Ventricle Segmentation in Radionuclide Imaging. Cybernetics and Physics, 14(1), 62-66. https://doi.org/10.35470/2226-4116-2024-14-1-62-66

Vancouver

Ларочкин ПВ, Котина ЕД, Гирдюк ДВ, Остроумов ЕН. Deep Learning Methods for Right Ventricle Segmentation in Radionuclide Imaging. Cybernetics and Physics. 2025 Jun 28;14(1):62-66. https://doi.org/10.35470/2226-4116-2024-14-1-62-66

Author

Ларочкин, Петр Викторович ; Котина, Елена Дмитриевна ; Гирдюк, Дмитрий Викторович ; Остроумов, Евгений Николаевич. / Deep Learning Methods for Right Ventricle Segmentation in Radionuclide Imaging. In: Cybernetics and Physics. 2025 ; Vol. 14, No. 1. pp. 62-66.

BibTeX

@article{25379d7af99b4ae8a82bb8d43f1d7a20,
title = "Deep Learning Methods for Right Ventricle Segmentation in Radionuclide Imaging",
abstract = "This study presents a deep learning-based approach for the automatic segmentation of the right ventricle (RV) in gated myocardial perfusion SPECT images (gSPECT). Unlike the left ventricle (LV), the RV poses significant segmentation challenges due to its complex anatomy, thinner walls, and lower perfusion. We manually annotated 384 SPECT volumes and propose the use of the ResUNetSE3D neural network, incorporating both anatomical and phase imaging data to enhance segmentation accuracy. The model achieved a Dice coefficient of 0.8272 and a Jaccard index of 0.7086. These results demonstrate the feasibility of fully automated RV segmentation, laying the groundwork for future clinical applications in quantitative cardiac assessment.",
keywords = "deep learning, mathematical modeling, myocardial perfusion imaging, image segmentation, gated single photon emission computed tomography, Gated single photon emission computed tomography (gSPECT), deep learning, image segmentation, mathematical modeling, myocardial perfusion imaging",
author = "Ларочкин, {Петр Викторович} and Котина, {Елена Дмитриевна} and Гирдюк, {Дмитрий Викторович} and Остроумов, {Евгений Николаевич}",
year = "2025",
month = jun,
day = "28",
doi = "10.35470/2226-4116-2024-14-1-62-66",
language = "English",
volume = "14",
pages = "62--66",
journal = "Cybernetics and Physics",
issn = "2223-7038",
publisher = "IPACS",
number = "1",

}

RIS

TY - JOUR

T1 - Deep Learning Methods for Right Ventricle Segmentation in Radionuclide Imaging

AU - Ларочкин, Петр Викторович

AU - Котина, Елена Дмитриевна

AU - Гирдюк, Дмитрий Викторович

AU - Остроумов, Евгений Николаевич

PY - 2025/6/28

Y1 - 2025/6/28

N2 - This study presents a deep learning-based approach for the automatic segmentation of the right ventricle (RV) in gated myocardial perfusion SPECT images (gSPECT). Unlike the left ventricle (LV), the RV poses significant segmentation challenges due to its complex anatomy, thinner walls, and lower perfusion. We manually annotated 384 SPECT volumes and propose the use of the ResUNetSE3D neural network, incorporating both anatomical and phase imaging data to enhance segmentation accuracy. The model achieved a Dice coefficient of 0.8272 and a Jaccard index of 0.7086. These results demonstrate the feasibility of fully automated RV segmentation, laying the groundwork for future clinical applications in quantitative cardiac assessment.

AB - This study presents a deep learning-based approach for the automatic segmentation of the right ventricle (RV) in gated myocardial perfusion SPECT images (gSPECT). Unlike the left ventricle (LV), the RV poses significant segmentation challenges due to its complex anatomy, thinner walls, and lower perfusion. We manually annotated 384 SPECT volumes and propose the use of the ResUNetSE3D neural network, incorporating both anatomical and phase imaging data to enhance segmentation accuracy. The model achieved a Dice coefficient of 0.8272 and a Jaccard index of 0.7086. These results demonstrate the feasibility of fully automated RV segmentation, laying the groundwork for future clinical applications in quantitative cardiac assessment.

KW - deep learning

KW - mathematical modeling

KW - myocardial perfusion imaging

KW - image segmentation

KW - gated single photon emission computed tomography

KW - Gated single photon emission computed tomography (gSPECT)

KW - deep learning

KW - image segmentation

KW - mathematical modeling

KW - myocardial perfusion imaging

UR - https://www.mendeley.com/catalogue/41196851-010c-3eff-9945-d2cd25ce5cf9/

U2 - 10.35470/2226-4116-2024-14-1-62-66

DO - 10.35470/2226-4116-2024-14-1-62-66

M3 - Article

VL - 14

SP - 62

EP - 66

JO - Cybernetics and Physics

JF - Cybernetics and Physics

SN - 2223-7038

IS - 1

ER -

ID: 137816613