Enhancing Surgical Guidance: Deep Learning-Based Liver Vessel Segmentation in Real-Time Ultrasound Video Frames

Muhammad Awais; Mais Al Taie; Caleb S. O’Connor; Austin H. Castelo; Belkacem Acidi; Hop S. Tran Cao; Kristy K. Brock

doi:10.3390/cancers16213674

Enhancing Surgical Guidance: Deep Learning-Based Liver Vessel Segmentation in Real-Time Ultrasound Video Frames

Muhammad Awais, Mais Al Taie, Caleb S. O’Connor, Austin H. Castelo, Belkacem Acidi, Hop S. Tran Cao, Kristy K. Brock

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

1 Scopus citations

Abstract

In liver surgery, the complex and individualized nature of liver vascular anatomy makes planning and execution challenging. Traditional 2D intraoperative ultrasonography (IOUS) often suffers from interpretability issues due to noise and artifacts. This paper introduces an AI-based model, the “2D-weighted U-Net model,” designed to enhance real-time IOUS navigation by accurately segmenting key blood vessels, including the inferior vena cava, hepatic veins, portal vein, and its major branches. Our deep learning model demonstrated high performance, with Dice scores ranging from 0.84 to 0.96 across different vessels. This advancement promises improved precision in liver resection procedures and sets the stage for future development of real-time multi-label segmentation for broader liver vasculature.

Original language	English (US)
Article number	3674
Journal	Cancers
Volume	16
Issue number	21
DOIs	https://doi.org/10.3390/cancers16213674
State	Published - Nov 2024

Keywords

2D-weighted U-Net model
deep learning
intraoperative ultrasound (IOUS) video frames
liver vessel segmentation

ASJC Scopus subject areas

Oncology
Cancer Research

Access to Document

10.3390/cancers16213674

Cite this

@article{106b02b5e1004b489a90d15e602da0c8,

title = "Enhancing Surgical Guidance: Deep Learning-Based Liver Vessel Segmentation in Real-Time Ultrasound Video Frames",

abstract = "In liver surgery, the complex and individualized nature of liver vascular anatomy makes planning and execution challenging. Traditional 2D intraoperative ultrasonography (IOUS) often suffers from interpretability issues due to noise and artifacts. This paper introduces an AI-based model, the “2D-weighted U-Net model,” designed to enhance real-time IOUS navigation by accurately segmenting key blood vessels, including the inferior vena cava, hepatic veins, portal vein, and its major branches. Our deep learning model demonstrated high performance, with Dice scores ranging from 0.84 to 0.96 across different vessels. This advancement promises improved precision in liver resection procedures and sets the stage for future development of real-time multi-label segmentation for broader liver vasculature.",

keywords = "2D-weighted U-Net model, deep learning, intraoperative ultrasound (IOUS) video frames, liver vessel segmentation",

author = "Muhammad Awais and {Al Taie}, Mais and O{\textquoteright}Connor, {Caleb S.} and Castelo, {Austin H.} and Belkacem Acidi and {Tran Cao}, {Hop S.} and Brock, {Kristy K.}",

note = "Publisher Copyright: {\textcopyright} 2024 by the authors.",

year = "2024",

month = nov,

doi = "10.3390/cancers16213674",

language = "English (US)",

volume = "16",

journal = "Cancers",

issn = "2072-6694",

publisher = "Multidisciplinary Digital Publishing Institute (MDPI)",

number = "21",

}

TY - JOUR

T1 - Enhancing Surgical Guidance

T2 - Deep Learning-Based Liver Vessel Segmentation in Real-Time Ultrasound Video Frames

AU - Awais, Muhammad

AU - Al Taie, Mais

AU - O’Connor, Caleb S.

AU - Castelo, Austin H.

AU - Acidi, Belkacem

AU - Tran Cao, Hop S.

AU - Brock, Kristy K.

PY - 2024/11

Y1 - 2024/11

N2 - In liver surgery, the complex and individualized nature of liver vascular anatomy makes planning and execution challenging. Traditional 2D intraoperative ultrasonography (IOUS) often suffers from interpretability issues due to noise and artifacts. This paper introduces an AI-based model, the “2D-weighted U-Net model,” designed to enhance real-time IOUS navigation by accurately segmenting key blood vessels, including the inferior vena cava, hepatic veins, portal vein, and its major branches. Our deep learning model demonstrated high performance, with Dice scores ranging from 0.84 to 0.96 across different vessels. This advancement promises improved precision in liver resection procedures and sets the stage for future development of real-time multi-label segmentation for broader liver vasculature.

AB - In liver surgery, the complex and individualized nature of liver vascular anatomy makes planning and execution challenging. Traditional 2D intraoperative ultrasonography (IOUS) often suffers from interpretability issues due to noise and artifacts. This paper introduces an AI-based model, the “2D-weighted U-Net model,” designed to enhance real-time IOUS navigation by accurately segmenting key blood vessels, including the inferior vena cava, hepatic veins, portal vein, and its major branches. Our deep learning model demonstrated high performance, with Dice scores ranging from 0.84 to 0.96 across different vessels. This advancement promises improved precision in liver resection procedures and sets the stage for future development of real-time multi-label segmentation for broader liver vasculature.

KW - 2D-weighted U-Net model

KW - deep learning

KW - intraoperative ultrasound (IOUS) video frames

KW - liver vessel segmentation

UR - http://www.scopus.com/inward/record.url?scp=85208406162&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85208406162&partnerID=8YFLogxK

U2 - 10.3390/cancers16213674

DO - 10.3390/cancers16213674

M3 - Article

C2 - 39518111

AN - SCOPUS:85208406162

SN - 2072-6694

VL - 16

JO - Cancers

JF - Cancers

IS - 21

M1 - 3674

ER -

Enhancing Surgical Guidance: Deep Learning-Based Liver Vessel Segmentation in Real-Time Ultrasound Video Frames

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this