Real-Time Liver Lesion Segmentation in Ultrasound Imaging using Deep Learning

Ultrasound (US) imaging is a widely used, non-invasive method for detecting liver tumors and assessing parenchymal changes. However, the inherent variability and noise in US images pose challenges for accurate lesion identification. This study aims to develop and evaluate a deep learning (DL) model...

Full description

Saved in:
Bibliographic Details
Main Authors: Marinela-Cristiana URHUŢ, Larisa Daniela SĂNDULESCU, Costin Teodor STREBA, Mădălin MĂMULEANU, Adriana CIOCÂLTEU, Sergiu Marian CAZACU, Suzana DĂNOIU
Format: Article
Language:English
Published: Iuliu Hatieganu University of Medicine and Pharmacy, Cluj-Napoca 2025-05-01
Series:Applied Medical Informatics
Subjects:
Online Access:https://ami.info.umfcluj.ro/index.php/AMI/article/view/1194
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850102044742909952
author Marinela-Cristiana URHUŢ
Larisa Daniela SĂNDULESCU
Costin Teodor STREBA
Mădălin MĂMULEANU
Adriana CIOCÂLTEU
Sergiu Marian CAZACU
Suzana DĂNOIU
author_facet Marinela-Cristiana URHUŢ
Larisa Daniela SĂNDULESCU
Costin Teodor STREBA
Mădălin MĂMULEANU
Adriana CIOCÂLTEU
Sergiu Marian CAZACU
Suzana DĂNOIU
author_sort Marinela-Cristiana URHUŢ
collection DOAJ
description Ultrasound (US) imaging is a widely used, non-invasive method for detecting liver tumors and assessing parenchymal changes. However, the inherent variability and noise in US images pose challenges for accurate lesion identification. This study aims to develop and evaluate a deep learning (DL) model capable of performing real-time segmentation of liver lesions in US scans. A dataset of 50 video examinations was used, from which frames were extracted and manually annotated by an experienced gastroenterologist. The segmentation process was conducted using a U-Net architecture with focal Tversky loss (FTL) to address class imbalance. Two versions of the model were trained with different FTL parameters: Model 1 (α = β = 0.5, γ = 1) and Model 2 (α = 0.7, β = 0.3, γ = 0.75). The models were assessed based on key performance metrics, including intersection over union (IoU), recall, and precision. Model 1 achieved a higher IoU score (0.84) than Model 2. Both models demonstrated inference times between 30 and 80 milliseconds, confirming their feasibility for real-time US applications. Visual analysis showed that Model 1 produced more precise and contiguous lesion segmentation, whereas Model 2 tended to separate lesions that were close together. These findings suggest that the proposed DL models are effective in real-time liver lesion segmentation in US imaging. Model 1, which utilized balanced FTL parameters, demonstrated superior segmentation accuracy.
format Article
id doaj-art-5c41e7f90cec4fcfb98e61ba5ba76532
institution DOAJ
issn 2067-7855
language English
publishDate 2025-05-01
publisher Iuliu Hatieganu University of Medicine and Pharmacy, Cluj-Napoca
record_format Article
series Applied Medical Informatics
spelling doaj-art-5c41e7f90cec4fcfb98e61ba5ba765322025-08-20T02:39:51ZengIuliu Hatieganu University of Medicine and Pharmacy, Cluj-NapocaApplied Medical Informatics2067-78552025-05-0147Suppl. 1Real-Time Liver Lesion Segmentation in Ultrasound Imaging using Deep LearningMarinela-Cristiana URHUŢ0Larisa Daniela SĂNDULESCU1Costin Teodor STREBA2Mădălin MĂMULEANU3Adriana CIOCÂLTEU4Sergiu Marian CAZACU5Suzana DĂNOIU6Doctoral School, University of Medicine and Pharmacy of Craiova, Petru Rareş, no. 2, 200349 Craiova, RomaniaDepartment of Gastroenterology, Research Center of Gastroenterology and Hepatology, University of Medicine and Pharmacy of Craiova, Petru Rareş, no. 2, 200349 Craiova, RomaniaUniversity of Medicine and Pharmacy of CraiovaDepartment of Automatic Control and Electronics, University of Craiova, A. I. Cuza Str., no. 13, 200585 Craiova, Romania.Department of Gastroenterology, Research Center of Gastroenterology and Hepatology, University of Medicine and Pharmacy of Craiova, Petru Rareş, no. 2, 200349 Craiova, RomaniaDepartment of Gastroenterology, Research Center of Gastroenterology and Hepatology, University of Medicine and Pharmacy of Craiova, Petru Rareş, no. 2, 200349 Craiova, RomaniaDepartment of Pathophysiology, University of Medicine and Pharmacy of Craiova, Petru Rareş, no. 2, 200349 Craiova, Romania Ultrasound (US) imaging is a widely used, non-invasive method for detecting liver tumors and assessing parenchymal changes. However, the inherent variability and noise in US images pose challenges for accurate lesion identification. This study aims to develop and evaluate a deep learning (DL) model capable of performing real-time segmentation of liver lesions in US scans. A dataset of 50 video examinations was used, from which frames were extracted and manually annotated by an experienced gastroenterologist. The segmentation process was conducted using a U-Net architecture with focal Tversky loss (FTL) to address class imbalance. Two versions of the model were trained with different FTL parameters: Model 1 (α = β = 0.5, γ = 1) and Model 2 (α = 0.7, β = 0.3, γ = 0.75). The models were assessed based on key performance metrics, including intersection over union (IoU), recall, and precision. Model 1 achieved a higher IoU score (0.84) than Model 2. Both models demonstrated inference times between 30 and 80 milliseconds, confirming their feasibility for real-time US applications. Visual analysis showed that Model 1 produced more precise and contiguous lesion segmentation, whereas Model 2 tended to separate lesions that were close together. These findings suggest that the proposed DL models are effective in real-time liver lesion segmentation in US imaging. Model 1, which utilized balanced FTL parameters, demonstrated superior segmentation accuracy. https://ami.info.umfcluj.ro/index.php/AMI/article/view/1194Ultrasound ImagingLiver TumorSegmentationU-NetFocal Tversky LossReal-time detection of PCR products
spellingShingle Marinela-Cristiana URHUŢ
Larisa Daniela SĂNDULESCU
Costin Teodor STREBA
Mădălin MĂMULEANU
Adriana CIOCÂLTEU
Sergiu Marian CAZACU
Suzana DĂNOIU
Real-Time Liver Lesion Segmentation in Ultrasound Imaging using Deep Learning
Applied Medical Informatics
Ultrasound Imaging
Liver Tumor
Segmentation
U-Net
Focal Tversky Loss
Real-time detection of PCR products
title Real-Time Liver Lesion Segmentation in Ultrasound Imaging using Deep Learning
title_full Real-Time Liver Lesion Segmentation in Ultrasound Imaging using Deep Learning
title_fullStr Real-Time Liver Lesion Segmentation in Ultrasound Imaging using Deep Learning
title_full_unstemmed Real-Time Liver Lesion Segmentation in Ultrasound Imaging using Deep Learning
title_short Real-Time Liver Lesion Segmentation in Ultrasound Imaging using Deep Learning
title_sort real time liver lesion segmentation in ultrasound imaging using deep learning
topic Ultrasound Imaging
Liver Tumor
Segmentation
U-Net
Focal Tversky Loss
Real-time detection of PCR products
url https://ami.info.umfcluj.ro/index.php/AMI/article/view/1194
work_keys_str_mv AT marinelacristianaurhut realtimeliverlesionsegmentationinultrasoundimagingusingdeeplearning
AT larisadanielasandulescu realtimeliverlesionsegmentationinultrasoundimagingusingdeeplearning
AT costinteodorstreba realtimeliverlesionsegmentationinultrasoundimagingusingdeeplearning
AT madalinmamuleanu realtimeliverlesionsegmentationinultrasoundimagingusingdeeplearning
AT adrianaciocalteu realtimeliverlesionsegmentationinultrasoundimagingusingdeeplearning
AT sergiumariancazacu realtimeliverlesionsegmentationinultrasoundimagingusingdeeplearning
AT suzanadanoiu realtimeliverlesionsegmentationinultrasoundimagingusingdeeplearning