Mixture of Expert-Based SoftMax-Weighted Box Fusion for Robust Lesion Detection in Ultrasound Imaging DOI Creative Commons
Se-Yeol Rhyou,

Ming Yu,

Jae-Chern Yoo

et al.

Diagnostics, Journal Year: 2025, Volume and Issue: 15(5), P. 588 - 588

Published: Feb. 28, 2025

Background/Objectives: Ultrasound (US) imaging plays a crucial role in the early detection and treatment of hepatocellular carcinoma (HCC). However, challenges such as speckle noise, low contrast, diverse lesion morphology hinder its diagnostic accuracy. Methods: To address these issues, we propose CSM-FusionNet, novel framework that integrates clustering, SoftMax-weighted Box Fusion (SM-WBF), padding. Using raw US images from leading hospital, Samsung Medical Center (SMC), applied intensity adjustment, adaptive histogram equalization, low-pass, high-pass filters to reduce noise enhance resolution. Data augmentation generated ten per one image, allowing training 10 YOLOv8 networks. The [email protected] each network was used SoftMax-derived weights SM-WBF. Threshold-lowered bounding boxes were clustered using Density-Based Spatial Clustering Applications with Noise (DBSCAN), outliers managed within clusters. SM-WBF reduced redundant boxes, padding enriched features, improving classification Results: accuracy improved 82.48% 97.58% sensitivity reaching 100%. increased 56.11% 95.56% after clustering Conclusions: CSM-FusionNet demonstrates potential significantly improve reliability US-based detection, aiding precise clinical decision-making.

Language: Английский

Mixture of Expert-Based SoftMax-Weighted Box Fusion for Robust Lesion Detection in Ultrasound Imaging DOI Creative Commons
Se-Yeol Rhyou,

Ming Yu,

Jae-Chern Yoo

et al.

Diagnostics, Journal Year: 2025, Volume and Issue: 15(5), P. 588 - 588

Published: Feb. 28, 2025

Background/Objectives: Ultrasound (US) imaging plays a crucial role in the early detection and treatment of hepatocellular carcinoma (HCC). However, challenges such as speckle noise, low contrast, diverse lesion morphology hinder its diagnostic accuracy. Methods: To address these issues, we propose CSM-FusionNet, novel framework that integrates clustering, SoftMax-weighted Box Fusion (SM-WBF), padding. Using raw US images from leading hospital, Samsung Medical Center (SMC), applied intensity adjustment, adaptive histogram equalization, low-pass, high-pass filters to reduce noise enhance resolution. Data augmentation generated ten per one image, allowing training 10 YOLOv8 networks. The [email protected] each network was used SoftMax-derived weights SM-WBF. Threshold-lowered bounding boxes were clustered using Density-Based Spatial Clustering Applications with Noise (DBSCAN), outliers managed within clusters. SM-WBF reduced redundant boxes, padding enriched features, improving classification Results: accuracy improved 82.48% 97.58% sensitivity reaching 100%. increased 56.11% 95.56% after clustering Conclusions: CSM-FusionNet demonstrates potential significantly improve reliability US-based detection, aiding precise clinical decision-making.

Language: Английский

Citations

0