Scholarlink

<p>Being among the leading causes of cancer mortality, liver cancer necessitates reliable techniques for the accurate identification and classification of Focal Liver Lesions (FLLs), which are essential for early diagnosis and treatment planning. Contrast-Enhanced Ultrasound (CEUS) is popularly adopted in liver imaging due to its safety and real-time capability; however, automated FLL classification is still made difficult by speckle noise and weak contrast and inter-class similarity among lesions. This study proposes a Parallel Efficient Convolutional Neural Network (PECN-Net) for accurate FLL classification from CEUS video frames. In this research, CEUS video is regarded as input, and candidate frames are extracted in the phase of frame extraction, and subsequently, the required regions are extracted using Region of Interest (RoI). Later, an Adaptive Dynamically Weighted Median Filter (ADWMF) is utilized to denoise the image, and then the denoised image is fed to the feature extraction stage to extract the features, like Local Texton XOR (LTXOR) with Gray-Level Co-Occurrence Matrix (GLCM). Subsequently, the FLL is detected through the Parallel Convolutional Neural Network (PCNN), and classified using the developed Parallel Efficient Convolutional Neural Network (PECN-Net) as Hepatocellular Carcinoma (HCC), Hemangioma (HEM), Focal Nodular Hyperplasia (FNH), and Intrahepatic Cholangiocarcinoma (ICC). The PECN-Net model is developed using two networks, such as PCNN and EfficientNet. Hence, the PECN-Net attained the True Positive Rate (TPR), accuracy, True Negative Rate (TNR), and F-Measure values of 96.28%, 94%, 92.88%, and 95.38% at the K-fold cross-validation with K = 8.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Novel LTXOR with GLCM Features and Parallel Efficient Convolutional Neural Network for Focal Liver Lesion Detection and Classification Using CEUS Video Clip

  • 期刊论文

  • 发表日期：

    2026年6月26日

查看全文
  • Nampally Vijay Kumar,
  • Suresh Kumar Krishnamoorthy,
  • Velumani Ramaswamy,
  • Sreenu Ponnada,
  • Jayasri Kotti,
  • Suganthi Nagarajan

摘要

Being among the leading causes of cancer mortality, liver cancer necessitates reliable techniques for the accurate identification and classification of Focal Liver Lesions (FLLs), which are essential for early diagnosis and treatment planning. Contrast-Enhanced Ultrasound (CEUS) is popularly adopted in liver imaging due to its safety and real-time capability; however, automated FLL classification is still made difficult by speckle noise and weak contrast and inter-class similarity among lesions. This study proposes a Parallel Efficient Convolutional Neural Network (PECN-Net) for accurate FLL classification from CEUS video frames. In this research, CEUS video is regarded as input, and candidate frames are extracted in the phase of frame extraction, and subsequently, the required regions are extracted using Region of Interest (RoI). Later, an Adaptive Dynamically Weighted Median Filter (ADWMF) is utilized to denoise the image, and then the denoised image is fed to the feature extraction stage to extract the features, like Local Texton XOR (LTXOR) with Gray-Level Co-Occurrence Matrix (GLCM). Subsequently, the FLL is detected through the Parallel Convolutional Neural Network (PCNN), and classified using the developed Parallel Efficient Convolutional Neural Network (PECN-Net) as Hepatocellular Carcinoma (HCC), Hemangioma (HEM), Focal Nodular Hyperplasia (FNH), and Intrahepatic Cholangiocarcinoma (ICC). The PECN-Net model is developed using two networks, such as PCNN and EfficientNet. Hence, the PECN-Net attained the True Positive Rate (TPR), accuracy, True Negative Rate (TNR), and F-Measure values of 96.28%, 94%, 92.88%, and 95.38% at the K-fold cross-validation with K = 8.