Scholarlink

<p>In advanced surgical technology, the precise segmentation of tumor in the kidney is essential for radiomic analysis. Traditionally, visual assessment of images is utilized for the segmentation of tumors in the kidney, and these visuals are collected using computed tomography (CT) scans. The success rate of using this method is based on previous experiments, and well-trained doctors are needed to analyse the kidney tumor. However, due to the large volume of CT images, a detailed explanation is necessary for diagnosis, which can lead to inconsistent results. Therefore, neural networks are implemented for the “automatic detection of tumors in the kidney using CT images”. In most approaches, the network architecture is customized to enhance the exactness of detection. Initially, the CT images were collected using recognized benchmark datasets. Segmentation of the gathered CT images is now underway. In this work, the adaptive 3D Trans-DenseUNet+ (A3D-TDUNet+) model is implemented as the segmentation model. The parameters in the A3D-TDUnet+ scheme are optimized with the aid of the improved artificial gorilla troops optimizer (IAGTO). The segmented images from the A3D-TDUNet+ are then fed to the “vision transformer (ViT)-based multi-scale EfficientNetB7 (ViT-MS-ENetB7)”. From the ViT-MS-ENetB7 model, the final kidney tumor detection outcome is obtained. A series of experiments is executed to validate the performance of the implemented ViT-MS-ENetB7 in the kidney tumor detection method.</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 vision transformer-based multiscale EfficientNetB7 model for detecting kidney tumor from CT segmented images using IAGTO

  • 期刊论文

  • 发表日期：

    2026年3月6日

查看全文
  • Anupkumar Bhatulal Jayaswal,
  • Mahesh Bhimsham Dembrani

摘要

In advanced surgical technology, the precise segmentation of tumor in the kidney is essential for radiomic analysis. Traditionally, visual assessment of images is utilized for the segmentation of tumors in the kidney, and these visuals are collected using computed tomography (CT) scans. The success rate of using this method is based on previous experiments, and well-trained doctors are needed to analyse the kidney tumor. However, due to the large volume of CT images, a detailed explanation is necessary for diagnosis, which can lead to inconsistent results. Therefore, neural networks are implemented for the “automatic detection of tumors in the kidney using CT images”. In most approaches, the network architecture is customized to enhance the exactness of detection. Initially, the CT images were collected using recognized benchmark datasets. Segmentation of the gathered CT images is now underway. In this work, the adaptive 3D Trans-DenseUNet+ (A3D-TDUNet+) model is implemented as the segmentation model. The parameters in the A3D-TDUnet+ scheme are optimized with the aid of the improved artificial gorilla troops optimizer (IAGTO). The segmented images from the A3D-TDUNet+ are then fed to the “vision transformer (ViT)-based multi-scale EfficientNetB7 (ViT-MS-ENetB7)”. From the ViT-MS-ENetB7 model, the final kidney tumor detection outcome is obtained. A series of experiments is executed to validate the performance of the implemented ViT-MS-ENetB7 in the kidney tumor detection method.