Scholarlink

<p>A brain tumor refers to the irregular growth of cells inside brain or neighboring structures. Brain tumors can be benign or malignant, and treatment varies based on tumor size, tumor location, and patient health. Early and precise segmentation is important for active management and treatment planning. To address this, a Stock Exchange Trading Dingo Optimization Algorithm-based Medical Transformer U-Net++ (SETDO-MedNet++) is established for segmenting brain tumors using multi-modal imaging data, including T2-weighted, T2 Fluid Attenuated Inversion Recovery (FLAIR), and contrast T1-weighted (T1Gd) images. Initially, the Native T1 image is pre-processed using an adaptive median filter, followed by image transformation using Stationary Wavelet (SWT). Brain tumor segmentation is then performed by MedNet++, which integrates U-Net + + and Medical Transformer. Then, the weights of MedNet + + are fine-tuned by the SETDO. This process is repeated for all modalities, and the outputs are fused using the Deep Q Network (DQN) trained by SETDO to obtain the final output. The approach was evaluated on two standard datasets, namely the BraTS 2018 dataset and BraTS 2020 dataset and achieved an Intersection Over Union (IOU) of 94.172%, segmentation accuracy of 95.622%%, and dice coefficient of 96.98%. The proposed SETDO-MedNet + + method can help radiologists in early diagnosis and treatment planning by providing accurate and reliable tumor segmentation. Limitations include class imbalance, which may affect segmentation performance for underrepresented tumor types, and reliance on multi-modal inputs, which may reduce performance if some imaging modalities are missing.</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

Multi-Modal Brain Tumor Segmentation Using Optimized MedNet++

  • 期刊论文

  • 发表日期：

    2025年12月15日

查看全文
  • Mohana Saranya S,
  • Rajalaxmi Rajammal Ramasamy,
  • Dinesh Komarasamy,
  • Suganya T

摘要

A brain tumor refers to the irregular growth of cells inside brain or neighboring structures. Brain tumors can be benign or malignant, and treatment varies based on tumor size, tumor location, and patient health. Early and precise segmentation is important for active management and treatment planning. To address this, a Stock Exchange Trading Dingo Optimization Algorithm-based Medical Transformer U-Net++ (SETDO-MedNet++) is established for segmenting brain tumors using multi-modal imaging data, including T2-weighted, T2 Fluid Attenuated Inversion Recovery (FLAIR), and contrast T1-weighted (T1Gd) images. Initially, the Native T1 image is pre-processed using an adaptive median filter, followed by image transformation using Stationary Wavelet (SWT). Brain tumor segmentation is then performed by MedNet++, which integrates U-Net + + and Medical Transformer. Then, the weights of MedNet + + are fine-tuned by the SETDO. This process is repeated for all modalities, and the outputs are fused using the Deep Q Network (DQN) trained by SETDO to obtain the final output. The approach was evaluated on two standard datasets, namely the BraTS 2018 dataset and BraTS 2020 dataset and achieved an Intersection Over Union (IOU) of 94.172%, segmentation accuracy of 95.622%%, and dice coefficient of 96.98%. The proposed SETDO-MedNet + + method can help radiologists in early diagnosis and treatment planning by providing accurate and reliable tumor segmentation. Limitations include class imbalance, which may affect segmentation performance for underrepresented tumor types, and reliance on multi-modal inputs, which may reduce performance if some imaging modalities are missing.