Scholarlink

<p>Accurate and automatic brain tumor segmentation (BTS) plays a vital role in radiation therapy, treatment planning and extending patients lives. Early diagnosis and correct treatment significantly improve cancer survival rates. Segmenting these lesions is a challenging task because brain tumors have significant structural variability. Studies on BTS have become more prevalent in the fields of computer science and healthcare, with good accuracy and great interpretability. However, comprehensive and up-to-date reviews summarizing recent (2017-2023) advancements, particularly focusing on CNN, GAN, and hybrid DL strategies for multimodal BTS, are scarce. This review aims to address this gap by providing a systematic synthesis of the current state-of-the-art approaches. The primary objective of this review is to systematically analyze and consolidate recent research on CNN, GAN, U-Net, and hybrid methods for BTS developed between 2017 and 2023. Our approach involves evaluating key aspects like hyperparameters, data processing techniques, modalities, evaluation metrics, datasets, loss functions, and DL libraries documented in over 200 selected scientific articles. Moreover, the diverse DL and hybrid-based BTS strategies of researchers are summarized by focusing on the main considerations, advantages, limitations and performance measures, to provide a better understanding of the existing methods. It highlights how deep learning and machine learning techniques can be combined to improve the accuracy and efficacy of brain tumor segmentation, which ultimately helps to enhance diagnostic and treatment plans for patients. Finally, the paper discusses research issues, their limitations and demanding situations for future research that assist radiologists in deciding the analysis of brain tumors. Furthermore, this review synthesizes the diverse strategies, evaluates the comparative efficacy of different approaches, highlights successful combinations of machine learning and deep learning, and concludes by identifying current research limitations and future directions.</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

Multimodal brain image segmentation: a recent review, challenges and future perspectives

  • 期刊论文

  • 发表日期：

    2025年8月4日

查看全文
  • Berlin Shaheema,
  • Naresh Babu Muppalaneni

摘要

Accurate and automatic brain tumor segmentation (BTS) plays a vital role in radiation therapy, treatment planning and extending patients lives. Early diagnosis and correct treatment significantly improve cancer survival rates. Segmenting these lesions is a challenging task because brain tumors have significant structural variability. Studies on BTS have become more prevalent in the fields of computer science and healthcare, with good accuracy and great interpretability. However, comprehensive and up-to-date reviews summarizing recent (2017-2023) advancements, particularly focusing on CNN, GAN, and hybrid DL strategies for multimodal BTS, are scarce. This review aims to address this gap by providing a systematic synthesis of the current state-of-the-art approaches. The primary objective of this review is to systematically analyze and consolidate recent research on CNN, GAN, U-Net, and hybrid methods for BTS developed between 2017 and 2023. Our approach involves evaluating key aspects like hyperparameters, data processing techniques, modalities, evaluation metrics, datasets, loss functions, and DL libraries documented in over 200 selected scientific articles. Moreover, the diverse DL and hybrid-based BTS strategies of researchers are summarized by focusing on the main considerations, advantages, limitations and performance measures, to provide a better understanding of the existing methods. It highlights how deep learning and machine learning techniques can be combined to improve the accuracy and efficacy of brain tumor segmentation, which ultimately helps to enhance diagnostic and treatment plans for patients. Finally, the paper discusses research issues, their limitations and demanding situations for future research that assist radiologists in deciding the analysis of brain tumors. Furthermore, this review synthesizes the diverse strategies, evaluates the comparative efficacy of different approaches, highlights successful combinations of machine learning and deep learning, and concludes by identifying current research limitations and future directions.