Scholarlink

In recent years, the emergence of deep learning (DL) has triggered a transformative revolution within the realm of medical imaging. Its profound impact lies in its ability to confront complex medical image analysis tasks with unmatched precision and efficiency. This paper delves into the domain of DL, focusing on its application to optical coherence tomography (OCT), a critical modality in biomedical imaging. The core of this research revolves around the utilization of a multi-class dataset tailored specifically for OCT, featuring four distinct categories. In parallel, the study extends its reach to DermaMNIST, a comprehensive dataset featuring a diverse array of common pigmented skin lesions for multi-class classification. Experiments are conducted using MedMNIST v2, a comprehensive, lightweight benchmark for 2D biomedical image classification. It not only illuminates the untapped potential of DL within the realm of OCT and dermatological imaging but also underscores the pivotal role of harnessing AI-driven solutions in the ever-evolving landscape of medical imaging. This study focuses on a deep feature ensemble methodology, which combines distinctive features from two pre-trained DL models such as DenseNet121 and InceptionV3 to enhance predictive accuracy. The research presents a compelling narrative that emphasizes the broader implications and applications of DL in the intricate field of biomedical image classification, spanning both the domains of OCT and dermatology. The source codes related to this work are available at https://github.com/rajatrajoria/A-Deep-Feature-Ensemble-Methodology-for-2D-Biomedical-Image-Classification .

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 Deep Feature Ensemble Methodology for 2D Biomedical Image Classification

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Rajat Rajoria,
  • Balmukund Kanodia,
  • Debam Saha,
  • Pawan Kumar Singh

摘要

In recent years, the emergence of deep learning (DL) has triggered a transformative revolution within the realm of medical imaging. Its profound impact lies in its ability to confront complex medical image analysis tasks with unmatched precision and efficiency. This paper delves into the domain of DL, focusing on its application to optical coherence tomography (OCT), a critical modality in biomedical imaging. The core of this research revolves around the utilization of a multi-class dataset tailored specifically for OCT, featuring four distinct categories. In parallel, the study extends its reach to DermaMNIST, a comprehensive dataset featuring a diverse array of common pigmented skin lesions for multi-class classification. Experiments are conducted using MedMNIST v2, a comprehensive, lightweight benchmark for 2D biomedical image classification. It not only illuminates the untapped potential of DL within the realm of OCT and dermatological imaging but also underscores the pivotal role of harnessing AI-driven solutions in the ever-evolving landscape of medical imaging. This study focuses on a deep feature ensemble methodology, which combines distinctive features from two pre-trained DL models such as DenseNet121 and InceptionV3 to enhance predictive accuracy. The research presents a compelling narrative that emphasizes the broader implications and applications of DL in the intricate field of biomedical image classification, spanning both the domains of OCT and dermatology. The source codes related to this work are available at https://github.com/rajatrajoria/A-Deep-Feature-Ensemble-Methodology-for-2D-Biomedical-Image-Classification .