Scholarlink

<p>In the medical field, the limited availability of large-scale datasets and labor-intensive annotation processes hinder the performance of deep models. Diffusion-based generative augmentation approaches present a promising solution to this issue, having been proven effective in advancing downstream medical recognition tasks. Nevertheless, existing works lack sufficient semantic and sequential steerability for challenging video/3D sequence generation, and neglect quality control of noisy synthesized samples, resulting in unreliable synthetic databases and severely limiting the performance of downstream tasks. In this work, we present <i>Ctrl-GenAug</i>, a novel and general generative augmentation framework that enables highly semantic- and sequential-customized sequence synthesis and suppresses incorrectly synthesized samples, to aid medical sequence classification. Specifically, we first design a multimodal conditions-guided sequence generator for controllably synthesizing diagnosis-promotive samples. A sequential augmentation module is integrated to enhance the temporal/stereoscopic coherence of generated samples. Then, we propose a noisy synthetic data filter to suppress unreliable cases at the semantic and sequential levels. Extensive experiments on 5 medical datasets with 4 different modalities, including comparisons with 15 augmentation methods and evaluations using 11 networks trained on 3 paradigms, comprehensively demonstrate the effectiveness and generality of <i>Ctrl-GenAug</i>, particularly with pronounced performance gains in underrepresented high-risk populations and out-domain conditions. Codes, models, and synthetic databases are available at <a href="https://github.com/XinRuiZhou0106/Ctrl-GenAug">https://github.com/XinRuiZhou0106/Ctrl-GenAug</a>.</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

Ctrl-GenAug: Controllable Generative Augmentation for Medical Sequence Classification

  • 期刊论文

  • 发表日期：

    2026年4月7日

查看全文
  • Xinrui Zhou,
  • Yuhao Huang,
  • Haoran Dou,
  • Shijing Chen,
  • Ao Chang,
  • Jia Liu,
  • Weiran Long,
  • Jian Zheng,
  • Erjiao Xu,
  • Jie Ren,
  • Alejandro F. Frangi,
  • Ruobing Huang,
  • Jun Cheng,
  • Xiaomeng Li,
  • Wufeng Xue,
  • Dong Ni

摘要

In the medical field, the limited availability of large-scale datasets and labor-intensive annotation processes hinder the performance of deep models. Diffusion-based generative augmentation approaches present a promising solution to this issue, having been proven effective in advancing downstream medical recognition tasks. Nevertheless, existing works lack sufficient semantic and sequential steerability for challenging video/3D sequence generation, and neglect quality control of noisy synthesized samples, resulting in unreliable synthetic databases and severely limiting the performance of downstream tasks. In this work, we present Ctrl-GenAug, a novel and general generative augmentation framework that enables highly semantic- and sequential-customized sequence synthesis and suppresses incorrectly synthesized samples, to aid medical sequence classification. Specifically, we first design a multimodal conditions-guided sequence generator for controllably synthesizing diagnosis-promotive samples. A sequential augmentation module is integrated to enhance the temporal/stereoscopic coherence of generated samples. Then, we propose a noisy synthetic data filter to suppress unreliable cases at the semantic and sequential levels. Extensive experiments on 5 medical datasets with 4 different modalities, including comparisons with 15 augmentation methods and evaluations using 11 networks trained on 3 paradigms, comprehensively demonstrate the effectiveness and generality of Ctrl-GenAug, particularly with pronounced performance gains in underrepresented high-risk populations and out-domain conditions. Codes, models, and synthetic databases are available at https://github.com/XinRuiZhou0106/Ctrl-GenAug.