![[2409.17091] Ctrl-GenAug: Controllable Generative Augmentation for Medical Sequence Classification](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2409.17091] Ctrl-GenAug: Controllable Generative Augmentation for Medical Sequence Classification
Summary
The paper presents Ctrl-GenAug, a novel framework for controllable generative augmentation in medical sequence classification, addressing dataset limitations and enhancing model performance.
Why It Matters
With the increasing demand for accurate medical sequence classification, Ctrl-GenAug offers a solution to the challenges posed by limited datasets and noisy synthetic data. This framework enhances the quality and reliability of generated samples, which is crucial for improving diagnostic accuracy and patient outcomes in healthcare.
Key Takeaways
- Ctrl-GenAug improves the controllability of generative augmentation for medical sequences.
- The framework integrates a multimodal conditions-guided generator for better sample synthesis.
- A noisy synthetic data filter enhances the reliability of generated samples.
- Extensive experiments demonstrate effectiveness across various medical datasets.
- The approach is particularly beneficial for underrepresented high-risk populations.
Computer Science > Computer Vision and Pattern Recognition arXiv:2409.17091 (cs) [Submitted on 25 Sep 2024 (v1), last revised 18 Feb 2026 (this version, v3)] Title:Ctrl-GenAug: Controllable Generative Augmentation for Medical Sequence Classification Authors: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 View a PDF of the paper titled Ctrl-GenAug: Controllable Generative Augmentation for Medical Sequence Classification, by Xinrui Zhou and 15 other authors View PDF HTML (experimental) Abstract: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 sequen...
