研究
OTTER: Optimized Training with Trustworthy Enhanced Replication via Diffusion and Federated VMUNet for Privacy-Aware Medical Segmentation (Oct 2025)
- 描述 In medical domain, the scarcity and sensitivity of imaging data present significant challenges to the development of accurate and privacy-preserving segmentation models. In this paper, an Optimized Training framework with Trustworthy Enhanced Replication (OTTER) via diffusion and federated VMUNet is proposed to solve these problems. OTTER leverages diffusion-based generative models to synthesize diverse and high-fidelity medical images that reflect the statistical distribution of real data while avoiding direct exposure of patient information. These synthetic samples are further filtered using a Bayesian quality estimator to ensure training on only reliable and representative data. These trusted replicas are integrated into a federated learning pipeline built on VMUNet, a state-space-based segmentation architecture that enables decentralized training across medical centers without sharing raw data. Experiments on multiple public medical image segmentation datasets have demonstrated that OTTER achieves competitive or superior performance in terms of accuracy, robustness, and generalization, while maintaining strong privacy guarantees. Our results highlight the potential of combining generative data augmentation with federated architectures to advance privacy-aware medical AI.
- Haocheng Kan, Yuesheng Zhu, Guibo Luo and Hanwen Zhang. OTTER: Optimized Training with Trustworthy Enhanced Replication via Diffusion and Federated VMUNet for Privacy-Aware Medical Segmentation
- 關鍵字 Federated Learning 、Privacy-Preserving 、Diffusion Models 、Medical Image Segmentation
學業成就分析系統之研究:使用集群分析與視覺化 (Aug 2015 – Sep 2017)
- 描述 該研究專案開發了一個交互式學業成績分析系統,並分析 2008 年以後進入龍華科技大學管理學院的學生在不同類型的課程中的表現,包含他們的經濟狀況和畢業後的就業情況進行分析。 從而可以根據個別學生的初始情況預測畢業後的職業發展。
- Hao-Cheng Kan & Fang-Ling Lin (10 Dec 2016). The Impact of Student Loans on Academic Performance: A Cluster and Visual Analysis. presented at the 22th CSIM IMP, Taichung, Taiwan.
- 程式語言和軟體使用 R、 RShiny Server、 MariaDB 和 Ubuntu Server