arXiv：cs.LG· Xinye Yang, Zhusi Zhong, Scott Collins, Michael Bernstein, Grayson Baird, Terrence Healey, Michael Atalay, Mahesh Jayaraman, Xuyu Wang, Zhicheng Jiao·· 3 小时前AI 评分39
VRM：面向医学影像的置信门控云边级联分诊框架
Confidence-Gated Cloud-Edge Cascade Triage via Variational Risk Minimization for Medical Imaging
AI 导读
研究者提出变分风险最小化（VRM）蒸馏框架，将 LVLM 生成的报告变体视为潜在临床解读的蒙特卡洛样本，从变分边缘化的教师分布中学习，实现缺失模态下的不确定性感知监督。在紧凑边缘学生实例中，置信门控级联达到 AUC 0.941，平均延迟 103ms，云端升级率 20.3%。该工作为云边临床工作流提供了明确的可靠性-延迟操作点，已被 Smart Health 41 (2026) 100689 接收。
正文
Abstract:Emergency chest X-ray (CXR) triage has a structural modality gap: reports arrive after triage decisions, yet multimodal foundation models require image-text inputs. We present Variational Risk Minimization (VRM), a distillation framework that treats LVLM-generated report variants as Monte Carlo samples of latent clinical interpretations. Rather than distilling from a single teacher target, VRM learns from a variationally marginalized teacher distribution, enabling uncertainty-aware supervision under missing-modality constraints. Under matched encoder families, VRM outperforms direct fine-tuning baselines and improves calibration with strong recovery from hallucinated supervision. Marginalized supervision reduces report-selection instability. In our compact edge-student instantiation, a confidence-gated cascade reaches AUC 0.941 at 103ms average latency with 20.3% cloud escalation, yielding an explicit reliability-latency operating point for cloud-edge clinical workflows.
| Comments: | 14 pages, 6 figures, 14 tables. Accepted manuscript of the article published in Smart Health 41 (2026) 100689. Presented as an oral at IEEE/ACM CHASE 2026. Code: this https URL |
| Subjects: | Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.02269 [eess.IV] |
| (or arXiv:2610.02269v1 [eess.IV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.02269 arXiv-issued DOI via DataCite |
|
| Journal reference: | Smart Health 41 (2026) 100689 |
| Related DOI: | https://doi.org/10.1016/j.smhl.2026.100689
DOI(s) linking to related resources |
Submission history
From: Xinye Yang [view email]
[v1]
Thu, 1 Oct 2026 06:15:02 UTC (967 KB)
来源：arXiv：cs.LG · arxiv.org