Representation Is Not Enough: Body-Localized Thermal Evidence for Contactless Stress and Craving Sensing in Opioid Use Disorder
作者: Sachin Deb, Harshit Sharma, Asif Salekin
分类: cs.CV, cs.LG
发布日期: 2026-08-17
💡 一句话要点
提出FABLE-Therm以解决无接触压力与渴望感知问题
🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 无接触监测 压力感知 阿片类药物使用障碍 弱监督学习 局部证据融合 生理信号处理 个性化医疗
📋 核心要点
- 现有的生理监测方法依赖可穿戴设备,缺乏无接触监测的时空定位能力,导致压力感知的准确性不足。
- FABLE-Therm通过弱监督架构,保留身体各区域的局部证据,解决了无接触压力感知中的证据定位问题。
- 在阿片类药物使用障碍的实验中,FABLE-Therm实现了0.938的AUROC,且其学习到的表示能够有效转移到自我报告的渴望感知上。
📝 摘要(中文)
本研究探讨了无接触压力感知的挑战,提出了FABLE-Therm架构,该架构在身体区域、时间和编码器特征上保留局部证据,解决了弱监督证据定位问题。FABLE-Therm通过在嵌入层融合冻结的基础模型编码器,展示了局部融合优于特征连接和预测平均的理论基础。研究表明,该方法在阿片类药物使用障碍(OUD)中取得了0.938的AUROC,并首次证明了可以从无接触热视频中恢复渴望感知,支持个体层面的分析,揭示了模型失败的原因。
🔬 方法详解
问题定义:本研究旨在解决无接触压力感知中的证据定位问题。现有方法依赖可穿戴设备,无法有效捕捉压力反应的时空信息,导致监测效果不佳。
核心思路:FABLE-Therm架构通过弱监督学习,保留身体各区域的局部证据,确保在最终决策时能够追溯到产生信号的具体时刻和部位,从而提高感知的准确性。
技术框架:FABLE-Therm的整体架构包括多个模块,首先是对固定热视频的处理,然后通过融合多个冻结的基础模型编码器,在嵌入层进行局部证据的整合,最后进行决策输出。
关键创新:FABLE-Therm的主要创新在于局部证据的融合策略,理论上证明了局部融合优于传统的特征连接和预测平均方法,这一设计使得模型在处理复杂的生理信号时更具鲁棒性。
关键设计:在模型设计中,采用了特定的损失函数以优化局部证据的保留,同时在网络结构上,结合了多个基础模型的编码器,以增强特征的表达能力。
🖼️ 关键图片
📊 实验亮点
FABLE-Therm在阿片类药物使用障碍的实验中实现了0.938的AUROC,显示出其在无接触压力感知中的高效性。此外,该模型的学习表示能够有效转移到自我报告的渴望感知上,首次证明了无接触热视频的应用潜力。
🎯 应用场景
该研究的潜在应用领域包括心理健康监测、成瘾治疗和个性化医疗。通过无接触的方式监测压力和渴望,可以为患者提供更灵活的支持,尤其是在早期康复阶段,减少对可穿戴设备的依赖,提升治疗效果。
📄 摘要(原文)
Removing wearables from physiological monitoring also removes their supervision: the signal indicating where and when a stress response occurred. Contactless stress sensing therefore becomes a weakly supervised evidence-localization problem, where a clip-level label must be traced to the body regions and moments that produced it. We address this with FABLE-Therm, a weakly supervised architecture that preserves localized evidence across body regions, time, and encoder-specific representations until the final decision. FABLE-Therm fuses frozen foundation-model encoders at the embedding level, with theory explaining why localized fusion can outperform feature concatenation and prediction averaging. We study this problem in opioid use disorder (OUD), where stress is a major relapse trigger and sustained wearable use can be difficult during early recovery. Using fixed thermal video, FABLE-Therm achieves 0.938 AUROC on held-out participants, and its learned representation transfers to self-reported craving, providing, to our knowledge, the first evidence that craving can be recovered from contactless thermal video. Localized evidence also enables participant-level analysis of deployment failure. We find that improving representation alone is insufficient for equitable deployment: additional data from the underserved group would recover only about half of the cohort gap, while the remainder reflects person-to-person heterogeneity. This modality-agnostic decomposition applies to models with identifiable subpopulations. Together with the first cohort-structured contactless thermal OUD benchmark, our results show that preserving localized evidence supports both accurate sensing and principled analysis of who a model fails and why.