PPG-Bench: A Unified Benchmark for PPG Foundation Models
作者: Arvasu Kulkarni, Mahir Jain, Aditya Ray Mishra, Sandeep Singh, Siddharth Panwar
分类: cs.LG
发布日期: 2026-10-09
🔗 代码/项目: GITHUB
💡 一句话要点
提出PPG-Bench以统一评估PPG基础模型的性能
🎯 匹配领域: 支柱八:物理动画 (Physics-based Animation) 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 光电容积脉搏波 基础模型 性能评估 统一基准 健康监测 机器学习
📋 核心要点
- 现有PPG基础模型在不同评估中使用不同的数据集和任务定义,导致比较困难,缺乏统一标准。
- 本文提出PPG-Bench,作为一个统一的基准,评估多个PPG基础模型在不同任务中的表现,提供开放源码。
- 实验结果显示,35.3%的任务组合未能显示出模型的优势,表明当前PPG基础模型的通用性和可转移性不足。
📝 摘要(中文)
光电容积脉搏波(PPG)信号在可穿戴和临床设备中被广泛记录,然而现有的PPG基础模型(FMs)在不同评估中使用了不同的数据集、任务定义和预处理方法,导致比较困难。为了解决这一问题,本文提出了PPG-Bench,一个统一的开源基准,评估五个公共PPG FMs在九个开放数据集和26个任务中的表现。我们在四种下游任务适应性下评估模型,并与随机基线、脉率变异性特征集及通用时间序列模型进行对比。结果显示,35.3%的任务组合未能显示出模型的优势,83.8%的情况下PPG特定模型未能超越简单基线。当前PPG FMs尚不能提供通用的PPG信号表示,未来评估应包括简单控制的检查。
🔬 方法详解
问题定义:本文旨在解决现有PPG基础模型在评估时缺乏统一标准的问题,现有方法在不同数据集和任务定义下的比较存在困难。
核心思路:PPG-Bench通过统一评估框架,比较不同PPG基础模型在多个任务上的表现,提供开放源码以促进研究者的使用和扩展。
技术框架:整体架构包括数据集选择、模型评估、下游任务适应性测试及对比分析,涵盖了线性探测、随机森林、冻结MLP探测器和端到端微调等四种下游任务适应性。
关键创新:PPG-Bench的创新在于其系统性地评估了五个公共PPG基础模型在26个任务上的表现,提供了对比基线和控制组,揭示了当前模型的局限性。
关键设计:在评估过程中,采用了模型特定的超参数优化和受试者不重叠的交叉验证,确保评估的公正性和准确性。
🖼️ 关键图片
📊 实验亮点
实验结果显示,在68个任务-适应性组合中,有35.3%的组合未能显示出模型相对于随机基线的优势,83.8%的情况下,PPG特定模型未能超越脉率变异性特征集和通用时间序列模型,表明当前模型的有效性不足。
🎯 应用场景
该研究的潜在应用领域包括健康监测、运动分析和睡眠研究等。通过提供统一的评估基准,PPG-Bench可以帮助研究人员更好地比较和改进PPG基础模型,推动相关技术的发展和应用。
📄 摘要(原文)
Photoplethysmography (PPG) is increasingly recorded by wearable and clinical devices at enormous scale, and a growing number of PPG foundation models (FMs) have been developed to compute reusable representations for various downstream tasks. These models are difficult to compare: different evaluations use different datasets, task definitions, preprocessing, downstream heads, hyperparameter tuning, and baselines. We present PPG-Bench, a unifying open-source benchmark evaluating five public PPG FMs across nine open-access datasets and 26 tasks spanning cardiovascular state estimation, activity, sleep, screening, and identification. We evaluate models under four downstream regimes of increasing adaptability (linear probing, random forests, frozen MLP probes, and end-to-end finetuning), with model-specific hyperparameter optimization and subject-disjoint cross-validation. We also evaluate against three controls: random/constant baseline, a deterministic pulse-rate-variability (PRV) feature set, and a generic time-series FM (MOMENT). Across all 68 task-regime combinations, 35.3% show no meaningful FM advantage over the random/constant baseline, and in 83.8% no PPG-specific FM outperforms both PRV and MOMENT. We also find that added downstream head capacity generally improves performance, but can cause significant degradation if the model is performing at chance/constant level. Current PPG FMs therefore cannot yet claim to provide generalised and transferrable representations of PPG signals, and future evaluation in this field should include checks against simple, non-learned controls. Code for the benchmark and instructions to evaluate more models can be found here - https://github.com/mannas-ai/ppg-bench.