发表机构
University of Liverpool; Shanghai Artificial Intelligence Laboratory; Liverpool Heart and Chest Hospital（利物浦大学; 上海人工智能实验室; 利物浦心胸医院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究针对ECG到PPG知识蒸馏多学习身份而非心血管状态的问题，提出固定效应知识蒸馏方法消除记录均值特质，使状态一致性翻倍，在多骨干模型和数据库上均有效，助力可穿戴设备追踪个体心血管状态变化。
AI 中文摘要
心电图（ECG）被广泛用于仅使用光电容积描记（PPG）的模型的训练，但其所传递的信息尚未得到充分研究。可穿戴设备的价值在于追踪个体心血管状态的变化，但ECG到PPG的知识蒸馏大多学习的是个体身份。每个记录的均值（即“特质”）占冻结ECG教师模型目标的40%-59%，而合并训练的学生模型会记住该特质，却无法将其迁移至新记录。原始对齐余弦相似度未能捕捉到这一点，因为常数预测器的得分为0.793。在34次运行中，学生模型记住的身份信息越多，其学习到的状态信息就越少。固定效应知识蒸馏会从预测值和目标值中减去每个记录的均值，从而精确消除该特质，而合并锚点则保留该特质。状态一致性提升了一倍以上，个体内标签得到改善，而年龄和性别相关的标签未受影响，且该增益在两种骨干模型和两个额外数据库上均成立。基于记录进行条件设置可使知识蒸馏转向可穿戴设备所监测的个体内变化。
英文摘要
ECG is widely used to teach PPG-only models, yet what it teaches is unexamined. Wearables are valued for tracking how a person's cardiovascular state changes, but ECG-to-PPG distillation mostly learns who the person is. A per-recording mean, the trait, holds 40-59% of a frozen ECG teacher's target, and pooled students memorise it without carrying it to new recordings. The raw alignment cosine misses this, since a constant predictor scores 0.793. Across 34 runs, the more identity a student memorises, the less state it learns. Fixed-effects distillation subtracts each recording's mean from prediction and target, so the trait cancels exactly, while a pooled anchor keeps it. State agreement more than doubles, within-person labels improve while age and sex do not, and the gain holds on two backbones and two further databases. Conditioning on the recording turns distillation toward the within-person changes that wearables monitor.
