Efficient Personalized Adaptation for Physiological Signal Foundation Model

Time series analysis is crucial across various fields like energy, environment, transportation, finance and health. Deep learning has significantly advanced this field, particularly, the Time Series Foundation Model (TSFM) excels in multiple domains due to extensive pre-training. In this work, we focus on TSFM's challenges in medical practice: limited computing resources and medical data privacy. TSFM variants include fine-tuned models and those pre-trained for rapid deployment on diverse data. There may not be enough computing resources to train physiological signals locally in hospitals, and generalized TSFM is still inferior to task-specific methods on private, imbalanced local data. To address this, we propose PhysioPFM, a framework for efficiently personalizing TSFM. Our approach involves low-rank pre-training on public datasets, generator training by trained LoRA weights, and efficient weight generation via local data. Experimental results demonstrate that integrating generated models with TSFM enhances performance, and transferability, and reduces the need for additional sensitive data training.

We teach computers to diagnose and classify by training them on a large number of labeled sensor records from medical devices. However, large-scale foundation models require a lot of training resources, which are difficult to meet in clinical practice. For privacy concerns, patient data cannot be sent to the cloud for analysis. We trained a powerful generator on massive public data. It can be personalized for the large time series foundation model, relying on private data. This will help to quickly identify medical time series data and accurately classify the patient's condition type status.