QualityFM: a Multimodal Physiological Signal Foundation Model with Self-Distillation for Signal Quality Challenges in Critically Ill Patients
This addresses signal quality challenges in ICU and OR settings for critically ill patients, representing a novel method for a known bottleneck with strong specific gains.
The paper tackled the problem of poor signal quality in photoplethysmogram (PPG) and electrocardiogram (ECG) signals in critically ill patients, which can cause false alarms or diagnostic errors, by introducing QualityFM, a multimodal foundation model pre-trained on over 21 million waveforms and 179,757 hours of data, achieving efficacy in transfer learning on clinical tasks like ventricular tachycardia detection, atrial fibrillation identification, and arterial blood pressure estimation.
Photoplethysmogram (PPG) and electrocardiogram (ECG) are commonly recorded in intesive care unit (ICU) and operating room (OR). However, the high incidence of poor, incomplete, and inconsistent signal quality, can lead to false alarms or diagnostic inaccuracies. The methods explored so far suffer from limited generalizability, reliance on extensive labeled data, and poor cross-task transferability. To overcome these challenges, we introduce QualityFM, a novel multimodal foundation model for these physiological signals, designed to acquire a general-purpose understanding of signal quality. Our model is pre-trained on an large-scale dataset comprising over 21 million 30-second waveforms and 179,757 hours of data. Our approach involves a dual-track architecture that processes paired physiological signals of differing quality, leveraging a self-distillation strategy where an encoder for high-quality signals is used to guide the training of an encoder for low-quality signals. To efficiently handle long sequential signals and capture essential local quasi-periodic patterns, we integrate a windowed sparse attention mechanism within our Transformer-based model. Furthermore, a composite loss function, which combines direct distillation loss on encoder outputs with indirect reconstruction loss based on power and phase spectra, ensures the preservation of frequency-domain characteristics of the signals. We pre-train three models with varying parameter counts (9.6 M to 319 M) and demonstrate their efficacy and practical value through transfer learning on three distinct clinical tasks: false alarm of ventricular tachycardia detection, the identification of atrial fibrillation and the estimation of arterial blood pressure (ABP) from PPG and ECG signals.