PulseDB: A large, cleaned dataset based on MIMIC-III and VitalDB for benchmarking cuff-less blood pressure estimation methods

Summary

PulseDB, from the Rutgers University group (Laleh Najafizadeh), is the largest cleaned and standardised PPG-ECG-blood pressure dataset available, designed specifically to enable fair benchmarking of cuffless BP estimation models. The dataset contains 5,245,454 10-second segments of simultaneous ECG, PPG, and arterial blood pressure (ABP) waveforms from 5,361 subjects (MIMIC-III + VitalDB sources), totalling 14,570 hours. A critical contribution is the standardised train/validation/test split following BHS and AAMI clinical protocols, which addresses the pervasive problem of incompatible splits across published papers that makes fair comparison impossible.