BUET (Ibtehaz & Rahman)
Translates a raw PPG waveform into a full continuous arterial blood-pressure waveform, from which systolic, diastolic, and mean arterial pressure are derived. Uses a two-stage cascaded 1D convolutional network in a U-Net style, with a coarse approximation stage followed by a refinement stage. Meets BHS Grade A and AAMI accuracy standards for diastolic and mean arterial pressure. Developed at BUET.
Architecture
CNN (1D)
Two-stage cascaded fully-convolutional 1D network (ApproximateNetwork + RefinementNetwork, U-Net-style)
Framework
Keras
Added to catalog
2026-07-10
MIT
License for model weights only. Associated code may be licensed seperately, check code source for specific terms.
MIMIC-II Waveform Database open_in_new
ICU waveform database containing synchronized PPG and invasive arterial blood pressure (ABP) recordings, used to train PPG2ABP.
ABP waveform
Diastolic blood pressure (DBP)
Mean arterial pressure (MAP)
Systolic blood pressure (SBP)