Estimation of Continuous Blood Pressure from PPG via a Federated Learning Approach

. 2021 Sep 21;21(18):6311. doi: 10.3390/s21186311

Algorithm 1 FedT2TGAN Training Procedure.

Input: Training sample pairs of ABP and PPG $S_{n}$ = ( $X_{A B P 1, P P G 1} + \dots + X_{A B P n, P P G n}$ )
Output: GAN Model T2TGAN
Initialise global-model
Synchronise client-models with global-model
for num clients do
# Communication Round
Select 6 random client-models to train
for each client-model do
# client-model training
for num epochs do
# Calculate identity, cycle and adversarial losses
L = GetGANLosses( $X_{A B P, P P G}$ , ${\hat{X}}_{\hat{A B P}, \hat{P P G}}$ )
# Update weights of client-model
W = UpdateClientWeights(L)
end for
Aggregate client-model’s weights with global-model
end for
end for
Generate ABP waveforms from unseen PPG using trained global-model