Fig. 5: Acute MI detection algorithm architecture. | npj Digital Medicine

Fig. 5: Acute MI detection algorithm architecture.

From: AI-enhanced reconstruction of the 12-lead electrocardiogram via 3-leads with accurate clinical assessment

Fig. 5

The architecture of the acute MI detection algorithm is organized into three different sections. The first section includes eight residual convolutional neural network (ResCNN) blocks extracting the features of each of the eight input leads. The second section includes a single ResCNN block receiving the aggregated lead feature as input and returning a unique multi-dimensional vector as output. In the last section, the vector obtained is processed through a Feed-Forward Neural Network (FNN) and, finally, a sigmoid function, returning the probability of detecting an acute MI.

Back to article page