Model to detect ECG at high risk of heart failure from ECG presenting LBBB
- tensorflow 2.4.1
- numpy 1.19.2
- pandas 1.3.4
Put the input file in the same directory as LBBB_model_code.py as numpy array. The input should be formated as a 3d numpy array with shape 2500,12,1 (time,induction,1). The ECG should be in 250 Hz recording with voltage unit = mV.
The model weights are not publicly available because it may contain patient information. The web interface to run the full model is available at http://onebraveideaml.org/
The model consists of a layer of 2D convolutional neural network (CNN) layer followed by 20 layers of multi_conv2D module, which consists of 3 different-depth 2D-CNN layers. The first CNN layer has a kernel shape of (7x3) whille all remaining CNN layers have (3x3). The final CNN layer is followed by a global average pooling and a single fully connected layer. The model has 258,754,113 parameters (258,546,625 trainable)