26. 心電圖競賽心電圖介紹我是誰 心電圖未來
陳在民 專案處-顧問
原創的MetforNet模型架構
1. Combination of CNN and bidirectional RNN with Attention
layer.
2. 5 CNN-blocks including 2 convolution layers that follow the
convolution-pooling layer
3. Dropout was driven by randomly dropping 20% of the
connections to the next block or layer.
4. In the last CNN-block, we connected it into a bidirectional RNN
with Attention layer, and applied the batch normalization
before it was connected to the fully-connected layer.
5. LeakyReLU activation function was used for each layer, except
for the last fully-connected layer, where Sigmoid activation
function was used.
6. Model was trained with categorical-cross-entropy loss
function and ADAM optimizer.
走在時代尖端的架構
30. 心電圖競賽心電圖介紹我是誰 心電圖未來
陳在民 專案處-顧問
三位心臟科醫師的復驗
PatientID Conf. Model Cardiologist1 Cardiologist2 Cardiologist3 Consensus
6268 STE Normal Normal SR (normal) SR (normal) Normal
6215 STD AF
narrow-QRS tachycardia, SVT,
RBBB, RAD
PSVT PSVT PSVT
4237 RBBB I-AVB I-AVB I-AVB, SR, TWI (V1-V4), rSR’ (V1) I-AVB,SR I-AVB
6380 STE LBBB LBBB LBBB,SR,LAE LBBB,SR LBBB
1452 AF RBBB AF, MVR, RBBB
AF, RBBB, Q wave, STE with reciprocal change,
w/o old MI
AF, RBBB RBBB
5398 PVC PAC PAC PAC PAC,SR PAC
5963 STD PVC PVC PVC,SR,STD(V3-V6) PVC,SR PVC
4278 RBBB STD normal SR, minimal STTC(II, III, aVF) STD-like, SR Normal
504 Normal STE STE-like SR, early repolarization SR Normal