AI for Cardiology: ECG and Cardiac Imaging

Module: Healthcare AI | Difficulty: Advanced

Heart Rate Variability

Ejection Fraction

Cardiac AI Performance

import torch
import torch.nn as nn

class ECGClassifier(nn.Module):
    def __init__(self, num_classes=5, in_channels=12):
        super().__init__()
        self.conv1 = nn.Conv1d(in_channels, 64, kernel_size=50, stride=5)
        self.conv2 = nn.Conv1d(64, 128, kernel_size=20, stride=2)
        self.conv3 = nn.Conv1d(128, 256, kernel_size=10, stride=2)
        self.bn1 = nn.BatchNorm1d(64)
        self.bn2 = nn.BatchNorm1d(128)
        self.bn3 = nn.BatchNorm1d(256)
        self.attention = nn.MultiheadAttention(256, num_heads=8, batch_first=True)
        self.classifier = nn.Linear(256, num_classes)

    def forward(self, x):
        x = torch.relu(self.bn1(self.conv1(x)))
        x = torch.relu(self.bn2(self.conv2(x)))
        x = torch.relu(self.bn3(self.conv3(x)))
        x = x.permute(0, 2, 1)
        attn_out, _ = self.attention(x, x, x)
        x = attn_out[:, -1, :]
        return self.classifier(x)

model = ECGClassifier(num_classes=5)
ecg = torch.randn(1, 12, 5000)
output = model(ecg)
print(f'ECG classification output: {output.shape}')

rr_intervals = np.random.uniform(0.6, 1.2, 100)
rmssd = np.sqrt(np.mean(np.diff(rr_intervals) ** 2))
print(f'RMSSD: {rmssd:.4f}')

Research Insight: ECG AI models can detect conditions beyond their training labels: models trained for arrhythmia detection can identify atr fibrillation, predict future AF episodes, and even estimate left ventricular ejection fraction from standard 12-lead ECGs. This suggests ECG signals contain far more cardiac information than traditionally extracted.