RPeaksDetectionMetrics

class torch_ecg.components.RPeaksDetectionMetrics(thr: float = 0.075, extra_metrics: Callable | None = None)

Bases: Metrics

Metrics for the task of R peaks detection, as proposed in CPSC2019.

Parameters:

• thr (float, default 0.075) – Threshold for a prediction to be truth positive, with units in seconds,

• extra_metrics (callable, optional) –

  Extra metrics to compute, has to be a function with signature

  def extra_metrics(
      labels : Sequence[Union[Sequence[int], np.ndarray]],
      outputs : Sequence[Union[Sequence[int], np.ndarray]],
      fs : int
  ) -> dict
  

compute(labels: Sequence[ndarray | Sequence[int]], outputs: Sequence[ndarray | Sequence[int]], fs: int, thr: float | None = None) → RPeaksDetectionMetrics

QRS accuracy score, proposed in CPSC2019.

Parameters:

• labels (array_like) – array of ground truths of rpeaks locations (indices) from multiple records.

• outputs (array_like) – predictions of ground truths of rpeaks locations (indices) for multiple records.

• fs (numbers.Real) – Sampling frequency of ECG signal

• thr (float, optional, defaults to self.thr) – Threshold for a prediction to be truth positive, with units in seconds.

Returns:

self – The metrics object itself with the computed metrics.

Examples

>>> labels = [np.array([500, 1000])]
>>> outputs = [np.array([500, 700, 1000])]  # a false positive at 700
>>> metrics = RPeaksDetectionMetrics()
>>> metrics = metrics.compute(labels, outputs, fs=500)
>>> metrics.qrs_score
0.7

Return type:

RPeaksDetectionMetrics

extra_repr_keys() → List[str]

Extra keys for __repr__() and __str__().