LR-HSMM

Logistic Regression - Hidden Semi-Markov Model segmenter

Based on: D. B. Springer, L. Tarassenko and G. D. Clifford, “Logistic Regression-HSMM-Based Heart Sound Segmentation,” in IEEE Transactions on Biomedical Engineering, vol. 63, no. 4, pp. 822-832, April 2016, doi: 10.1109/TBME.2015.2475278.

class pyPCG.lr_hsmm.LR_HSMM

Main segmenter object

signal_fs

Sampling frequency of the input signal. Default: 1000 [Hz]

Type:

int

feature_fs

Sampling frequency for feature calculations. Default: 50 [Hz]

Type:

int

mean_s1_len

Average S1 duration. Default: 122 [ms]

Type:

float

mean_s2_len

Average S2 duration. Default: 99 [ms]

Type:

float

std_s1_len

Standard deviation of S1 durations. Default: 22 [ms]

Type:

float

std_s2_len

Standard deviation of S2 durations. Default: 22 [ms]

Type:

float

bandpass_frq

Cutoff frequencies for pre-processing band-pass filtering (Butterworth, 4th order). Default: (25,400) [Hz]

Type:

tuple[float,float]

expected_hr_range

Minimum and maximum expected heartrates. Default: (30,120) [bpm]

Type:

tuple[float,float]

hsmm_model

State predictor model

Type:

hsmmlearn.hsmm.HSMMModel

lr_model

Probability emissions for the HSMM states. Includes a LogisticRegression model for each state

Type:

hsmmlearn.emissions.AbstractEmissions

load_model(filename: str) → None

Loads the model parameters from a json file

Parameters:

filename (str) – Name of file containing model parameters

save_model(filename: str) → None

Saves the model parameters to a json file

Parameters:

filename (str) – Name of file to be saved

segment_single(sig: ndarray[tuple[int, ...], dtype[float64]], recalc_timing: bool = False) → tuple[ndarray[tuple[int, ...], dtype[float64]], ndarray[tuple[int, ...], dtype[float64]]]

Predicts the states for the given PCG signal

Parameters:

sig (np.ndarray) – Input signal to be segmented

Returns:

Predicted state for each sample and the calculated features

Return type:

tuple[np.ndarray,np.ndarray]

train_model(train_data: ndarray[tuple[int, ...], dtype[float64]] | list[float], train_s1_annot: ndarray[tuple[int, ...], dtype[float64]] | list[float], train_s2_annot: ndarray[tuple[int, ...], dtype[float64]] | list[float], multiprocess: int | None = None) → None

Trains the model on the specified data with S1 and S2 location annotations

Parameters:

• train_data (np.ndarray) – Array of input signals

• train_s1_annot (np.ndarray) – Array of S1 annotations for each signal

• train_s2_annot (np.ndarray) – Array of S2 annotations for each signal