Catheter location determination in paediatric patients
When inserting a catheter or other medical equipment into a child or adolescent or other paediatric patient, ECG signals may be recorded from the catheter and the location of the catheter determined by analysing the ECG signals. A signal processor and user interface may receive recorded signals in real-time from the catheter while the catheter is inserted into the paediatric patient. The signal processor may analyse the ECG signals to determine the location of the catheter in the paediatric patient. The user interface may display the location of the catheter and other pertinent information to a user while the user is inserting the catheter. One method for determining the location may include determining R-wave and P-wave peaks of the ECG signal and determining the location from an average location of the R-wave and P-wave peaks in the ECG signal.
1. A machine learning engine, comprising:
a signal quality determination module trained to determine signal qualities of intravascular electrocardiogram (ECG) signals from a tip of a catheter inserted in a patient as clean, no signal, and noisy signal,
the determination based on a frequency response of the intravascular ECG signals and P/R ratios determined from feature extraction from the intravascular ECG signals, and
the signal quality determination module being previously trained using a set of pre-recorded intravascular ECG tracings labelled with signal qualities comprising clean, no signal, and noisy signal; and
a location determination module trained to determine a relative location of the tip of the catheter in the patient based on modelled associations of locations with characteristics of P-wave and R-wave peaks extracted from the intravascular ECG signals,
the characteristics of P-wave and R-wave peaks comprising one or more of:
a P/R ratio of the averaged P-wave and R-wave peaks, and
normalised averaged P-wave and R-wave peaks, each normalization obtained by a ratio between an averaged surface ECG P-wave peak and an averaged intravascular P-wave peak or an averaged surface ECG R-wave peak and an averaged intravascular R-wave peak, respectively, and the surface ECG peaks being captured prior to receiving the intravascular ECG signals, and
the location determination module being previously trained using a set of pre-recorded intravascular ECG tracings obtained from catheters and labelled with the known locations of the catheters in patients' bodies,
wherein the signal quality determination module is configured to pass the intravascular ECG signals to the location determination module if they are determined to have a clean signal quality.
2. The machine learning engine of claim 1 , wherein the signal quality determination module is further configured to generate an alert if the intravascular ECG recordings are determined to have a noisy or empty signal quality.
3. The machine learning engine of claim 1 , wherein the machine learning engine uses at least one of an artificial neural network algorithm, a deep learning algorithm, a Bayesian network algorithm, a decision tree learning algorithm, and a rule-based learning algorithm.
4. A machine learning method, comprising:
receiving intravascular electrocardiogram (ECG) signals from a tip of a catheter inserted in a patient;
inputting the intravascular ECG signals to a machine learning engine comprising a signal quality determination module and a location determination module,
wherein the signal quality determination module is trained to determine signal qualities of the intravascular ECG recordings as clean, no signal, and noisy signal,
the determination based on a frequency response of the intravascular ECG signals and P/R ratios determined from feature extraction from the intravascular ECG signals,
the signal quality determination module being previously trained using a set of pre-recorded intravascular ECG tracings labelled with signal qualities comprising clean, no signal, and noisy signal,
wherein the location determination module is trained to determine a relative location of the tip of the catheter in the patient based on modelled associations of locations with characteristics of P-wave and R-wave peaks extracted from the intravascular ECG signals,
the characteristics of P-wave and R-wave peaks comprising one or more of:
a P/R ratio of the averaged P-wave and R-wave peaks, and
normalised averaged P-wave and R-wave peaks, each normalization obtained by a ratio between an averaged surface ECG P-wave peak and an averaged intravascular P-wave peak or an averaged surface ECG R-wave peak and an averaged intravascular R-wave peak, respectively, and the surface ECG peaks being captured prior to receiving the intravascular ECG signals, and
the location determination module being previously trained using a set of pre-recorded intravascular ECG tracings obtained from catheters and labelled with the known locations of the catheters in patients' bodies; and
passing the intravascular ECG signals determined to have a clean signal quality from the signal quality determination module to the location determination module.
5. The machine learning method of claim 4 , wherein the signal quality determination module is further configured to generate an alert if the intravascular ECG recordings are determined to have a noisy or empty signal quality.
6. The machine learning method of claim 4 , wherein the machine learning engine uses at least one of an artificial neural network algorithm, a deep learning algorithm, a Bayesian network algorithm, a decision tree learning algorithm, and a rule-based learning algorithm.