Method and apparatus for producing thermodilution cardiac output measurements utilizing a neural network

• US 5,579,778 A
• Filed: 08/23/1994
• Issued: 12/03/1996
• Est. Priority Date: 09/09/1993
• Status: Expired due to Term

• Current Assignee: University of Utah Research Foundation (State of Utah)
• Original Assignee: University of Utah Research Foundation (State of Utah)
• Inventors: Westenskow, Dwayne R., Johnson, Royce W., Orr, Joseph, Baker, Phillip D.
• Primary Examiner(s): Sykes, Angela D.
• Assistant Examiner(s): NASSER, ROBERT L
• Application Number:
  US08/294,938
• Time in Patent Office: 833 Days
• Field of Search: 128/668, 128/713, 128/692, 128/691
• US Class Current: 600/504
• CPC Class Codes: A61B 5/028 by thermo-dilution
• CPC Class Codes: A61B 5/7267 involving training the classification device
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients
• CPC Class Codes: Y10S 128/925 Neural network

Abstract

A method and device for indirect, quantitative estimation of cardiac output utilizing invasive, indirect techniques. The method of practice includes (i) generating a sequence of signals which are quantitatively dependent upon cardiac output, (ii) transmitting and processing the signals within a computer system and associated neural network capable of generating a single output signal for the combined input signals, (iii) directly determining an actual value for the parameter concurrent with the invasive generation of signals of the previous steps, (iv) applying weighting factors within the neural network at interconnecting nodes to force the output signal of the neural network to match the known value of the parameter as determined invasively, (v) recording the input signals, weighting factors and known value as training data within memory of the computer, and (vi) repeating the previous steps to develop sufficient training data to enable the neural network to accurately estimate parameter value upon future receipt of on-line input signals. Procedures are also described for preclassification of signals and artifact rejection. Following training of the neural network, further direct measurement is unnecessary and the system is ready for diagnostic application and invasive estimation of parameter values.

Claims

1. A method for on-line calculation of cardiac output of a patient, said method comprising the steps of:

1.1) coupling at least one sensor to the patient, said sensor being responsive to register changes in blood temperature as part of a thermodilution procedure which includes bolus injection without volume loading of the patient;

1.2) activating the sensor to generate a sequence of on-line signals which register changes occurring in the blood temperature of a patient through direct indicator dilution signal measurements;

1.3) transmitting the on-line signals as input signals to a computer system, including input nodes of a neural network supported by the computer system, which neural network is capable of calculating a continuous output signal corresponding to a parameter value from the on-line, input signals;

1.4) preprocessing the input signals to produce a scaled waveform;

1.5) processing the input signals within the neural network to convert the sequence of input signals to an on-line output signal corresponding to a cardiac output value by applying fixed weighting factors to the input signals; and

1.6) retrieving said fixed weighting factors which were previously generated by applying a training algorithm with respect to previously collected training data comprising neural network input signals and corresponding known cardiac output values.

2. A method as defined in claim 1, further comprising the steps of:

2.1) selecting a plurality of sample signals from the sequence of signals for processing through a neural network which has been trained to associate such sample signals with a related value for the cardiac output; and

2.2) identifying at least one feature within the sample signals which can be processed through the neural network as a feature signal.

3. A method as defined in claim 2, wherein step 2.2 includes the step of identifying a blood temperature measurement as the feature which defines the feature signal, said method further including the step of developing a waveform based on the sampled signals generated in step 1.3, said waveform being represented by the pulmonary artery blood temperature measurements graphed over a time period comprising a single diagnostic measurement procedure.

4. A method as defined in claim 2 wherein step 2.2 comprises the more specific step of selecting the average of a plurality of blood temperature measurements as a single measurement procedure and processing the temperature signal of these sample signals to estimate the cardiac output without processing all pulmonary artery blood temperature measurements being generated.

5. A method as defined in claim 2, wherein step 2.2 includes the more specific steps as follows:

7. A method as defined in claim 5, wherein step 5.1 comprises the more specific step of selecting the average of approximately 25 blood temperature measurements as a single measurement procedure and processing the temperature signal of these sample signals to apply on-line signals at the cardiac output neural network to estimate the cardiac output without processing all blood temperature measurements being generated.

8. A method as defined in claim 2 further comprising the more specific steps of:

7.1) generating a sequence of pulmonary artery blood temperature signals from a temperature sensing means within the patient'"'"'s pulmonary artery;

7.2) identifying as a feature within the sampled pulmonary artery blood temperature signals to be a temperature value measured once every second over a time defined as the period of the waveform;

7.3) measuring and recording the pulmonary artery blood temperature values within the temperature sensing means to correspond to the signals of step 7.2);

7.4) transmitting the pulmonary artery blood temperature values to the computer system and input nodes of the neural network supported by the computer system for identification of the estimated cardiac output value based on comparison of on-line sample signals with a data base of training signals stored within computer memory; and

7.5) processing the on-line sample signals within the neural network to identify the estimated value of cardiac output associated with the sample signals.

9. A method as defined in claim 7, comprising the more specific step of measuring and recording pulmonary artery blood temperature values at predetermined time intervals over temperature ranges from approximately 23 to 43 degrees celsius.

10. A method as defined in claim 7, including the more specific step of selecting less than all generated signals of step 7.1 for transmittal to the input nodes of the neural network.

11. A method as defined in claim 9, comprising the more specific step of selecting a representative sampling of the generated temperature signals, said temperature signals to consist of approximately 32 points defining a single thermodilution curve waveform, where each point is the average of approximately 25 measurements, where the approximately 25 measurements occur within approximately a one second interval, thereby estimating the waveform developed by the signals transmitted to the input nodes of the neural network without requiring processing of all signals through all interactive nodes of the neural network to generate the desired output signal corresponding to the cardiac output.

14. A method as defined in claim 11, further comprising the steps of processing the output of the at least one hidden layer node from step 11.5g) by repeating steps 11.5c) through 11.5g) until the at least one node output defined in step 11.5f) representing the estimated physiological parameter has been displayed.

15. A method as defined in claim 11, wherein the step of 11.5b) retrieves previously determined weighting factors, comprising the more specific step of generating said previously determined weighting factors by:

13.4) determining an error between the known cardiac output value and the value generated by the neural network, and adjusting the weighting factors to modify the value of the output signal to match the known cardiac output value.

6. 1) developing a waveform from the diagnostic measurement procedures comprising a predetermined number of sample signals, which number corresponds approximately to the number of input nodes existing in the cardiac output neural network;

5.2) storing in memory the sample signals; and

5.3) transmitting the stored sample signals of the waveform to respective input nodes of the cardiac output neural network.

12. A method for on-line calculation of a cardiac output of a patient, said method comprising the steps of:

11.1) coupling at least one sensor to the patient, said sensor being responsive to register changes in blood temperature with the patient as part of a thermodilution procedure which includes bolus injection without volume loading of the patient;

11.2) activating the sensor to generate a sequence of on-line signals which register changes in the blood temperature through direct indicator dilution signal measurements;

11.3) transmitting the on-line signals as input signals to a computer system, including input nodes of a neural network supported by the computer system, which neural network is capable of calculating a continuous output signal corresponding to the cardiac output from the on-line, input signals;

11.4) preprocessing the input signals to produce a scaled waveform;

11.5) processing the input signals within the neural network to convert the sequence of input signals to an on-line output signal corresponding to cardiac output value in accordance with the following substeps;

11.5a) processing the input signals within the neural network through at least one neural network layer having at least one node by applying fixed weighting factors to the input signals;

11.5b) retrieving said fixed weighting factors which were previously determined by applying a training algorithm with respect to previously collected training data comprising neural network input signals and corresponding known cardiac output values to generate said fixed weighting factors;

13. 5c) for each input signal of each node within the neural network layer, calculating a product of the input signal and fixed weighting factor corresponding to each input signal and node combination;

11.5d) for each node within the neural network layer, summing the products of each input signal and fixed weighting factor combination calculated in the previous step 11.5c);

11.5e) for each node within the neural network layer, calculating a node output by applying an input/output function to the sum calculated in the previous step 11.5d);

11.5f) where the output of each node calculated in step 11.5e) represents the neural network output, displaying at least one node output as an estimated cardiac output value, or

11.5g) where the output of each node calculated in step 11.5e) represents the output of at least one hidden layer node, passing at least one output from outputs calculated in 11.5e) as input to any subsequent layer of nodes in the neural network.