Medical machine synthetic data and corresponding event generation

1. A synthetic time series data generation apparatus comprising:

memory storing instructions; and

at least one processor to execute the instructions to at least:

a) generate a synthetic data set including i) multi-channel one-dimensional time-series data and ii) associated annotation using a first artificial intelligence network model, the multi-channel one-dimensional time-series data including synthetic waveform signal data, the synthetic data set generated and having a first classification;

b) analyze the synthetic data set with respect to a real data set using a second artificial intelligence network model to classify the synthetic data set, the real data set collected from at least one patient data source and having a second classification, the second classification different from the first classification at least in that the first classification indicates synthetic data generation by a model and the second classification indicates data captured fora patient;

c) when the second artificial intelligence network model classifies the synthetic data set as having the first classification, adjust the first artificial intelligence network model, rather than the second artificial intelligence network model, using feedback from the second artificial intelligence network model to tune the first artificial intelligence network model and repeat a)-b), the feedback including the first classification of the synthetic data set by the second artificial intelligence network model, the feedback including a loss function computed between the real data set and the synthetic data set and an error gradient associated with the loss function to adjust at least one parameter of the first artificial intelligence network model; and

d) when the second artificial intelligence network model classifies the synthetic data set as having the second classification, output the synthetic data set for use in association with the second classification,

wherein use of the synthetic data set includes to impute synthetic waveform signal data as missing data to complete a captured waveform.

2. The apparatus of claim 1 , wherein the first artificial intelligence network model includes a generator of a generative adversarial network model, and wherein the second artificial intelligence network model includes a discriminator of the generative adversarial network model.

3. The apparatus of claim 1 , wherein the annotation includes an event associated with the time-series data.

4. The apparatus of claim 3 , wherein the event includes at least one of a first event external to a signal represented by the time-series data ora second event internal to the signal represented by the time-series data.

5. The apparatus of claim 1 , wherein the processor is to generate the synthetic data set using one or more latent input vectors for a plurality of times.

6. The apparatus of claim 1 , wherein the synthetic data set is to be provided to a model generator to at least one of train or test a third artificial intelligence network model.

7. The apparatus of claim 1 , wherein the synthetic data set is classified by the at least one processor using a decision aggregator.

8. At least one tangible computer-readable storage medium comprising instructions that, when executed, cause at least one processor to at least:

a) generate a synthetic data set including i) multi-channel one-dimensional time-series data and ii) associated annotation using a first artificial intelligence network model, the multi-channel one-dimensional time-series data including synthetic waveform signal data, the synthetic data set generated and having a first classification;

b) analyze the synthetic data set with respect to a real data set using a second artificial intelligence network model to classify the synthetic data set, the real data set collected from at least one patient data source and having a second classification, the second classification different from the first classification at least in that the first classification indicates synthetic data generation by a model and the second classification indicates data captured fora patient;

c) when the second artificial intelligence network model classifies the synthetic data set as having the first classification, adjust the first artificial intelligence network model, rather than the second artificial intelligence network model, using feedback from the second artificial intelligence network model to tune the first artificial intelligence network model and repeat a)-b), the feedback including the first classification of the synthetic data set by the second artificial intelligence network model, the feedback including a loss function computed between the real data set and the synthetic data set and an error gradient associated with the loss function to adjust at least one parameter of the first artificial intelligence network model; and

d) when the second artificial intelligence network model classifies the synthetic data set as having the second classification, output the synthetic data set for use in association with the second classification,

wherein use of the synthetic data set includes to impute synthetic waveform signal data as missing data to complete a captured waveform.

9. The at least one tangible computer-readable storage medium of claim 8 , wherein the first artificial intelligence network model includes a generator of a generative adversarial network model, and wherein the second artificial intelligence network model includes a discriminator of the generative adversarial network model.

10. The at least one tangible computer-readable storage medium of claim 8 , wherein the annotation includes an event associated with the time-series data.

11. The at least one tangible computer-readable storage medium of claim 10 , wherein the event includes at least one of a first event external to a signal represented by the time-series data or a second event internal to the signal represented by the time-series data.

12. The at least one tangible computer-readable storage medium of claim 8 , wherein the instructions, when executed, cause the processor to generate the synthetic data set using one or more latent input vectors for a plurality of times.

13. The at least one tangible computer-readable storage medium of claim 8 , wherein the instructions, when executed, cause the processor to provide the synthetic data set to a model generator to at least one of train or test a third artificial intelligence network model.

14. A computer-implemented method to generate synthetic time series data and associated annotation, the method comprising:

a) generating, using at least one processor, a synthetic data set including i) multi-channel one-dimensional time-series data and ii) associated annotation using a first artificial intelligence network model, the multi-channel one-dimensional time-series data including synthetic waveform signal data, the synthetic data set generated and having a first classification;

b) analyzing, using the at least one processor, the synthetic data set with respect to a real data set using a second artificial intelligence network model to classify the synthetic data set, the real data set collected from at least one patient data source and having a second classification, the second classification different from the first classification at least in that the first classification indicates synthetic data generation by a model and the second classification indicates data capture fora patient;

c) when the second artificial intelligence network model classifies the synthetic data set as having the first classification, adjusting, using the at least one processor, the first artificial intelligence network model, rather than the second artificial intelligence network model, using feedback from the second artificial intelligence network model to tune the first artificial intelligence network model and repeat a)-b), the feedback including the first classification of the synthetic data set by the second artificial intelligence network model, the feedback including a loss function computed between the real data set and the synthetic data set and an error gradient associated with the loss function to adjust at least one parameter of the first artificial intelligence network model; and

d) when the second artificial intelligence network model classifies the synthetic data set as having the second classification, outputting, using the at least one processor, the synthetic data set for use in association with the second classification,

wherein use of the synthetic data set includes to impute synthetic waveform signal data as missing data to complete a captured waveform.