SCALABLE ARTIFICIAL INTELLIGENCE MODEL GENERATION SYSTEMS AND METHODS FOR HEALTHCARE

• US 20200134446A1
• Filed: 10/31/2018
• Published: 04/30/2020
• Est. Priority Date: 10/31/2018
• Status: Active Grant

• Current Assignee: General Electric Company
• Original Assignee: General Electric Company
• Inventors: Soni, Ravi, Zhang, Min, Avinash, Gopal
• Granted Patent: US 11,507,822 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 11/3447 Performance evaluation by modeling
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 3/047 Probabilistic or stochastic networks
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 3/084 Backpropagation, e.g. using gradient descent
• CPC Class Codes: G16H 30/20 for handling medical images, e.g. DICOM, HL7 or PACS
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems

Abstract

Systems and methods to generate artificial intelligence models with synthetic data are disclosed. An example system includes a deep neural network (DNN) generator to generate a first DNN model using first real data. The example system includes a synthetic data generator to generate first synthetic data from the first real data, the first synthetic data to be used by the DNN generator to generate a second DNN model. The example system includes an evaluator to evaluate performance of the first and second DNN models to determine whether to generate second synthetic data. The example system includes a synthetic data aggregator to aggregate third synthetic data and fourth synthetic data from a plurality of sites to form a synthetic data set. The example system includes an artificial intelligence model deployment processor to deploy an artificial intelligence model trained and tested using the synthetic data set.

Claims

1. An artificial intelligence model generation system comprising:

a deep neural network (DNN) generator to generate a first DNN model using first real data;

a synthetic data generator to generate first synthetic data from the first real data, the first synthetic data to be used by the DNN generator to generate a second DNN model;

an evaluator to evaluate performance of the first DNN model and the second DNN model based on output of the first DNN model and the second DNN model, the evaluator to determine whether to generate second synthetic data based on a comparison of a first output of the first DNN model and a second output of the second DNN model, the synthetic data generator to generate third synthetic data from a first site when the comparison indicates that performance of the first DNN model and performance of the second DNN model are aligned;

a synthetic data aggregator to aggregate the third synthetic data from the first site and fourth synthetic data from a second site to form a synthetic data set; and

an artificial intelligence model generator to generate and deploy an artificial intelligence model trained and tested using the synthetic data set.

2. The system of claim 1, further including a generative adversarial network (GAN) generator to generate a GAN model to be used by the synthetic data generator to generate the third synthetic data.

3. The system of claim 1, wherein, when the evaluator determines to generate second synthetic data, the synthetic data generator is to generate the second synthetic data using second real data obtained from a real data store.

4. The system of claim 1, wherein the synthetic data aggregator is to aggregate the third synthetic data and the fourth synthetic data across a plurality of sites to form the synthetic data set to be deployed to each of the plurality of sites.

5. The system of claim 4, wherein the artificial intelligence model generator is to generate and deploy the artificial intelligence model to each of the plurality of sites.

6. The system of claim 1, wherein the DNN generator is to generate the first DNN model according to a clinical purpose.

7. The system of claim 1, wherein the evaluator is to compare the first output of the first DNN model and the second output of the second DNN model by evaluating an accuracy of the second output compared to the first output.

8. The system of claim 1, wherein the first synthetic data resembles the first real data in content and format but does not include any portion of the first real data.

9. A computer-readable storage medium comprising instructions which, when executed, cause at least one processor to implement at least:

a deep neural network (DNN) generator to generate a first DNN model using first real data;

a synthetic data generator to generate first synthetic data from the first real data, the first synthetic data to be used by the DNN generator to generate a second DNN model;

an evaluator to evaluate performance of the first DNN model and the second DNN model based on output of the first DNN model and the second DNN model, the evaluator to determine whether to generate second synthetic data based on a comparison of a first output of the first DNN model and a second output of the second DNN model, the synthetic data generator to generate third synthetic data from a first site when the comparison indicates that performance of the first DNN model and performance of the second DNN model are aligned;

a synthetic data aggregator to aggregate the third synthetic data from the first site and fourth synthetic data from a second site to form a synthetic data set; and

an artificial intelligence model generator to generate and deploy an artificial intelligence model trained and tested using the synthetic data set.

10. The computer-readable storage medium of claim 9, wherein the instructions, when executed, further cause the at least one processor to implement a generative adversarial network (GAN) generator to generate a GAN model to be used by the synthetic data generator to generate the first synthetic data.

11. The computer-readable storage medium of claim 9, wherein, when the evaluator determines to generate second synthetic data, the synthetic data generator is to generate the second synthetic data using second real data obtained from a real data store.

12. The computer-readable storage medium of claim 9, wherein the synthetic data aggregator is to aggregate the third synthetic data and the fourth synthetic data across a plurality of sites to form the synthetic data set to be deployed to each of the plurality of sites.

13. The computer-readable storage medium of claim 12, wherein the artificial intelligence model generator is to generate and deploy the artificial intelligence model to each of the plurality of sites.

14. The computer-readable storage medium of claim 9, wherein the DNN generator is to generate the first DNN model according to a clinical purpose.

15. The computer-readable storage medium of claim 9, wherein the evaluator is to compare the first output of the first DNN model and the second output of the second DNN model by evaluating an accuracy of the second output compared to the first output.

16. The computer-readable storage medium of claim 9, wherein the first synthetic data resembles the first real data in content and format but does not include any portion of the first real data.

17. A method comprising:

generating, using at least one processor, a first DNN model using first real data;

generating, using the at least one processor, a first synthetic data from the first real data;

generating, using the at least one processor, a second DNN model using the first synthetic data;

evaluating, using the at least one processor, performance of the first DNN model and the second DNN model based on output of the first DNN model and the second DNN model, the evaluating to determine whether to generate second synthetic data based on a comparison of a first output of the first DNN model and a second output of the second DNN model;

generating, using the at least one processor when the comparison indicates that performance of the first DNN model and performance of the second DNN model align, third synthetic data from a first site;

aggregating, using the at least one processor, the third synthetic data from the first site and fourth synthetic data from a second site to form a synthetic data set; and

deploying, using the at least one processor, an artificial intelligence model trained and tested using the synthetic data set.

18. The method of claim 17, further including generating a generative adversarial network (GAN) model to be used to generate the first synthetic data.

19. The method of claim 17, further including generating the second synthetic data using second real data obtained from a real data store.

20. The method of claim 17, wherein aggregating the first synthetic data and the second synthetic data includes aggregating synthetic data across a plurality of sites to form the synthetic data set and deploying the synthetic data set to each of the plurality of sites.