# US 11,443,218 B2 · Model guided deep learning approach towards prediction of physical system behavior

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 11,443,218 B2
- Document ID: US11443218B2
- Status: Granted
- Kind: B2
- Application number: 16/226,461
- Filed: 2018-12-19
- Published: 2019-06-20
- Granted: 2022-09-13
- Earliest claimed: 2017-12-20
- Assignees: Arizona Board of Regents on behalf of Arizona State University
- Applicants: Sandeep Gupta; Ayan Banerjee
- Inventors: Sandeep Gupta; Ayan Banerjee
- Web page: https://patentgrep.com/patent/US11443218B2

## CPC

- G06N 3/08: Learning methods
- G06N 7/00: Computing arrangements based on specific mathematical models
- A61B 5/1118: {Determining activity level}
- A61B 5/14532: {for measuring glucose, e.g. by tissue impedance measurement}
- A61B 5/7275: {Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor}
- G06N 3/048: Activation functions
- G06N 3/0499: Feedforward networks
- G06N 3/09: Supervised learning
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G16H 10/60: for patient-specific data, e.g. for electronic patient records
- G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems
- A61B 5/7267: {involving training the classification device}
- A61B 2562/0219: Inertial sensors, e.g. accelerometers, gyroscopes, tilt switches
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- A61B: DIAGNOSIS; SURGERY; IDENTIFICATION (usefulness limited to only animals A61D)
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA

## Abstract

Systems and methods are provided for controlling predictive medical monitoring. A non-linear-predictive-guide model estimates a patient parameter that is used as a guide in a deep neural network for improving accuracy of estimation by the deep neural network. The guide model generates a guiding first estimated patient parameter based on the guide model and patient input data. The deep neural network generates a second estimated patient parameter based on the deep neural network, the patient input data, and the guiding first estimated patient parameter. The deep neural network includes an input layer that receives the guiding first estimated patient parameter, and hidden layers including respective artificial neurons configured to perform a linear or nonlinear transformation on output of at least one artificial neuron from an adjacent layer in the deep neural network. An output layer receives at least one output from a hidden layer.

## Claims

### Claim 1

A control system for a predictive medical monitor, the control system comprising:
an electronic processor; and
a memory communicatively coupled to the electronic processor, the memory storing:
a guide model including a non-linear predictive model for estimating a guiding first patient parameter;
a deep neural network, the deep neural network including:
an input layer configured to receive the guiding first estimated patient parameter;
a plurality of hidden layers, wherein each hidden layer comprises a respective plurality of artificial neurons, each artificial neuron in a hidden layer being configured to perform a linear or nonlinear transformation on output of at least one artificial neuron from an adjacent layer in the deep neural network; and
an output layer configured to receive at least one output from a hidden layer;
the memory further storing instructions that when executed by the electronic processor configure the electronic processor to:
receive patient input data;
input the patient input data into the guide model and generate the guiding first estimated patient parameter based on the guide model and the patient input data; and
input the patient input data and the guiding first estimated patient parameter into the deep neural network input layer and generate the second estimated patient parameter based on the deep neural network, the patient input data, and the guiding first estimated patient parameter.

### Claim 2 (depends on 1)

The control system of claim 1, wherein the guide model is a physiological model.

### Claim 3 (depends on 1)

The control system of claim 1, wherein the guide model is a modified Bergman Minimal Model.

### Claim 4 (depends on 1)

The control system of claim 1, wherein the guiding first estimated patient parameter includes a first estimated future glucose level and the second estimated patient parameter includes a second estimated future glucose level.

### Claim 5 (depends on 1)

The control system of claim 1, wherein the patient input data includes a measured blood glucose level.

### Claim 6 (depends on 1)

The control system of claim 1, wherein the patient input data includes an endogenous glucose production factor.

### Claim 7 (depends on 1)

The control system of claim 1, wherein the patient input data includes one or both of a physical activity factor and a meal intake factor.

### Claim 8 (depends on 1)

The control system of claim 1, wherein the second estimated patient parameter is utilized to determine an input parameter in a closed loop control system.

### Claim 9

A method for controlling a predictive medical monitor, the method comprising:
storing, in a memory, a guide model including a non-linear predictive model for estimating a guiding first patient parameter;
storing, in the memory, a deep neural network, the deep neural network including:
an input layer configured to receive the guiding first estimated patient parameter;
a plurality of hidden layers, wherein each hidden layer comprises a respective plurality of artificial neurons, each artificial neuron in a hidden layer being configured to perform a linear or nonlinear transformation on output of at least one artificial neuron from an adjacent layer in the deep neural network; and
an output layer configured to receive at least one output from a hidden layer; and with an electronic processor coupled to the memory:
receiving patient input data;
inputting the patient input data into the guide model and generating the guiding first estimated patient parameter based on the guide model and the patient input data; and
inputting the patient input data and the guiding first estimated patient parameter into the deep neural network input layer and generating the second estimated patient parameter based on the deep neural network, the patient input data, and the guiding first estimated patient parameter.

### Claim 10 (depends on 9)

The method of claim 9, wherein the guide model is a physiological model.

### Claim 11 (depends on 9)

The method of claim 9, wherein the guide model is a modified Bergman Minimal Model.

### Claim 12 (depends on 9)

The method of claim 9, wherein the guiding first estimated patient parameter includes a first estimated future glucose level and the second estimated patient parameter includes a second estimated future glucose level.

### Claim 13 (depends on 9)

The method of claim 9, wherein the patient input data includes one or more of a measured blood glucose level, an endogenous glucose production factor, a physical activity factor, or a meal intake factor.

### Claim 14 (depends on 9)

The method of claim 9, wherein the second estimated patient parameter is utilized to determine an input parameter in a closed loop control system.

### Claim 15

A transitory computer-readable medium comprising instructions that when executed by an electronic processor, cause the electronic processor to:
store, in a memory, a guide model including a non-linear predictive model for estimating a guiding first patient parameter;
store, in the memory, a deep neural network, the deep neural network including:
an input layer configured to receive the guiding first estimated patient parameter;
a plurality of hidden layers, wherein each hidden layer comprises a respective plurality of artificial neurons, each artificial neuron in a hidden layer being configured to perform a linear or nonlinear transformation on output of at least one artificial neuron from an adjacent layer in the deep neural network; and
an output layer configured to receive at least one output from a hidden layer;
receive patient input data;
input the patient input data into the guide model and generate the guiding first estimated patient parameter based on the guide model and the patient input data; and
input the patient input data and the guiding first estimated patient parameter into the deep neural network input layer and generate the second estimated patient parameter based on the deep neural network, the patient input data, and the guiding first estimated patient parameter.

### Claim 16 (depends on 15)

The transitory computer-readable medium of claim 15, wherein the guide model is a physiological model.

### Claim 17 (depends on 15)

The transitory computer-readable medium of claim 15, wherein the guide model is a modified Bergman Minimal Model.

### Claim 18 (depends on 15)

The transitory computer-readable medium of claim 15, wherein the guiding first estimated patient parameter includes a first estimated future glucose level and the second estimated patient parameter includes a second estimated future glucose level.

### Claim 19 (depends on 15)

The transitory computer-readable medium of claim 15, wherein the patient input data includes one or more of a measured blood glucose level, an endogenous glucose production factor, a physical activity factor, or a meal intake factor.

### Claim 20 (depends on 15)

The transitory computer-readable medium of claim 15, wherein the second estimated patient parameter is utilized to determine an input parameter in a closed loop control system.