# US 2024/0005154 A1 · Superclass-Conditional Gaussian Mixture Model for Personalized Prediction on Dialysis Events

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

- Number: US 2024/0005154 A1
- Document ID: US20240005154A1
- Status: Published application
- Kind: A1
- Application number: 18/370,092
- Filed: 2023-09-19
- Published: 2024-01-04
- Earliest claimed: 2021-09-23
- Assignees: NEC Laboratories America, Inc.
- Applicants: NEC Laboratories America, Inc.
- Inventors: Jingchao Ni; Wei Cheng; Haifeng Chen
- Web page: https://patentgrep.com/patent/US20240005154A1

## CPC

- G06N 3/08: Learning methods
- G06N 3/044: Recurrent networks, e.g. Hopfield networks
- G06N 3/045: Combinations of networks
- G06N 3/084: Backpropagation, e.g. using gradient descent
- G06N 7/01: Probabilistic graphical models, e.g. probabilistic networks
- G16H 20/10: relating to drugs or medications, e.g. for ensuring correct administration to patients
- G16H 50/70: for mining of medical data, e.g. analysing previous cases of other patients
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA

## Abstract

A computer-implemented method for model building is provided. The method includes receiving a training set of medical records and model hyperparameters. The method further includes initializing an encoder as a Dual-Channel Combiner Network (DCNN) and initialize distribution related parameters. The method also includes performing, by a hardware processor, a forward computation to (1) the DCNN to obtain the embeddings of the medical records, and (2) the distribution related parameters to obtain class probabilities. The method additionally includes checking by a convergence evaluator if the iterative optimization has converged. The method further includes performing model personalization responsive to model convergence by encoding the support data of a new patient and using the embeddings and event subtype labels to train a personalized classifier.

## Claims

### Claim 1

A computer-implemented method for model building, comprising:
receiving a training set of dialysis records and model hyperparameters;
initializing an encoder distribution related parameters;
performing, by a hardware processor, a forward computation to (1) the encoder to obtain the embeddings of the dialysis records, and (2) the distribution related parameters to obtain class probabilities;
checking by a convergence evaluator if the iterative optimization has converged; and
performing model personalization responsive to model convergence by encoding the support data of a new dialysis patient and using the embeddings and event subtype labels including different unstable patterns of blood pressure to train a personalized classifier.

### Claim 2 (depends on 1)

The computer-implemented method of claim 1, wherein performing a forward computation comprises computing a loss function with the embeddings and the class probabilities for optimization.

### Claim 3 (depends on 1)

The computer-implemented method of claim 1, wherein performing a forward computation comprises performing alternate optimization for a predefined number of iterations between (1) a step for obtaining posterior probabilities, and (2) a step for obtaining updated model parameters.

### Claim 4 (depends on 1)

The computer-implemented method of claim 1, further comprising performing model testing by encoding the test data of the new dialysis patient and using the embeddings and the personalized classifier to predict event subtypes.

### Claim 5 (depends on 4)

The computer-implemented method of claim 4, further comprising performing a dialysis event on a dialysis patient responsive to the predicted event subtypes.

### Claim 6 (depends on 1)

The computer-implemented method of claim 1, wherein the encoder includes a static channel for encoding static patient profiles, and a temporal channel to encode temporal patient status features, wherein outputs of the static channel and the temporal channel are concatenated and projected to a compact embedding used for prediction by a combination layer of the encoder.

### Claim 7 (depends on 6)

The computer-implemented method of claim 6, wherein the static channel comprises a multilayer perceptron (MLP), and wherein the temporal channel comprises one or more Long Short-Term Memories (LSTMs).

### Claim 8

A computer program product for model building, the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform a method comprising:
receiving a training set of dialysis records and model hyperparameters;
initializing an encoder distribution related parameters;
performing, by a hardware processor, a forward computation to (1) the encoder to obtain the embeddings of the dialysis records, and (2) the distribution related parameters to obtain class probabilities;
checking by a convergence evaluator if the iterative optimization has converged; and
performing model personalization responsive to model convergence by encoding the support data of a new dialysis patient and using the embeddings and event subtype labels to train a personalized classifier.

### Claim 9 (depends on 8)

The computer program product of claim 8, wherein performing a forward computation comprises computing a loss function with the embeddings and the class probabilities for optimization.

### Claim 10 (depends on 8)

The computer program product of claim 8, wherein performing a forward computation comprises performing alternate optimization for a predefined number of iterations between (1) a step for obtaining posterior probabilities, and (2) a step for obtaining updated model parameters.

### Claim 11 (depends on 8)

The computer program product of claim 8, further comprising performing model testing by encoding the test data of the new dialysis patient and using the embeddings and the personalized classifier to predict event subtypes.

### Claim 12 (depends on 11)

The computer program product of claim 11, further comprising performing a dialysis event on a dialysis patient responsive to the predicted event subtypes.

### Claim 13 (depends on 8)

The computer program product of claim 8, wherein the encoder includes a static channel for encoding static patient profiles, and a temporal channel to encode temporal patient status features, wherein outputs of the static channel and the temporal channel are concatenated and projected to a compact embedding used for prediction by a combination layer of the encoder.

### Claim 14 (depends on 13)

The computer program product of claim 13, wherein the static channel comprises a multilayer perceptron (MLP), and wherein the temporal channel comprises one or more Long Short-Term Memories (LSTMs).

### Claim 15

A computer processing system for model building, comprising:
a memory device for storing program code; and
a hardware processor operatively coupled to the memory device for storing program code to
receive a training set of dialysis records and model hyperparameters;
initialize an encoder distribution related parameters;
perform, by a hardware processor, a forward computation to (1) the encoder to obtain the embeddings of the dialysis records, and (2) the distribution related parameters to obtain class probabilities;
check by a convergence evaluator if the iterative optimization has converged; and
perform model personalization responsive to model convergence by encoding the support data of a new dialysis patient and using the embeddings and event subtype labels to train a personalized classifier.

### Claim 16 (depends on 15)

The computer processing system of claim 15, wherein performing a forward computation comprises computing a loss function with the embeddings and the class probabilities for optimization.

### Claim 17 (depends on 15)

The computer processing system of claim 15, wherein performing a forward computation comprises performing alternate optimization for a predefined number of iterations between (1) a step for obtaining posterior probabilities, and (2) a step for obtaining updated model parameters.

### Claim 18 (depends on 15)

The computer processing system of claim 15, wherein the processor further runs the program code to perform model testing by encoding the test data of the new dialysis patient and using the embeddings and the personalized classifier to predict event subtypes.

### Claim 19 (depends on 18)

The computer processing system of claim 18, wherein the processor further runs the program code to control performing a dialysis event on a dialysis patient responsive to the predicted event subtypes.

### Claim 20 (depends on 15)

The computer processing system of claim 15, wherein the encoder includes a static channel for encoding static patient profiles, and a temporal channel to encode temporal patient status features, wherein outputs of the static channel and the temporal channel are concatenated and projected to a compact embedding used for prediction by a combination layer of the encoder.
