# US 2022/0037023 A1 · Systems and methods for treatment selection

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

- Number: US 2022/0037023 A1
- Document ID: US20220037023A1
- Status: Published application
- Kind: A1
- Application number: 17/493,070
- Filed: 2021-10-04
- Published: 2022-02-03
- Earliest claimed: 2020-04-03
- Assignees: Aifred Health
- Inventors: Caitrin ARMSTRONG; David BENRIMOH; Robert FRATILA; Adam KAPELNER; Akiva KLEINERMAN; Joseph MEHLTRETTER; Ariel ROSENFELD
- Web page: https://patentgrep.com/patent/US20220037023A1

## CPC

- G16H 50/20: for computer-aided diagnosis, e.g. based on medical expert systems
- G06N 3/045: Combinations of networks
- G06N 3/0455: Auto-encoder networks; Encoder-decoder networks
- G06N 3/09: Supervised learning
- G06N 5/02: Knowledge representation; Symbolic representation
- G06N 5/022: Knowledge engineering; Knowledge acquisition
- G06N 20/00: Machine learning
- G16H 10/20: for electronic clinical trials or questionnaires
- G16H 10/60: for patient-specific data, e.g. for electronic patient records
- G16H 20/10: relating to drugs or medications, e.g. for ensuring correct administration to patients
- G16H 20/70: relating to mental therapies, e.g. psychological therapy or autogenous training
- G16H 40/67: for remote operation
- G16H 50/70: for mining of medical data, e.g. analysing previous cases of other patients
- G16H: HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS

## Abstract

There is disclosed a method and a system for predicting the efficacy of one or more treatments. A completed questionnaire may be received from a patient requiring treatment. The responses to the questionnaire may be input to a machine learning algorithm (MLA). The MLA may have been trained using labelled patient data. A predicted efficacy of one or more treatments and a prototype corresponding to the patient may be received from the MLA. An interface may be output indicating the predicted efficacy of the one or more treatments and the prototype.

## Claims

### Claim 1

A method for predicting a treatment efficacy for a patient, the method comprising:
training a machine learning algorithm (MLA) by:
receiving datasets from one or more sources corresponding to treatments for mental illness, wherein each data point in the datasets comprises questionnaire data corresponding to a patient and an indication of treatment efficacy corresponding to the respective patient;
normalizing results of the datasets, thereby generating normalized results;
generating, based on the normalized results, a training dataset;
selecting one or more features in the training dataset;
training, using the selected one or more features, the MLA to predict, for input patient data, an efficacy of each of a plurality of treatments; and
training the MLA to determine a prototype corresponding to the input patient data from a set of prototypes, wherein each prototype of the set of prototypes corresponds to a cluster of patient data, wherein the training is performed using a loss function, wherein the loss function determines a distance between prototypes, and wherein the loss function determines a variance in remission predictions between the prototypes, and
predicting the treatment efficacy for the patient by:
receiving questionnaire responses from the patient;
inputting the questionnaire responses into the MLA;
outputting, by the MLA, a predicted efficacy of each of the plurality of treatments for the patient;
outputting, by the MLA, a prototype of the set of prototypes corresponding to the patient;
generating, based on the predicted efficacy of the one or more treatments and the prototype, an interface; and
outputting for display the interface.

### Claim 2 (depends on 1)

The method of claim 1, wherein the interface comprises, for each of the one or more treatments, a predicted likelihood of remission.

### Claim 3 (depends on 1)

The method of claim 1, further comprising receiving, via the interface, user input indicating a treatment plan, wherein the treatment plan comprises at least one of the one or more treatments.

### Claim 4 (depends on 1)

The method of claim 1, wherein the loss function is configured to increase the distance between prototypes and increase the variance in remission predictions between the prototypes.

### Claim 5 (depends on 1)

The method of claim 1, wherein the prototype corresponding to the patient indicates a cluster of patients having similar features to the patient, wherein the interface provides an indication of a similarity between the patient and the cluster of patients, and wherein the interface explains how the prototype relates to the predicted efficacy.

### Claim 6 (depends on 1)

The method of claim 1, wherein the MLA comprises a neural network, wherein the prototype is determined by a layer of the neural network corresponding to prototypes, and wherein the layer of the neural network corresponding to prototypes outputs a distance between the patient and each prototype of the set of prototypes.

### Claim 7

A method comprising:
receiving questionnaire responses from a patient requiring treatment;
inputting the questionnaire responses into a machine learning algorithm (MLA), wherein the MLA was trained based on labelled patient data, wherein each data point in the labelled patient data comprises questionnaire data corresponding to a respective patient and a label indicating an efficacy of a treatment for the respective patient, and wherein the MLA is configured to generate a set of prototypes based on the labelled patient data;
receiving, from the MLA, a predicted efficacy of one or more treatments for the patient;
receiving, from the MLA, a prototype of the set of prototypes corresponding to the patient;
generating, based on the predicted efficacy of the one or more treatments and the prototype, an interface; and
outputting for display the interface.

### Claim 8 (depends on 7)

The method of claim 7, wherein the interface comprises:
a predicted likelihood of remission for each of the one or more treatments, and
an indication of a distance between the patient and the prototype.

### Claim 9 (depends on 7)

The method of claim 7, wherein each prototype of the set of prototypes corresponds to a cluster of patients generated from the labelled patient data.

### Claim 10 (depends on 7)

The method of claim 7, wherein the MLA comprises a neural network, and wherein the prototype is determined by a layer of the neural network corresponding to prototypes.

### Claim 11 (depends on 10)

The method of claim 10, wherein the layer of the neural network corresponding to prototypes outputs a distance between the patient and each prototype of the set of prototypes.

### Claim 12

A method for training a machine learning algorithm (MLA), the method comprising:
receiving datasets from one or more sources corresponding to treatments for mental illness, wherein each data point in the datasets comprises questionnaire data corresponding to a patient in and an indication of treatment efficacy corresponding to the respective patient;
normalizing results of the datasets, thereby generating normalized results;
generating, based on the normalized results, a training dataset;
selecting one or more features in the training dataset;
training, using the selected one or more features, the MLA to predict, for input patient data, an efficacy of each of the treatments; and
training the MLA to determine a prototype corresponding to the input patient data from a set of prototypes, wherein each prototype of the set of prototypes corresponds to a cluster of patient data, wherein the training is performed using a loss function, wherein the loss function determines a distance between prototypes of the set of prototypes, and wherein the loss function determines a variance in remission predictions between the prototypes of the set of prototypes.

### Claim 13 (depends on 12)

The method of claim 12, wherein training the MLA to determine a prototype corresponding to the input patient data comprises training the MLA based at least in part on a prototype sample distance variance.

### Claim 14 (depends on 12)

The method of claim 12, further comprising determining a prototype sample distance variance based at least in part on a variance of distances between a set of nearest samples for a given prototype and the given prototype itself.

### Claim 15 (depends on 12)

The method of claim 12, further comprising determining a prototype sample distance variance based at least in part on variance of pairwise distances between the set of prototypes.

### Claim 16 (depends on 12)

The method of claim 12, further comprising determining a prototype remission prediction based at least in part on variance of differential remission predictions for the set of prototypes across a treatment type.

### Claim 17 (depends on 12)

The method of claim 12, further comprising determining a prototype remission prediction based at least in part on variance of differential remission predictions for a given prototype across a plurality of treatment types.

### Claim 18 (depends on 12)

The method of claim 12, wherein the training is performed using a loss function, wherein the loss function determines a difference between a predicted likelihood of remission and a labeled occurrence of remission, wherein the loss function determines an autoencoder loss indicating a distance between an original sample and a decoded sample, and wherein the loss function is configured to increase the distance between prototypes and increase the variance in remission predictions between the prototypes.

### Claim 19 (depends on 12)

The method of claim 12, wherein normalizing the results of the datasets comprises grouping questions in different datasets relating to a same feature.

### Claim 20 (depends on 12)

The method of claim 12, wherein normalizing the results of the datasets comprises converting categorical responses in the datasets to binary responses.
