Systems and methods for predicting pain level

Patent metadata

Field	Value
Publication number	US-11537888-B2
Application number	US-202016875041-A
Country	US
Kind code	B2
Filing date	May 15, 2020
Priority date	May 15, 2019
Publication date	Dec 27, 2022
Grant date	Dec 27, 2022

Abstract

Devices and methods for learning and/or predicting the self-reported pain improvement levels of osteoarthritis (OA) patients are provided. A device or apparatus can include a processor and a machine-readable medium in operable communication with the processor and having stored thereon an algorithm and a unique set of features. The algorithm and set of features can enable building one or more models that learn the self-reported pain improvement levels of OA patients.

First claim

What is claimed is:

1. A method for predicting a pain level of an osteoarthritis (OA) patient, the method comprising:
developing, by a processor, a set of classifiers, the set of classifiers comprising three classifiers corresponding to a first category, a second category, and a third category, respectively;
training, by the processor, the set of classifiers;
testing, by the processor, the set of classifiers; and
using, by the processor, the set of classifiers to predict the pain level of the OA patient at a future visit intended to assess the pain level,
the first category being that pain has improved for the OA patient since a previous visit, the second category being that pain has remained unchanged for the OA patient since the previous visit, and the third category being that pain has worsened for the OA patient since the previous visit,
the developing, training, testing, and using of the set of classifiers comprising using a machine learning (ML) technique that factors in sex, age, body mass index, injury factors, occupation factors, medical history, strength performance measures, and physical activity factors for the OA patient,
the developing of the set of classifiers comprising feature selection,
the feature selection comprising utilizing a Fisher coefficient and a squared Euclidean distance on features of the set of classifiers,
the ML technique being a recurrent neural network (RNN) with three single class, multi-label RNN classifiers,
the training of the set of classifiers comprising training the set of classifiers using a dataset with known values,
the testing of the set of classifiers comprising testing the set of classifiers using the dataset with known values,
the dataset being broken into a first sub-dataset to be used for the training of the set of classifiers and a second sub-dataset to be used for the testing of the set of classifiers,
the training of the set of classifiers comprising normalization of data obtained from the dataset,
the using of the set of classifiers to predict the pain level of the OA patient comprising using the set of classifiers to predict the pain level of the OA patient at an Nth visit based on features of the set of classifiers reported on all visits up to an (N−1)th visit,
the training of the set of classifiers comprising rescaling the age and body mass index for the OA patient to a unified range from 0 to 1,
the feature selection comprising decomposing first features, of the features of the set of classifiers, into independent constituents and aggregating second features, of the features of the set of classifiers, into an aggregated group of features, and
the feature selection further comprising binarizing all of the features of the set of classifiers using a binary threshold function where a feature value of 1 is assigned to a respective feature if it is higher than a threshold of the binary threshold function and a feature value of 0 is assigned to the respective feature if it is lower than the threshold of the binary threshold function.

2. The method according to claim 1 ,
the dataset with known values being the Osteoarthritis Initiative (OAI) dataset.

3. The method according to claim 1 , the using of the set of classifiers comprising choosing a classifier from the set of classifiers to predict the pain level of the OA patient,
where, if all classifiers of the set of classifiers predict a positive class, the chosen classifier is that with a highest F1-score,
where, if only one classifier of the set of classifiers predicts a positive class, the chosen classifier is the classifier that predicts a positive result, and
where, if more than one classifier, but less than all classifiers, of the set of classifiers predicts a positive class, the chosen classifier is the classifier that with a highest F1-score that also predicts a positive result.

4. A system for predicting a pain level of an osteoarthritis (OA) patient, the system comprising:
a processor; and
a machine-readable medium in operable communication with the processor and having instructions stored thereon that, when executed by the processor, perform the following steps:
developing a set of classifiers, the set of classifiers comprising three classifiers corresponding to a first category, a second category, and a third category, respectively;
training the set of classifiers;
testing the set of classifiers; and
using the set of classifiers to predict the pain level of the OA patient at a future visit intended to assess the pain level,
the first category being that pain has improved for the OA patient since a previous visit, the second category being that pain has remained unchanged for the OA patient since the previous visit, and the third category being that pain has worsened for the OA patient since the previous visit,
the developing, training, testing, and using of the set of classifiers comprising using a machine learning (ML) technique that factors in sex, age, body mass index, injury factors, occupation factors, medical history, strength performance measures, and physical activity factors for the OA patient,
the ML technique being an RNN with three single class, multi-label RNN classifiers,
the training of the set of classifiers comprising training the set of classifiers using a dataset with known values,
the testing of the set of classifiers comprising testing the set of classifiers using the dataset with known values,
the dataset being broken into a first sub-dataset to be used for the training of the set of classifiers and a second sub-dataset to be used for the testing of the set of classifiers,
the developing of the set of classifiers comprising feature selection,
the training of the set of classifiers comprising normalization of data obtained from the dataset,
the using of the set of classifiers to predict the pain level of the OA patient comprising using the set of classifiers to predict the pain level of the OA patient at an Nth visit based on features of the set of classifiers reported on all visits up to an (N−1)th visit,
the feature selection comprising utilizing a Fisher coefficient and a squared Euclidean distance on the features of the set of classifiers,
the training of the set of classifiers comprising rescaling the age and body mass index for the OA patient to a unified range from 0 to 1,
the feature selection comprising decomposing first features, of the features of the set of classifiers, into independent constituents and aggregating second features, of the features of the set of classifiers, into an aggregated group of features, and
the feature selection further comprising binarizing all of the features of the set of classifiers using a binary threshold function where a feature value of 1 is assigned to a respective feature if it is higher than a threshold of the binary threshold function and a feature value of 0 is assigned to the respective feature if it is lower than the threshold of the binary threshold function.

5. The system according to claim 4 , the using of the set of classifiers comprising choosing a classifier from the set of classifiers to predict the pain level of the OA patient,
where, if all classifiers of the set of classifiers predict a positive class, the chosen classifier is that with a highest F1-score,
where, if only one classifier of the set of classifiers predicts a positive class, the chosen classifier is the classifier that predicts a positive result, and
where, if more than one classifier, but less than all classifiers, of the set of classifiers predicts a positive class, the chosen classifier is the classifier that with a highest F1-score that also predicts a positive result.

Assignees

• Banisakher Deya
• Rishe Naphtali
• Finlayson Mark
• The Florida International Univ Board Of Trustees

Inventors

• Banisakher Deya
• Rishe Naphtali
• Finlayson Mark

Classifications

• G06N20/10

  using kernel methods, e.g. support vector machines [SVM] · CPC title

• G16H50/20 Primary

  for computer-aided diagnosis, e.g. based on medical expert systems · CPC title

• G06N20/20

  Ensemble learning · CPC title

• G06N3/084

  Backpropagation, e.g. using gradient descent · CPC title

• G16H70/60

  relating to pathologies · CPC title