CN114496211A - Machine learning-based nursing decision assistance system for dementia - Google Patents

Abstract

The invention relates to a cognitive disorder nursing decision auxiliary system based on machine learning, which can provide a personalized nursing scheme for a cognitive disorder patient, and comprises the following components: the machine learning module is used for constructing a multidimensional cognition disorder classifier for classifying the cognition disorder types of the elder with the cognition disorder by using a machine learning algorithm; the input module is configured to receive input of data to be evaluated of the cognitive disorder patient; the classification module is used for classifying and evaluating the input data to be evaluated of the cognitive disorder patient based on the multidimensional cognitive disorder classifier, so that each matching probability of the cognitive disorder patient conforming to each cognitive disorder type corresponding to each classifier of the multidimensional cognitive disorder classifier is calculated, and a maximum matching probability is extracted; an output module for outputting the set of each matching probability calculated by the classification module as a result; and a countermeasure module for making countermeasures corresponding to the output result of the output module.

Description

Machine learning-based nursing decision assistance system for dementia

technical field

The present invention relates to a dementia nursing decision assistance system based on machine learning.

Background technique

With the improvement of economic and technological level and the arrival of an aging society, the nursing requirements for the elderly (elderly) people with dementia are getting higher and higher. Due to the different symptoms of the elderly with dementia, it is necessary to provide individualized care services for the elderly with dementia. However, the current nursing service mainly relies on the experience of the nursing staff and the subjective judgment of the patient's condition, and lacks the precise grasp of the situation of the elderly with dementia. own personalised care services and reduce the workload of caregivers accordingly.

SUMMARY OF THE INVENTION

The object of the present invention is to solve the above-mentioned problems at least to some extent.

To this end, the present invention provides a machine learning-based nursing decision assistance system for dementia. The system includes: a machine learning module, which uses a machine learning algorithm to construct a system for determining the type of dementia of the elderly with dementia. A classified multidimensional dementia classifier; an input module configured to receive input of the data to be evaluated for the elderly with dementia; The above-mentioned data to be evaluated is classified and evaluated, so as to calculate each matching probability of the dementia type corresponding to each classifier of the above-mentioned multi-dimensional dementia classifier, and extract a maximum matching probability for the elderly with dementia. probability; an output module, outputting the set of each matching probability calculated by the classification module as a result; and a countermeasure module, making corresponding countermeasures according to the result output by the output module.

In addition, on the basis of the above-mentioned machine learning-based nursing decision assistance system for dementia, the multi-dimensional dementia classifier includes: a cognitive impairment type classifier, which calculates that the elderly with dementia belongs to the cognitive impairment type dementia The matching probability of the environmental discomfort type classifier, calculate the matching probability that the above-mentioned dementia elderly belong to the environmental discomfort type dementia, and the water-deficient body disorder type classifier: calculate that the above-mentioned dementia elderly belong to the dehydration-induced physical disorder The matching probability of dysregulated dementia, the constipation-induced physical irregularity classifier: Calculate the matching probability of the above-mentioned dementia elders belonging to the constipation-induced physical irregularity type, and the low physical strength caused the physical irregularity type classifier: Calculate the matching probability that the above-mentioned dementia elders belong to the physical disorder caused by low physical strength. The matching probability of cognitive disorders, the inhibitory paradoxical classifier, calculate the matching probability that the above-mentioned dementia elders belong to the inhibitory paradoxical dementia, the loneliness paradoxical classifier: calculate that the above-mentioned dementia elders belong to the autistic paradoxical dementia The matching probability of , the free classifier: calculates the matching probability that the above-mentioned elderly with dementia belong to the isolated type of dementia, and the regression classifier: calculates the matching probability that the above-mentioned elderly with dementia belong to the regression type dementia.

In addition, on the basis of the above-mentioned machine learning-based nursing decision assistance system for dementia, the above-mentioned machine learning algorithm includes one or more of the following: linear regression learning algorithm, decision tree learning algorithm, ensemble learning algorithm, Bayesian optimization algorithm , neural network learning algorithms, and support vector machine learning algorithms.

In addition, on the basis of the above-mentioned machine learning-based nursing decision assistance system for dementia, the above-mentioned machine learning module includes an tuning module, and the above-mentioned tuning module tunes the system, so that the dementia corresponding to the above-mentioned maximum matching probability The type conforms to the type of dementia corresponding to the maximum matching probability calculated by each of the single dementia classifiers included in the multi-dimensional dementia classifier.

According to the above-mentioned machine learning-based nursing decision assistance system for dementia provided by the present invention, it is possible to efficiently provide personalized nursing services for the elderly with dementia and correspondingly reduce the workload of nursing staff. In addition, to a certain extent, it can solve the overfitting of the single-feature dementia training set, the care plan is poorly considered, it is difficult to implement due to the physical and mental conditions of the elderly, and the rehabilitation effect is limited by the short board effect. The problem.

Description of drawings

FIG. 1 is an exemplary block diagram showing the configuration of a dementia care decision support system 100 according to an embodiment of the present invention.

2 is a flowchart showing an example of the flow of training and processing performed by the dementia care decision support system 100 according to the embodiment of the present invention.

3 is a graph summarizing the types of dementia types according to the embodiment of the present invention.

Detailed ways

Hereinafter, the present invention will be described based on the embodiments of the present invention, but the following embodiments do not limit the invention according to the claims.

FIG. 1 is an exemplary block diagram showing the configuration of a dementia care decision support system 100 according to an embodiment of the present invention. The dementia nursing decision assistance system 100 includes a machine learning module 10 , an input module 20 , a classification module 30 , an output module 40 , and a countermeasure module 50 . The machine learning module 10 trains the training sample data through a preset machine learning algorithm, and after the training, constructs a multi-dimensional dementia classifier for classifying the type of dementia of the elderly with dementia (this In an embodiment, a 10-dimensional dementia classifier). In the training process of the machine learning module 10, a suitable machine learning algorithm can be selected for training as required, wherein the machine learning algorithm includes one or more of the following: linear regression learning algorithm, decision tree learning algorithm, ensemble learning algorithm, Bayesian optimization Algorithms, Neural Network Learning Algorithms, and Support Vector Machine Learning Algorithms. The input module 20 is configured to receive input of data to be assessed for the dementia elderly. The classification module 30, based on the 10-dimensional dementia classifier constructed by the machine learning module 10, respectively classifies and evaluates the input data to be evaluated of the elderly with dementia, so as to calculate that the elderly with dementia conforms to the 10-dimensional cognition Each matching probability of the dementia type corresponding to each classifier of the symptom classifier is extracted, and a maximum matching probability is extracted. The output module 40 outputs each set of matching probabilities calculated by the classification module 30 as a result. The countermeasure module 50 makes corresponding countermeasures according to the result output by the output module 40 .

2 is a flowchart showing an example of the flow of training and processing performed by the dementia care decision support system 100 according to the embodiment of the present invention. First, take most of the real elderly data with single and typical features of dementia as the original data samples, and filter out the unqualified ones that do not conform to the data protocol according to the physical assessment information of multiple elderly people who have determined the type of dementia. Data, construct a sampling sample pool for bootstrap to have back random sampling. Then, train the machine learning model to train a random forest model composed of N decision trees: according to the N (N=10) classification of the elderly with dementia, at the internal nodes of the N decision trees, according to the Shannon entropy The principle of the fastest decline selects a feature with the best classification effect from the feature vector of the sample, and divides the node into two branches. The Shannon entropy is defined as follows:

Repeat this step for each internal node of the decision tree until the decision tree can accurately classify the training samples with high probability, and repeat the above steps for each training sample, so that the decision tree corresponding to the training sample is constructed. In the process of model optimization, the elderly data with multiple features of dementia are added as training/testing samples, and the parameters of the random forest model are adjusted, including but not limited to the number of categories of classification results, the number of decision trees, and feature subsets. Sampling strategy, purity calculation strategy (gini, Gini coefficient, entropy, entropy), maximum tree level, random seed, maximum number of feature bins, etc., adjust the random forest classification model so that the single classification output with the maximum probability has a high probability of meeting the test set. In addition, the original typical random forest model is improved, and the last layer of the random forest classifier is adjusted to take the probability output of N decision trees (ie: the probability that an elderly person belongs to a certain type of dementia) as input, not only The decision tree number with the maximum probability value is selected as the output (that is, the maximum probability of an elderly person belongs to a certain type of dementia), and the random forest model classifier is improved to take the probability output of N decision trees as input, The N-dimensional feature vector is directly output, and the ith element Ni of the vector is the probability output of the ith decision tree, that is, the probability that an elderly person belongs to a certain type of dementia. Thus, the training and tuning process of the machine learning module is completed.

Among them, the 10-dimensional random forest model classifier specifically includes: cognitive impairment type classifier for 10 types of cognitive disorders, calculating the matching probability that the elderly with dementia belong to cognitive impairment type dementia, environmental discomfort type classifier, calculating The matching probability that the elderly with dementia belong to the environmental discomfort type dementia, and the physical disorder caused by water shortage classifier: Calculate the matching probability that the elderly with dementia belong to the physical disorder caused by water shortage, and the physical disorder caused by constipation is calculated. Disordered type classifier: Calculate the matching probability that the elderly with dementia belong to the disordered type of dementia caused by constipation, and the physical disorder caused by low physical strength The matching probability of dementia, the disease or injury-induced physical disorder type classifier: calculate the matching probability of the dementia elderly belonging to the disease or injury-induced physical disorder type dementia, suppress the contradictory classifier, calculate the dementia The matching probability that the elderly belong to the inhibitory paradoxical dementia, the loneliness paradox classifier: calculate the matching probability that the elderly with dementia belong to the loneliness paradoxical dementia, the free classifier: calculate that the elderly with dementia belong to the free type The matching probability of dementia, and the regression classifier: Calculate the matching probability that the elderly with dementia belong to the regression type of dementia. For the above 10 types of cognitive disorders, please refer to the chart shown in Figure 3.

When the dementia nursing decision assistance system 100 after training is put into use, the relevant information of the elderly with dementia is input through the input module 20, and the output module 40 of the dementia nursing decision assistance system 100 outputs a 10-dimensional feature vector. The module 50 generates a nursing plan strategy according to the 10-dimensional feature vector. For quantifiable standards, such as water intake, food intake, etc., the highest value under the limitation of basic pathological conditions is taken. The care measures included in the classified care plan, such as dedicated care, cognitive recovery training, etc., are emphasized in the final care plan and designated to be responsible. For the care schedule that needs to be carried out during special periods, it does not violate the first two. On a basis, prioritize. Specific examples are given below. Case 1: The elder is 60% [cognitive disorder] and 40% [physical disorder], and the corresponding plan: Based on the four basic nursing requirements corresponding to [cognitive disorder], for the condition of [physical disorder] , and develop further care plans, such as increasing the daily water intake target, and emphasizing hydration before the afternoon and evening every day. Case 2: The elder is 40% [environmental discomfort] and 25% [free type], and its corresponding plan: emphasizes the practice of improving the cognitive ability of the elderly, which is very important for both types of dementia, and eliminates the The "fuse" for the onset of these two types of elderly people with dementia is to arrange for the caller to approach the elderly consciously and in a planned way, and try to let the elderly do some familiar tasks within their ability.

Although the embodiment of the present invention has been described, the embodiment is presented as an example and is not intended to limit the scope of the invention. The new embodiment can be implemented in other various forms, and various abbreviations, substitutions, and changes can be made in the range which does not deviate from the summary of invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the scope equivalent thereto.

In addition, in the present embodiment, the learning model is set as a 10-dimensional dementia classifier based on 10 types of dementia, but it is not limited to this. For example, the classifier can also be set to have more than 10 dimensions, or more than 2 dimensions and less than 10 dimensions, as required.

Description of reference numbers:

100...dementia nursing decision assistance system; 10...machine learning module; 20...input module; 30...classification module; 40...output module; 50...countermeasure module.

Claims (4)

1. A machine learning-based cognitive impairment care decision assistance system, comprising:

the machine learning module is used for constructing a multidimensional cognition disorder classifier for classifying the cognition disorder types of the elder with the cognition disorder by using a machine learning algorithm;

the input module is configured to receive input of data to be evaluated of the cognitive disorder patient;

the classification module is used for performing classification evaluation on the input data to be evaluated of the cognitive disorder owner respectively based on the multidimensional cognitive disorder classifier so as to calculate each matching probability of the cognitive disorder owner according with each cognitive disorder type corresponding to each classifier of the multidimensional cognitive disorder classifier and extract a maximum matching probability;

an output module that outputs the set of each matching probability calculated by the classification module as a result; and

and the countermeasure module is used for making countermeasures corresponding to the result output by the output module.

2. The cognitive disorder care decision assistance system according to claim 1,

the multi-dimensional cognitive disorder classifier includes:

a cognitive impairment type classifier for calculating a matching probability that the cognitive impairment type person belongs to the cognitive impairment type person,

an environment-inappropriate type classifier for calculating a matching probability that the cognitive disorder person belongs to the environment-inappropriate type cognitive disorder,

lack of water and cause abnormal type sorter: calculating the matching probability of the cognition symptom persons belonging to the abnormal body type cognition symptom caused by water shortage,

sorter of abnormal body caused by constipation: calculating the matching probability of the cognition symptom patients belonging to the abnormal body type cognition symptom caused by constipation,

the physical strength is low to cause the abnormal shape classifier: calculating the matching probability of the cognition symptom of the person with the cognitive disorder who belongs to the body disorder caused by low physical strength,

disease or injury caused by abnormal body type classifier: calculating the matching probability of the cognition disorder patients belonging to diseases or injuries and causing the body disorder cognition disorder,

a conflict suppression classifier for calculating a matching probability that the cognitive disorder keeper belongs to conflict suppression cognitive disorders,

a lone spear shield classifier: calculating the matching probability of the cognition symptom long person belonging to the autism contradiction type cognition symptom,

free type classifier: calculating the probability of the cognitive disorder-free cognitive disorder-belonging match of the cognitive disorder-free person, and

a regression classifier: and calculating the matching probability of the cognitive disorder owner belonging to regression type cognitive disorders.

3. The cognitive disorder care decision assistance system according to claim 1,

the machine learning algorithm includes one or more of: linear regression learning algorithms, decision tree learning algorithms, ensemble learning algorithms, bayesian optimization algorithms, neural network learning algorithms, and support vector machine learning algorithms.

4. The cognitive disorder care decision assistance system according to claim 2,

the machine learning module comprises an adjusting module, and the adjusting module adjusts a system so that the cognition type corresponding to the maximum matching probability is in accordance with the cognition type corresponding to the maximum matching probability calculated when each single cognition classifier included in the multidimensional cognition classifier works respectively.

Images