Method for generating learned model, system for generating learned model, prediction device, and prediction system
A method for generating a learned model applied to a prediction device that predicts a probability that a subject develops delirium based on a moving image in which the subject appears is provided. The method includes: acquiring first data corresponding to the moving image in which the subject appears; generating, based on the first data, second data corresponding to changes over time in relative positions of a plurality of feature points in a body of the subject in the moving image; generating third data indicating a determination result as to whether the subject develops delirium based on the moving image; and generating the learned model by causing a neural network to learn using the second data and the third data.
1. A method for generating a learned model applied to a prediction device that predicts a probability that a subject develops delirium based on a moving image including a plurality of frame images in which the subject appears, the method comprising:
acquiring first data corresponding to the frame images each of which includes a plurality of feature points in a body of the subject;
generating, based on the first data, second data corresponding to changes over time in relative positions of the feature points extracted from the frame images;
generating third data indicating a determination result as to whether the subject develops delirium based on the frame images; and
generating the learned model by causing a neural network to learn using the second data and the third data.
2. The method for generating the learned model according to claim 1 ,
wherein the second data is generated by applying, to the first data, a skeleton model in which the plurality of feature points set for at least four limbs of a human body are connected based on a predetermined relationship.
3. The method for generating the learned model according to claim 1 ,
wherein the second data is generated by applying, to the first data, an expression detection model in which the plurality of feature points are set to a face.
4. The method for generating the learned model according to claim 1 , further comprising:
acquiring fourth data corresponding to at least one of physiological information acquired from the subject, body motion information of the subject, and voice information of the subject; and
generating the learned model by causing the neural network to learn using the fourth data as well.
5. The method for generating the learned model according to claim 1 , further comprising:
acquiring fifth data corresponding to at least one of background information of the subject, the background information of the subject including at least one of sex, age, height, weight, past history, or medication information of the subject; and
generating the learned model by causing the neural network to learn using the fifth data as well.
6. A prediction device comprising:
an input interface configured to receive behavior data corresponding to changes over time in relative positions of a plurality of feature points in a body of a subject in a moving image including a plurality of frame images in which the subject appears, the moving image being generated based on image data corresponding to the frame images each of which includes the feature points;
a processor configured to acquire prediction data corresponding to a probability that the subject develops delirium by inputting the behavior data to a learned model generated by the method for generating the learned model according to claim 1 ; and
an output interface configured to output the prediction data.
7. The prediction device according to claim 6 ,
wherein the input interface is configured to receive supplementary data including at least one of physiological information acquired from the subject, body motion information of the subject, and voice information of the subject, and
wherein the processor is configured to acquire the prediction data by inputting the supplementary data in addition to the behavior data to the learned model generated by the method for generating the learned model, which further includes acquiring fourth data corresponding to at least one of physiological information acquired from the subject, body motion information of the subject, and voice information of the subject; and generating the learned model by causing the neural network to learn using the fourth data as well.
8. The prediction device according to claim 6 ,
wherein the input interface is configured to receive background data including background information of the subject, the background information of the subject including at least one of sex, age, height, weight, past history, or medication information of the subject and
wherein the processor is configured to acquire the prediction data by inputting the background data in addition to the behavior data to the learned model, the
generated by the method for generating the learned model, which further includes acquiring fifth data corresponding to at least one of background information of the subject; and
generating the learned model by causing the neural network to learn using the fifth data as well.
9. A non-transitory computer-readable recording medium storing a computer program executable by a processor of a prediction device, the computer program being executed to, by the prediction device:
receive behavior data corresponding to changes over time in relative positions of a plurality of feature points in a body of a subject in a moving image including a plurality of frame images in which the subject appears, the moving image being generated based on image data corresponding to the frame images each of which includes the feature points;
acquire prediction data corresponding to a probability that the subject develops delirium by inputting the behavior data to a learned model generated by the method for generating the learned model according to claim 1 ; and
output the prediction data.
10. A prediction system comprising:
an image processing device configured to generate, based on image data corresponding to a moving image including a plurality of frame images each of which includes a plurality of feature points in a body of a subject, behavior data corresponding to changes over time in relative positions of the feature points extracted from the frame images; and
a prediction device configured to acquire prediction data corresponding to a probability that the subject develops delirium by inputting the behavior data to a learned model generated by the method for generating the learned model according to claim 1 , and output the prediction data.
11. The prediction system according to claim 10 , further comprising:
a user interface configured to input a determination result as to whether the subject develops delirium, the determination result being made by a medical worker based on the frame images corresponding to the image data; and
a training data generation device configured to generate the third data based on the determination result.
12. A system for generating a learned model applied to a prediction device that predicts a probability that a subject develops delirium based on a moving image including a plurality of frame images in which the subject appears, the system comprising:
an image processing device configured to generate, based on first data corresponding to the frame images each of which includes a plurality of feature points in a body of the subject, second data corresponding to changes over time in relative positions of the feature points extracted from the frame images; and
a model generation device configured to generate the learned model by causing a neural network to learn using the second data and third data indicating a determination result as to whether the subject develops delirium, the determination result being made based on the frame images.
13. A non-transitory computer-readable storage medium storing a computer program executable in a system for generating a learned model applied to a prediction device that predicts a probability that a subject develops delirium based on a moving image including a plurality of frame images in which the subject appears, the computer program being executed to:
generate, by an image processing device included in the system, based on first data corresponding to the frame images each of which includes a plurality of feature points in a body of the subject, second data corresponding to changes over time in relative positions of the feature points in extracted from the frame images; and
generate, by a model generation device included in the system, the learned model by causing a neural network to learn using the second data and third data indicating a determination result as to whether the subject develops delirium, the determination result being made based on the frame images.