Method for predicting state of object on basis of dynamic image data and computing device performing same

1 . A method for predicting a state of an object based on dynamic image data, the method being performed by an apparatus and comprising:

obtaining a plurality of pieces of image data for training;

extracting a plurality of pieces of dynamic image data for training from the plurality of pieces of image data for training, the plurality of pieces of dynamic image data for training including a playback interval image from a first time when a blood flow influence starts to decrease to a predetermined reference time point, after a time point when a drug is injected into a learning object included in each of the plurality of pieces of image data for training;

obtaining blood flow image information and disease-specific information corresponding to the plurality of pieces of dynamic image data for training;

performing learning using a first image prediction model for diagnosis to generate first training data, based on the blood flow image information and the disease-specific information corresponding to the plurality of pieces of dynamic image data for training; and

predicting the blood flow image information and the disease-specific information corresponding to a diagnosis object using dynamic image data for diagnosis and the first training data, when the dynamic image data for diagnosis corresponding to the diagnosis object is obtained.

2 . The method of claim 1 , further comprising:

obtaining early dynamic image data including a playback interval image from the time point when the drug is injected to the first time, the early dynamic image data corresponding to the dynamic image data for training, and delay image data including a playback interval image after the reference time point,

wherein the obtaining of the blood flow image information and the disease-specific information corresponding to the plurality of pieces of dynamic image data for training includes:

obtaining the blood flow image information based on the early dynamic image data and obtaining the disease-specific information based on the delay image data.

3 . The method of claim 2 , wherein the predicting of the blood flow image information and the disease-specific information corresponding to the diagnosis object includes:

predicting new early dynamic image data and new delay image data corresponding to the dynamic image data for diagnosis based on the dynamic image data for diagnosis and the first training data.

4 . The method of claim 3 , wherein the predicting of the blood flow image information and the disease-specific information corresponding to the dynamic image data for diagnosis includes:

normalizing the image data for training and image data for diagnosis based on a maximum value or an average value between brightness of the dynamic image data for training and brightness of the dynamic image data for diagnosis and predicting the new early dynamic image data and the new delay image data corresponding to the image data for diagnosis.

5 . The method of claim 3 , wherein the forming of the first training data includes:

matching and learning a change in each pixel in the dynamic image data for training over time with each pixel in the early dynamic image data and the delay image data.

6 . The method of claim 1 , wherein the image data for training is positron emission tomography image data.

7 . The method of claim 1 , wherein, when the drug used to capture the image data for training is a tracer bound to a specific target area, the dynamic image data for training is formed as image data obtained continuously or discontinuously from a time point when a dose ratio difference value between the target area and a reference area is greater than or equal to a specific value or an amount of change in dose in the reference area due to blood flow decreases to a time point when the dose ratio difference value is maximized.

8 . The method of claim 7 , wherein the predetermined reference time point is determined based on a type of the tracer.

9 . The method of claim 2 , further comprising:

predicting early dynamic image data and delay image data using the dynamic image data for training by means of a second image prediction model for diagnosis; and

using the predicted early dynamic image data and the predicted delay image data together to form second training data for predicting the blood flow image information and the disease-specific information.

10 . The method of claim 9 , further comprising:

predicting the blood flow image information and the disease-specific information, using an image prediction model for diagnosis, the image prediction model having higher accuracy between accuracy of the first image prediction model for diagnosis and accuracy of the second image prediction model for diagnosis, the first image prediction model for diagnosis and the second image prediction model for diagnosis corresponding to the image data.

11 . A computing device, comprising:

a display;

a memory storing an image prediction model for diagnosis; and

at least one processor configured to communicate with the display and the memory,

wherein the at least one processor is configured to:

obtain a plurality of pieces of image data for training;

extract a plurality of pieces of dynamic image data for training from the plurality of pieces of image data for training, the plurality of pieces of dynamic image data for training including a playback interval image from a first time when a blood flow influence starts to decrease to a predetermined reference time point, after a time point when a drug is injected into a learning object included in each of the plurality of pieces of image data for training;

obtain blood flow image information and disease-specific information corresponding to the plurality of pieces of dynamic image data for training;

perform learning using a first image prediction model for diagnosis to generate first training data, based on the blood flow image information and the disease-specific information corresponding to the plurality of pieces of dynamic image data for training; and

predict the blood flow image information and the disease-specific information corresponding to a diagnosis object using dynamic image data for diagnosis and the first training data, when the dynamic image data for diagnosis corresponding to the diagnosis object is obtained.

12 . The computing device of claim 11 , wherein the at least one processor is configured to:

further obtain early dynamic image data including a playback interval image from the time point when the drug is injected to the first time, the early dynamic image data corresponding to the dynamic image data for training, and delay image data including a playback interval image after the reference time point; and

obtain the blood flow image information based on the early dynamic image data and obtain the disease-specific information based on the delay image data, when obtaining the blood flow image information and the disease-specific information corresponding to the plurality of pieces of dynamic image data for training.

13 . The computing device of claim 12 , wherein the at least one processor is configured to:

predicting new early dynamic image data and new delay image data corresponding to the dynamic image data for diagnosis based on the dynamic image data for diagnosis and the first training data, when predicting the blood flow image information and the disease-specific information corresponding to the diagnosis object.

14 . The computing device of claim 13 , wherein the at least one processor is configured to:

normalize the image data for training and image data for diagnosis based on a maximum value or an average value between brightness of the dynamic image data for training and brightness of the dynamic image data for diagnosis and predict the new early dynamic image data and the new delay image data corresponding to the image data for diagnosis, when predicting the blood flow image information and the disease-specific information corresponding to the dynamic image data for diagnosis.

15 . The computing device of claim 13 , wherein the at least one processor is configured to:

match and learn a change in each pixel in the dynamic image data for training over time with each pixel in the early dynamic image data and the delay image data, when forming the first training data.