Machine learning device, estimation device, non-transitory computer readable medium, and learned model

1. A machine learning device comprising:

a generation unit generating a first shape model representing a shape of an object before deformation and a second shape model representing a shape of the object after the deformation based on measurement data before and after the deformation; and

a learning unit learning a feature amount including a difference value between each micro region and another micro region that constitute the first shape model, and a relation providing a displacement from the each micro region of the first shape model to each corresponding micro region of the second shape model.

2. The machine learning device according to claim 1 , wherein,

when the generation unit generates a third shape model with a change in positions of vertexes, which are elements of the first shape model, and a fourth shape model with a change in positions of vertexes, which are elements of the second shape model,

the learning unit also learns a feature amount including a difference value between each micro region and another micro region that constitute the third shape model, and a relation providing a displacement from the each micro region of the third shape model to each corresponding micro region of the fourth shape model.

3. The machine learning device according to claim 1 , wherein the feature amount is defined by at least one of coordinates giving each micro region, a gradient or a normal vector defining each micro region, a curvature defining each micro region, Voronoi area of each micro region, a shrinkage ratio of each micro region, and a deviation amount between each micro region and a corresponding region of an average shape model.

4. The machine learning device according to claim 1 , wherein the difference value giving the feature amount is calculated with the micro regions of at least 5% of total micro regions constituting the first shape model.

5. The machine learning device according to claim 1 , wherein the learning unit learns the relation by using a kernel regression model.

6. The machine learning device according to claim 1 , wherein the measurement data before and after the deformation is medical image data.

7. The machine learning device according to claim 6 , wherein the deformation is deaerated deformation of a lung, medical image data of a lung of a patient having treated pneumothorax is used as the measurement data before the deformation due to deaeration, and medical image data of the lung of the patient before treatment of pneumothorax is used as the measurement data after the deformation due to deaeration.

8. The machine learning device according to claim 6 , wherein the deformation includes deformation of an organ or movement of a position of an organ due to breathing, time variation, or differences in specimens.

9. The machine learning device according to claim 1 , wherein the first shape model and the second shape model are three-dimensional vertex models.

10. The machine learning device according to claim 1 , wherein

the first shape model and the second shape model are three-dimensional vertex models, and the difference value between each micro region and another micro region is a difference between vertexes.

11. A non-transitory computer readable medium storing a program causing a computer to execute:

a function of generating a first shape model representing a shape of an object before deformation and a second shape model representing a shape of the object after the deformation based on measurement data before and after the deformation; and

a function of learning a feature amount including a difference value between each micro region and another micro region of the first shape model, and a relation providing a displacement from the each micro region of the first shape model to each corresponding micro region of the second shape model.

12. An estimation device comprising:

an estimation unit, when any measurement data is given, using a learned model having learned a feature amount including a difference value between each micro region and another micro region in a first shape model generated from measurement data before deformation, and a relation providing a displacement from the each micro region of the first shape model to each corresponding micro region of a second shape model after the deformation, to thereby estimate a shape after the deformation corresponding to the any measurement data.

13. The estimation device according to claim 12 , wherein the estimation unit calculates a feature amount including a difference value between a micro region and another micro region of a shape model generated for the inputted measurement data, and estimates a displacement of a feature amount corresponding to each micro region by interpolating a relation stored in the learned model.

14. A non-transitory computer readable medium storing a program causing a computer to execute:

a function of, when any measurement data is given, using a learned model having learned a feature amount including a difference value between each micro region and another micro region in a first shape model generated from measurement data before deformation, and a relation providing a displacement from the each micro region of the first shape model to each corresponding micro region of a second shape model after the deformation, to thereby estimate a shape after the deformation corresponding to the any measurement data.

15. A non-transitory computer readable medium storing a program causing a computer to execute a function comprising:

providing a relation having a feature amount including a difference value between each micro region and another micro region in a first shape model generated from measurement data before deformation as an input, and a displacement from the each micro region of the first shape model to each corresponding micro region of a second shape model after the deformation as an output, and,

when any measurement data is given, computing a displacement corresponding to each micro region of a shape model corresponding to the any measurement data by using the relation, to thereby estimate a shape after the deformation corresponding to the any measurement data.

16. A machine learning device comprising:

a generation unit generating shape models representing individual shapes of a first object and an average shape model based on a plurality of measurement data items of the first object; and

a learning unit learning a distribution of a feature amount including a difference value between each micro region constituting the shape model representing the individual shape and each corresponding micro region of the average shape model, and a relation between a distribution of the difference value and a region where a second object having a constraint relation with the first object exists.

17. The machine learning device according to claim 16 , wherein the constraint relation includes a relation in which the second object is adjacent to or connected to the first object.

18. A non-transitory computer readable medium storing a program causing a computer to execute:

a function of generating shape models representing individual shapes of a first object and an average shape model based on a plurality of measurement data items of the first object; and

a function of learning a distribution of a feature amount including a difference value between each micro region constituting the shape model representing the individual shape and each corresponding micro region of the average shape model, and a relation between a distribution of the difference value and a region where a second object having a constraint relation with the first object exists.