Granted Patent B2

US 12,608,828 · App. 18/184,690 · Granted Apr 21, 2026

Inventor: Caihua Wang (Kanagawa, JP)

Assignee: FUJIFILM Corporation

G06T7/30G06T7/0012G06V10/761G06V10/764G16H30/40G16H50/20G06T2207/10088G06T2207/20081G06T2207/30016

Abstract

Claims (55)

1 . An image processing apparatus comprising:

a processor; and

a memory connected to or built in the processor,

wherein the processor is configured to

perform non-linear registration processing on a first medical image and a second medical image among a plurality of medical images to derive transformation amounts as a result of the non-linear registration processing, wherein the medical images are classified into classes,

generate transformation coefficients based on the classes of the first medical image and the second medical image,

multiply the transformation coefficients by the derived transformation amounts to obtain corrected transformation amounts, and

generate at least one new medical image that is used for training a machine learning model for the medical images by transforming at least one medical image of the first medical image or the second medical image using at least one of the corrected transformation amounts.

2 . The image processing apparatus according to claim 1 ,

wherein the transformation coefficients comprise a transformation coefficient α and a transformation coefficient β,

the transformation amounts comprise a transformation amount T_ 12 from the first medical image to the second medical image in the non-linear registration processing, and a transformation amount T_ 21 from the second medical image to the first medical image in the non-linear registration processing, and

the corrected transformation amounts comprise a corrected transformation amount αT_ 12 obtained by multiplying the transformation amount T_ 12 by the transformation coefficient α, and a corrected transformation amount βT_ 21 obtained by multiplying the transformation amount T_ 21 by the transformation coefficient β,

wherein the processor is configured to

set the first medical image as a first new medical image by applying, to the first medical image, the corrected transformation amount αT_ 12 , and

set the second medical image as a second new medical image by applying, to the second medical image, the corrected transformation amount βT_ 21 .

3 . The image processing apparatus according to claim 2 ,

wherein

the processor is configured to change values of the transformation coefficients α and β depending on whether the classes of the first medical image and the second medical image are the same or different from each other.

4 . The image processing apparatus according to claim 3 ,

wherein the processor is configured to

set a class of the first new medical image to be the same as the class of the first medical image, and

set a class of the second new medical image to be the same as the class of the second medical image.

5 . The image processing apparatus according to claim 2 ,

wherein the transformation coefficients α and β are random numbers according to a normal distribution.

6 . The image processing apparatus according to claim 5 ,

wherein

the processor is configured to

change values of the transformation coefficients α and β depending on whether the classes of the first medical image and the second medical image are the same or different from each other, and

a mean of a normal distribution in a case where the classes of the first medical image and the second medical image are different from each other is smaller than a mean of a normal distribution in a case where the classes of the first medical image and the second medical image are the same.

7 . The image processing apparatus according to claim 5 ,

wherein

the processor is configured to

change values of the transformation coefficients α and β depending on whether the classes of the first medical image and the second medical image are the same or different from each other,

set a class of the first new medical image to be the same as the class of the first medical image,

set a class of the second new medical image to be the same as the class of the second medical image, and

a mean of a normal distribution in a case where the classes of the first medical image and the second medical image are different from each other is smaller than a mean of a normal distribution in a case where the classes of the first medical image and the second medical image are the same.

8 . The image processing apparatus according to claim 1 ,

wherein the processor is configured to perform normalization processing of matching the first medical image and the second medical image with a reference medical image prior to the non-linear registration processing.

9 . The image processing apparatus according to claim 1 ,

wherein the medical image is an image in which a head of a patient appears, and

the machine learning model is a model that outputs a dementia opinion on the patient.

10 . An operation method of an image processing apparatus, the method comprising:

performing non-linear registration processing on a first medical image and a second medical image among a plurality of medical images to derive transformation amounts as a result of the non-linear registration processing, wherein the medical images are classified into classes,

generating transformation coefficients based on the classes of the first medical image and the second medical image,

multiplying the transformation coefficients by the derived transformation amounts to obtain corrected transformation amounts; and

generating at least one new medical image that is used for training a machine learning model for the medical images by transforming at least one medical image of the first medical image or the second medical image using at least one of the corrected transformation amounts.

11 . A non-transitory computer-readable storage medium storing an operation program of an image processing apparatus, the program causing a computer to execute a process comprising:

performing non-linear registration processing on a first medical image and a second medical image among a plurality of medical images to derive transformation amounts as a result of the non-linear registration processing, wherein the medical images are classified into classes,

generating transformation coefficients based on the classes of the first medical image and the second medical image,

multiplying the transformation coefficients by the derived transformation amounts to obtain corrected transformation amounts; and

generating at least one new medical image that is used for training a machine learning model for the medical images by transforming at least one medical image of the first medical image or the second medical image using at least one of the corrected transformation amounts.

12 . A non-transitory computer-readable storage medium storing a trained model that is trained by using a new medical image as a learning image, the new medical image being generated by transforming at least one medical image of a first medical image or a second medical image among a plurality of medical images using at least one of corrected transformation amounts, wherein the corrected transformation amounts is obtained by:

performing non-linear registration processing on the first medical image and the second medical image to derive transformation amounts as a result of the non-linear registration processing, wherein the medical images are classified into classes,

generating transformation coefficients based on the classes of the first medical image and the second medical image, and

multiplying the transformation coefficients by the derived transformation amounts to obtain the corrected transformation amounts.

Assignments (1)

Priority Claims (1)

Continuity (2)

References Cited (26)