US-12626333-B2 - Medical-image processing apparatus, medical-image processing method, and program for the same

US12626333B2US 12626333 B2US12626333 B2US 12626333B2US-12626333-B2

Abstract

A medical-image processing apparatus according to the present invention includes an obtaining unit configured to convert resolution of a medical image of a first resolution subjected to a noise reduction process to obtain a medical image of a second resolution lower than the first resolution and a training unit configured to train a learning model using training data including the medical image of the first resolution subjected to the noise reduction process and the medical image of the second resolution.

Inventors

Hiroyuki Omi

Assignees

CANON KABUSHIKI KAISHA

Dates

Publication Date: 20260512
Application Date: 20230403
Priority Date: 20201026

Claims (14)

1 . A medical-image processing apparatus comprising: one or more processors; and at least one memory storing executable instructions, which when executed by the one or more processors, cause the medical-image processing apparatus to perform operations comprising: converting resolution of a medical image of a first resolution subjected to a noise reduction process to obtain a medical image of a second resolution lower than the first resolution; and training a learning model using (A) training data including the medical image of the first resolution subjected to the noise reduction process and the medical image of the second resolution, and (B) a training parameter determined based on a noise amount of the medical image of the first resolution.
2 . The medical-image processing apparatus according to claim 1 , the operations further comprising: performing the noise reduction process on the medical image of the first resolution, wherein the converting converts the resolution of the medical image of the first resolution subjected to the noise reduction process to obtain the medical image of the second resolution lower than the first resolution.
3 . The medical-image processing apparatus according to claim 2 , wherein the performing performs the noise reduction process on the medical image of the first resolution using a nonlinear spatial filter.
4 . The medical-image processing apparatus according to claim 2 , wherein the performing performs the noise reduction process by averaging the medical image of the first resolution and a plurality of medical images captured under the same image-capturing conditions as conditions for the medical image of the first resolution.
5 . A medical-image processing apparatus comprising: one or more processors; and at least one memory storing executable instructions, which when executed by the one or more processors, cause the medical-image processing apparatus to perform operations comprising: estimating a noise amount of a medical image of a first resolution; converting resolution of the medical image of the first resolution selected based on the estimated noise amount to obtain a medical image of a second resolution lower than the first resolution; and training a learning model using (A) training data including the medical image of the first resolution and the medical image of the second resolution, and (B) a training parameter determined based on a noise amount of the medical image of the first resolution.
6 . The medical-image processing apparatus according to claim 5 , wherein the estimating estimates the noise amount of the medical image of the first resolution based on at least one of a standard deviation of a noise amount in a predetermined area of the medical image of the first resolution, an average of standard deviations in a plurality of areas of the medical image of the first resolution, medical physical properties, and a ratio of a signal value to the noise amount of the medical image of the first resolution.
7 . The medical-image processing apparatus according to claim 5 , wherein, when the estimated noise amount is less than a predetermined threshold, the converting converts the resolution of the medical image of the first resolution to obtain the medical image of the second resolution lower than the first resolution.
8 . The medical-image processing apparatus according to claim 1 , wherein the training parameter includes a training rate, and wherein, when the noise amount is a predetermined value or more, the training rate is less than a training rate when the noise amount is less than the predetermined value.
9 . The medical-image processing apparatus according to claim 8 , wherein the training parameter includes a loss function, and wherein, when the noise amount is the predetermined value or more, the loss function is smaller than a loss function when the noise amount is less than the predetermined value.
10 . The medical-image processing apparatus according to claim 1 , wherein the medical image comprises a radiographic image.
11 . A medical-image processing apparatus comprising: one or more processors; and at least one memory storing executable instructions, which when executed by the one or more processors, cause the medical-image processing apparatus to perform operations comprising: obtaining a medical image obtained by capturing an image of an examinee; and generating a medical image of a resolution higher than a resolution of the obtained medical image using a learning model that has learned using (A) training data including a medical image of a first resolution subjected to a noise reduction process and a medical image of a second resolution lower than the first resolution, the medical image of the second resolution being generated by converting resolution of the medical image of the first resolution subjected to the noise reduction process, and (B) a training parameter determined based on a noise amount of the medical image of the first resolution.
12 . The medical-image processing apparatus according to claim 11 , wherein the training parameter includes a training rate, and wherein, when the noise amount is a predetermined value or more, the training rate is less than a training rate when the noise amount is less than the predetermined value.
13 . The medical-image processing apparatus according to claim 12 , wherein the training parameter includes a loss function, and wherein, when the noise amount is the predetermined value or more, the loss function is smaller than a loss function when the noise amount is less than the predetermined value.
14 . The medical-image processing apparatus according to claim 12 , wherein the training parameter includes a loss function, wherein the loss function when the noise amount is more than the predetermined value is an absolute value of an error between an output image output from the learned model and a correct image, and wherein the loss function when the noise amount is less than the predetermined value is a square of the error.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS This application is a Continuation of International Patent Application No. PCT/JP2021/038600, filed Oct. 19, 2021, which claims the benefit of Japanese Patent Application No. 2020-179041, filed Oct. 26, 2020, both of which are hereby incorporated by reference herein in their entirety. TECHNICAL FIELD The present invention relates to a medical-image processing apparatus, a medical-image processing method, and a program for the same. BACKGROUND ART X-ray diagnosis and treatment based on radiography are widely performed in medical front, and digital diagnostic imaging based on radiographic images captured using a flat panel sensor (hereinafter referred to as “sensor”) is in widespread use all over the world. The sensor can image output immediately and can therefore capture not only still images but also moving images. Furthermore, an increase in the resolution of the sensor allows imaging that provides more detailed information. In contrast, reduced-resolution radiographic images are sometimes obtained to reduce radiation exposure to the examinee. One example is a use case in which X rays are applied for a long time, such as moving images. In this case, X-ray dose per pixel is increased by treating multiple pixels of output data from the sensor as one pixel. This allows the overall X-ray radiation to be reduced, thereby reducing radiation exposure to the examinee. However, the reduction in resolution causes loss of detailed information in the radiographic images, such as lesion information and information for accurate positioning of the imaging apparatus. One example of a process for decompressing detailed information in low-resolution images (increasing the resolution) is superresolution processing. Known examples of the superresolution processing include a method for converting multiple low-resolution images to a high-resolution image and a method for associating the features of a low-resolution image with the features of a high-resolution image and providing a high-resolution image on the basis of the information (PTL 1). A recent example method for associating features is machine learning. In particular, supervised learning using a convolutional neural network (CNN) is rapidly becoming popular because of their high performance (PTL 2). Superresolution processing using the CNN decompresses detailed information in input low-resolution images using training parameters created by means of supervised learning. The superresolution processing is also applied to medical images. Superresolution processing using the CNN makes an inference using a low-resolution image as an input and outputs a superresolution image as an inference. A high-resolution image is used as a correct image for training. For this reason, multiple sets of a high-resolution image and a low-resolution image are prepared as training data. An example of the image is a medical image. Preparing medical images allows the superresolution processing to be applied to medical images. However, for example, radiographic images are noise dominant images having a low signal/noise (S/N) ratio in a low dose area. For this reason, if the CNN is trained for radiographic images, the CNN learns such that not only the structure of the object, which is a signal component, but also a noise component is decompressed as information to be decompressed. Accordingly, superresolution processing using training parameters obtained by performing the learning generates a superresolution image in which noise is superimposed. Therefore, even if an image in a high dose area with low noise is input, a noisy superresolution image is output, causing degradation of image quality although with improved resolution. CITATION LIST Patent Literature PTL 1 Japanese Patent No. 4529804PTL 2 Japanese Patent No. 6276901PTL 3 Japanese Patent Laid-Open No. 2020-141908 SUMMARY OF INVENTION The present invention is made in view of the above problems, and an object is to build a learning model capable of outputting medical images of improved resolution and reduced noise. Another object of the present invention is to offer operational advantages that are provided using the configurations of the following embodiments and that are not provided using known techniques. A medical-image processing apparatus according to the present invention includes an obtaining unit configured to convert resolution of a medical image of a first resolution subjected to a noise reduction process to obtain a medical image of a second resolution lower than the first resolution and a training unit configured to train a learning model using training data including the medical image of the first resolution subjected to the noise reduction process and the medical image of the second resolution. Further features of the present invention will become apparent from the following description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF DRAWINGS FIG. 1