Granted Patent B2

US 12,749,246 · App. 17/701,826 · Granted Sep 29, 2026

Inventors: Keisuke Yamakawa (Kashiwa, JP); Taiga Goto (Kashiwa, JP)

Assignee: FUJIFILM Corporation

G06T12/30G06T5/50G06T7/0012G06T2207/10081

Abstract

Claims (37)

1 . A medical image processing apparatus comprising a processor and a non-transitory storage medium storing one or more programs and data for executing the programs, the programs being instructions executable by the processor to configure said medical image processing apparatus to perform a method comprising:

obtaining tomographic image data by reconstructing a tomographic image from projection data of an object under examination including a metal, the reconstructed tomographic image including metal artifacts;

obtaining machine learning output image data by acquiring a machine learning output image that includes reduced metal artifacts and is output when the tomographic image data is input to a machine learning engine that has machine-learned to reduce the metal artifacts in the tomographic image of the object under examination;

acquiring a weight map based on a beam hardening correction image obtained by applying beam hardening correction to the tomographic image, weight coefficients of the weight map being mapped, and the weight map indicating a probability distribution of a presence of metal artifacts; and

generating a composite image, by using the weight map based on the beam hardening correction image, to composite (i) the machine learning output image data and (ii) the tomographic image data of the tomographic image of the object under examination,

wherein for the machine learning output image data and the tomographic image data that are composited, an image quality of a region of the machine learning output image that is less affected by the metal artifacts than a corresponding region of the tomographic image that includes the metal artifacts is degraded relative to an image quality of a corresponding region in the tomographic image less affected by the metal artifacts, and an image quality of a counterpart region of the composite image corresponding to the region of the machine learning output image and the corresponding region in the tomographic image is preserved relative to an image quality of the corresponding region in the tomographic image which is composited with the machine learning output image, and

wherein the machine learning output image and the tomographic image are composited together by using a value obtained by multiplying the weight coefficient by an adjustment coefficient set in an adjustment coefficient setting portion,

the composite image is displayed in a same window as the adjustment coefficient setting portion and is updated every time the adjustment coefficient is set in the adjustment coefficient setting portion, and

the same window includes an input image display portion displaying the tomographic image and the machine learning output image.

2 . The medical image processing apparatus according to claim 1 , wherein the weight map indicates a distribution of absolute values of differences between the tomographic image and a linear interpolation image that is obtained by applying a linear interpolation technique to the tomographic image.

3 . The medical image processing apparatus according to claim 1 , wherein the weight coefficients become smaller with an increasing distance from a metal pixel extracted from the tomographic image.

4 . The medical image processing apparatus according to claim 3 , wherein the weight coefficient becomes larger as the metal pixel has a larger pixel value.

5 . The medical image processing apparatus according to claim 1 , wherein

each weight coefficient w is a real number between 0 and 1, and

for each pixel of the composite image,

(a) a pixel value of the machine learning output image is weighted by w,

(b) a pixel value of the tomographic image is weighted by (1−w), and

(c) the weighted pixel value of the machine learning output image and the weighted pixel value of the tomographic image are combined to obtain a pixel value of the composite image.

6 . The medical image processing apparatus according to claim 1 , wherein the weight map reflects a distribution of absolute values of differences between the tomographic image and the beam hardening correction image.

7 . A medical image processing method comprising the steps of:

obtaining tomographic image data by reconstructing a tomographic image from projection data of an object under examination including a metal;

obtaining machine learning output image data by acquiring a machine learning output image that includes reduced metal artifacts and is output when the tomographic image data of the object under examination and additionally including metal artifacts is input to a machine learning engine that has machine-learned to reduce metal artifacts;

acquiring a weight map in which weight coefficients are mapped, the weight map indicating a probability distribution of presence of metal artifacts;

generating a composite image, by using the weight map indicating the probability distribution of the presence of metal artifacts, to composite the machine learning output image data and the tomographic image data,

wherein for the machine learning output image data and the tomographic image data that are composited, an image quality of a region of the machine learning output image that is less affected by the metal artifacts than a corresponding region of the tomographic image that includes the metal artifacts is degraded relative to an image quality of a corresponding region in the tomographic image less affected by the metal artifacts, and an image quality of a counterpart region of the composite image corresponding to the region of the machine learning output image and the corresponding region in the tomographic image is preserved relative to an image quality of the corresponding region in the tomographic image which is composited with the machine learning output image, and

wherein the machine learning output image and the tomographic image are composited together by using a value obtained by multiplying the weight coefficient by an adjustment coefficient set in an adjustment coefficient setting portion;

displaying the composite image in a same window as the adjustment coefficient setting portion;

updating the composite image every time the adjustment coefficient is set in the adjustment coefficient setting portion; and

displaying the tomographic image and the machine learning output image in the same window.

8 . The medical image processing method according to claim 7 , wherein the weight map indicates a distribution of absolute values of differences between the tomographic image and a linear interpolation image that is obtained by applying a linear interpolation technique to the tomographic image.

9 . The medical image processing method according to claim 7 , wherein the weight coefficients become smaller with an increasing distance from a metal pixel extracted from the tomographic image.

10 . The medical image processing method according to claim 9 , wherein the weight coefficient becomes larger as the metal pixel has a larger pixel value.

11 . The medical image processing method according to claim 7 , wherein

each weight coefficient w is a real number between 0 and 1, and

the composite image is generated by compositing the machine learning output image data and the tomographic image data, by applying per-pixel weighting by w and (1−w) and combining weighted values.

12 . The medical image processing method according to claim 7 , wherein the weight map is acquired based on a beam hardening correction image that is obtained by applying a beam hardening correction to the tomographic image.

13 . The medical image processing method according to claim 7 , wherein the weight map reflects a distribution of absolute values of differences between the tomographic image and a beam hardening correction image that is obtained by applying a beam hardening correction to the tomographic image.

Assignments (2)

Priority Claims (1)

Continuity (1)

References Cited (14)