Image processing device, image processing method, medical imaging device, and program

Abstract

To estimate the contour of an area of an object in a captured image of the object while reflecting a statistical tendency of the shape of the area, and improve estimation accuracy.SOLUTION: An image processing apparatus acquires an input image obtained by imaging an object, designates a plurality of feature points related to the contour of the object in the input image, and estimates the contour of the object in the input image by using the positions of the feature points. The feature points include a fixed feature point which does not permit positional variation and a floating feature point which permits positional variation. In the contour estimation, the contour of the object is estimated under a restriction that the position of the floating feature point is varied while the position of the fixed feature point is fixed at a set position.SELECTED DRAWING: Figure 1

Claims

1. An acquisition means for acquiring a medical image of an object; a setting means for setting a plurality of feature points including fixed feature points that do not allow fluctuations in position, floating feature points that allow fluctuations in position, and a contour point group of the object based on an instruction received from a user; a normalization means for transforming the medical image together with the floating feature points into a normalized space so that the positions of the fixed feature points in the medical image have predetermined coordinate values; a learning model acquisition means for acquiring a learning model constructed by concatenating, for each of the medical images corresponding to a case, pixel value information in which pixel values ​​of the medical image in the normalized space are arranged, contour information in which coordinate values ​​of the contour point group are arranged, and floating feature point information in which coordinate values ​​of the floating feature points are arranged, as a single vector, and performing principal component analysis on vector groups corresponding to a plurality of cases; an estimation means for estimating coordinate value information of the contour point group in the normalized space in which the missing information has been interpolated by performing a matrix operation of a BPLP method in which the pixel value information and the floating feature point information in the normalized space are known information and the contour information including missing information is unknown information, and for acquiring coordinate values ​​of the contour point group in the medical image by applying an inverse transformation of the coordinate transformation to the estimated coordinate value information.

2. 2. The image processing device according to claim 1, wherein the learning model acquisition means increases the number of learning data samples by changing the positions of the correct contour points corresponding to the correct contour in the normalized space while keeping the positions of the fixed feature points fixed.

3. 3. The image processing apparatus according to claim 1, wherein the fixed feature point is a point corresponding to a specific portion on a contour of the object.

4. 4. The image processing apparatus according to claim 1, wherein the floating feature point is a point having a predetermined positional relationship with respect to the contour of the object.

5. 5. The image processing apparatus according to claim 1, wherein the floating feature point is determined based on the positions of a plurality of points on the contour of the object.

6. 6. The image processing apparatus according to claim 5, wherein the floating feature point is a center of gravity of the plurality of points.

7. 7. The image processing device according to claim 1, wherein the setting means sets up to a predetermined number of feature points specified on the medical image as the fixed feature points, and sets feature points exceeding the predetermined number as the floating feature points.

8. the setting means, when the predetermined number or less of feature points are designated, sets all the designated feature points as the fixed feature points; 8. The image processing device according to claim 7, wherein, when only the fixed feature point is set, the estimation means estimates the contour in the medical image under the constraint that the position of the fixed feature point is fixed to the position set by the setting means.

9. 9. The image processing apparatus according to claim 1, wherein the setting means sets each of the plurality of feature points to either the fixed feature point or the floating feature point based on an order in which the plurality of feature points are specified.

10. 10. The image processing device according to claim 1, wherein the setting means sets each of the plurality of feature points to either the fixed feature point or the floating feature point based on the positions of the plurality of feature points in the medical image.

11. 10. The image processing device according to claim 1, wherein the setting means determines a degree of certainty for each of the plurality of feature points based on a user operation, and determines whether each of the plurality of feature points is a fixed feature point or a floating feature point based on the degree of certainty.

12. 12. The image processing apparatus according to claim 11, wherein the setting means determines the degree of certainty based on the duration of a user operation to designate the position of each feature point.

13. An imaging means for capturing a medical image including an object; a setting means for setting a plurality of feature points including fixed feature points that do not allow fluctuations in position, floating feature points that allow fluctuations in position, and a contour point group of the object based on an instruction received from a user; a normalization means for transforming the medical image together with the floating feature points into a normalized space so that the positions of the fixed feature points in the medical image have predetermined coordinate values; a learning model acquisition means for acquiring a learning model constructed by concatenating, for each of the medical images corresponding to a case, pixel value information in which pixel values ​​of the medical image in the normalized space are arranged, contour information in which coordinate values ​​of the contour point group are arranged, and floating feature point information in which coordinate values ​​of the floating feature points are arranged, as a single vector, and performing principal component analysis on vector groups corresponding to a plurality of cases; an estimation means for estimating coordinate value information of the contour point group in the normalized space in which the missing information has been interpolated by performing a matrix operation of a BPLP method in which the pixel value information and the floating feature point information in the normalized space are known information and the contour information including missing information is unknown information, and for acquiring coordinate values ​​of the contour point group in the medical image by applying an inverse transformation of the coordinate transformation to the estimated coordinate value information.

14. An acquisition step of acquiring a medical image of an object; a setting step of setting a plurality of feature points including fixed feature points that do not allow fluctuations in position, floating feature points that allow fluctuations in position, and a contour point group of the object based on instructions received from a user; a normalization step of transforming the medical image together with the floating feature points into a normalized space so that the positions of the fixed feature points in the medical image have predetermined coordinate values; a learning model acquisition step of concatenating, for each medical image corresponding to a case, pixel value information in which pixel values ​​of the medical image in the normalized space are arranged, contour information in which coordinate values ​​of the contour point group are arranged, and floating feature point information in which coordinate values ​​of the floating feature points are arranged, as a single vector, and acquiring a learning model constructed by performing principal component analysis on vector groups corresponding to a plurality of cases; an estimation step of estimating coordinate value information of the contour point group in the normalized space in which the missing information has been interpolated by performing a matrix operation of the BPLP method in which the pixel value information and the floating feature point information in the normalized space are known information and the contour information including missing information is unknown information, and applying an inverse transformation of the coordinate transformation to the estimated coordinate value information to obtain the coordinate values ​​of the contour point group in the medical image.

15. A program for causing a computer to execute each step of the image processing method according to claim 14.

Images