DEVICE, METHOD AND RECORDING MEDIUM CONTAINING PROGRAM FOR SEPARATING IMAGE COMPONENT, AND DEVICE, METHOD AND RECORDING MEDIUM CONTAINING PROGRAM FOR GENERATING NORMAL IMAGE

• US 20080232667A1
• Filed: 03/21/2008
• Published: 09/25/2008
• Est. Priority Date: 03/22/2007
• Status: Active Grant

• Current Assignee: Fujifilm Corporation (Fujifilm Holdings Corporation)
• Original Assignee: Fujifilm Corporation (Fujifilm Holdings Corporation)
• Inventors: ITO, Wataru, KITAMURA, Yoshiro
• Granted Patent: US 8,391,576 B2
• Time in Patent Office: Days
• US Class Current: 382/132
• CPC Class Codes: G06T 2207/20016 Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
• CPC Class Codes: G06T 2207/30004 Biomedical image processing
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06V 10/20 Image preprocessing
• CPC Class Codes: G06V 2201/03 Recognition of patterns in medical or anatomical images

Abstract

A normal image representing a normal structure of a predetermined structure in an input medical image is generated with higher accuracy. Further, an abnormal component in the input medical image is separated with higher accuracy. A supervised learned filtering unit inputs an input image representing a predetermined structure to a supervised learned filter to generate an image representing a normal structure of the predetermined structure. The supervised learned filter is obtained through a learning process using supervisor images, each representing a normal structure of the predetermined structure in a subject (individual), and corresponding training images, each containing an abnormal component in the corresponding subject (individual). Further, a difference processing unit separates an abnormal component in the input image by calculating a difference between the input image and the image representing the normal structure.

Claims

1. An image component separating device comprising:

normal image generating means to generate, from an input medical image representing a predetermined structure in a subject, a normal image representing a normal structure of the structure in the subject; and

an abnormal component separating means to separate an abnormal component in the input medical image by calculating a difference between the input medical image and the normal image,

wherein the normal image generating means comprises a supervised learned filter obtained through a learning process using a plurality of training images and corresponding supervisor images, each training image representing the same structure as the predetermined structure in a subject of the same kind as the subject in the input medical image and containing an abnormal component, and each supervisor image representing a normal structure of the structure of the same subject as the subject in the corresponding training image, and the normal image generating means generates the normal image by inputting the input medical image to the supervised learned filter.

2. The image component separating device as claimed in claim 1, wherein

the supervised learned filter comprises a plurality of filters for different spatial frequency bands obtained through the learning process carried out for the respective spatial frequency bands, and

the normal image generating means generates the normal image by generating a plurality of band-limited images for the respective spatial frequency bands from the input medical image, each band-limited image representing a component for the corresponding spatial frequency band, inputting each band-limited image to corresponding one of the supervised learned filters for the spatial frequency band corresponding to the spatial frequency band of the band-limited image, and reconstructing outputted images for the respective spatial frequency bands outputted from the supervised learned filters.

3. The image component separating device as claimed in claim 1, wherein the predetermined structure is bone parts of the subject.

4. The image component separating device as claimed in claim 3, wherein the input medical image, the training images and the supervisor images are each obtained from two radiographic images by calculating a weighted sum for each combination of corresponding pixels between the two radiographic images using predetermined weighting factors, the two radiographic images being formed by two patterns of radiations having different energy distributions transmitted through a subject, and the two radiographic images representing degrees of transmission of the respective radiations through the subject.

5. The image component separating device as claimed in claim 4 further comprising:

input image generating means to generate the input medical image from two radiographic images by calculating a weighted sum for each combination of corresponding pixels between the two radiographic images using predetermined weighting factors, the two radiographic images being formed by two patterns of radiations having different energy distributions transmitted through a subject, and the two radiographic images representing degrees of transmission of the respective radiations through the subject.

6. The image component separating device as claimed in claim 1 further comprising:

image composing means to combine an abnormal component image representing the abnormal component separated by the abnormal component separating means and another image to be combined representing the same site of the same subject imaged from the same observation direction as that of the input medical image by calculating a weighted sum for each combination of corresponding pixels between the images using predetermined weighting factors.

7. The image component separating device as claimed in claim 6, wherein the other image to be combined is a soft part image representing soft parts of the subject.

8. The image component separating device as claimed in claim 7, wherein the soft part image is obtained from two radiographic images by calculating a weighted sum for each combination of corresponding pixels between the two radiographic images using predetermined weighting factors, the two radiographic images being formed by two patterns of radiations having different energy distributions transmitted through a subject, and the two radiographic images representing degrees of transmission of the respective radiations through the subject.

9. The image component separating device as claimed in claim 8, further comprising:

soft part image generating means to generate the soft part image from two radiographic images by calculating a weighted sum for each combination of corresponding pixels between the two radiographic images using predetermined weighting factors, the two radiographic images being formed by two patterns of radiations having different energy distributions transmitted through a subject, and the two radiographic images representing degrees of transmission of the respective radiations through the subject.

10. The image component separating device as claimed in claim 6, wherein the image composing means converts the color of the abnormal component in the abnormal component image into a different color from the color of the other image to be combined, and combines the converted abnormal component image and the other image.

11. The image component separating device as claimed in claim 6, wherein the image composing means applies gray-scale conversion to the abnormal component image so that the value of 0 is assigned to pixels of the abnormal component image having pixel values smaller than a predetermined threshold, and combines the converted abnormal component image and the other image.

12. The image component separating device as claimed in claim 1, further comprising:

display means to display at least one of an image containing only the abnormal component separated by the abnormal component separating means and an image in which the abnormal component is enhanced.

13. A normal image generating device comprising:

means to generate a normal image from an input medical image representing a predetermined structure in a subject by inputting the input medical image to a supervised learned filter, the supervised learned filter being obtained through a learning process using a plurality of training images and corresponding supervisor images, each training image representing the same structure as the predetermined structure in a subject of the same kind as the subject in the input medical image and containing an abnormal component, and each supervisor image representing a normal structure of the structure of the same subject as the subject in the corresponding training image.

14. The normal image generating device as claimed in claim 13, wherein

the supervised learned filter comprises a plurality of filters for different spatial frequency bands obtained through the learning carried out for the respective spatial frequency bands, and

the normal image generating means generates the normal image by generating a plurality of band-limited images for the respective spatial frequency bands from the input medical image, each band-limited image representing a component for the corresponding spatial frequency band, inputting each band-limited image to corresponding one of the supervised learned filters for the spatial frequency band corresponding to the spatial frequency band of the band-limited image, and reconstructing outputted images for the respective spatial frequency bands outputted from the supervised learned filters.

15. An image component separating method comprising:

a first step to generate, from an input medical image representing a predetermined structure in a subject, a normal image representing a normal structure of the structure in the subject; and

a second step to separate an abnormal component in the input medical image by calculating a difference between the input medical image and the normal image,

wherein the normal image is generated in the first step by inputting the input medical image to a supervised learned filter, the supervised learned filter being obtained through a learning process using a plurality of training images and corresponding supervisor images, each training image representing the same structure as the predetermined structure in a subject of the same kind as the subject in the input medical image and containing an abnormal component, and each supervisor image representing a normal structure of the structure of the same subject as the subject in the corresponding training image.

16. A normal image generating method comprising:

generating a normal image from an input medical image representing a predetermined structure in a subject by inputting the input medical image to a supervised learned filter, the supervised learned filter being obtained through a learning process using a plurality of training images and corresponding supervisor images, each training image representing the same structure as the predetermined structure in a subject of the same kind as the subject in the input medical image and containing an abnormal component, and each supervisor image representing a normal structure of the structure of the same subject as the subject in the corresponding training image.

17. A recording medium containing an image component separating program to cause a computer to carry out:

a first process to generate, from an input medical image representing a predetermined structure in a subject, a normal image representing a normal structure of the structure in the subject; and

a second process to separate an abnormal component in the input medical image by calculating a difference between the input medical image and the normal image,

wherein the first process generates the normal image by inputting the input medical image to a supervised learned filter, the supervised learned filter being obtained through a learning process using a plurality of training images and corresponding supervisor images, each training image representing the same structure as the predetermined structure in a subject of the same kind as the subject in the input medical image and containing an abnormal component, and each supervisor image representing a normal structure of the structure of the same subject as the subject in the corresponding training image.

18. A recording medium containing a normal image generating program to cause a computer to carry out:

a process to generate a normal image from an input medical image representing a predetermined structure in a subject by inputting the input medical image to a supervised learned filter, the supervised learned filter being obtained through a learning process using a plurality of training images and corresponding supervisor images, each training image representing the same structure as the predetermined structure in a subject of the same kind as the subject in the input medical image and containing an abnormal component, and each supervisor image representing a normal structure of the structure of the same subject as the subject in the corresponding training image.