Image region localization method, image region localization apparatus, and medical image processing device

1. An image region localization method, comprising:

acquiring a plurality of three-dimensional (3D) images of a target body part of a patient, the plurality of 3D images including a first 3D image set having a first magnetic resonant imaging (MRI) modality of a plurality of MRI modalities and a second 3D image set having a second MRI modality of the plurality of MRI modalities, wherein the first MRI modality is different from the second MRI modality and the plurality of MRI modalities are selected from the group consisting of Magnetic Resonance Angiography (MRA), diffusion weighted image (DWI), apparent diffusion coefficient (ADC), fat suppression (FS) imaging, and dynamic contrast-enhanced (DCE) imaging;

registering the first 3D image set having the first MRI modality with the second 3D image set having the second MRI modality so that images in the first 3D image set and images in the second 3D image set have the same image dimensions and the same coordinate system;

after the registering, extracting first image features from the first 3D image set having the first MRI modality and extracting second image features from the second 3D image set having the second MRI modality, wherein the first image features and the second image features are different image features;

fusing the extracted first and second image features to obtain fused features having multi-modality information based on the extracted first and second image features;

determining voxel types corresponding to voxels in the 3D images according to the fused features, each voxel type corresponding to a respective pathological type;

selecting, from the 3D images, target voxels having a preset voxel type corresponding to a predefined pathological type;

obtaining position information of the target voxels; and

localizing a target region within the target body part based on the position information of the target voxels.

2. The method according to claim 1 , wherein determining the voxel types further comprises:

for each of the voxels in the 3D images:

determining a respective fused feature corresponding the voxel;

calculating a fused feature probability corresponding to the respective fused feature;

determining a probability that the voxel corresponds to a respective voxel type of the voxel types based on the calculated fused feature probability; and

determining the preset voxel type to which the voxel corresponds based on the determined probability.

3. The method according to claim 1 , wherein fusing the extracted first and second image features further comprises:

acquiring preset feature weights corresponding to the first and second image features; and

weighting the image features of the plurality of 3D images based on the preset feature weights.

4. The method according to claim 3 , wherein weighting the image features of the plurality of 3D images based on the preset feature weights further comprises:

based on the position information of the target voxels, determining a first plurality of image features corresponding to voxels having the same position in the plurality of 3D images;

weighting the first plurality of image features with the preset feature weights to obtain a plurality of weighted image features; and

accumulating the plurality of weighted image features to obtain fused features of the voxels.

5. The method according to claim 1 , wherein localizing the target region further comprises:

selecting a target 3D image from the plurality of 3D images; and

localizing the target region on the target 3D image based on the position information of the target voxels.

6. The method according to claim 5 , wherein localizing the target region further comprises:

determining corresponding voxels on the target 3D image based on the position information of the target voxels to obtain the corresponding voxels; and

setting a voxel value of each of the corresponding voxels to a preset value to identify the target region.

7. The method according to claim 5 , wherein localizing the target region further comprises:

calculating a mean of the position information of all of the target voxels to obtain position information of a central point of the target region; and

localizing the central point of the target region on the target 3D image based on the position information of the central point.

8. The method according to claim 1 , wherein extracting the image features of the plurality of 3D images comprises:

preprocessing the plurality of 3D images to obtain a plurality of preprocessed 3D images; and

extracting image features of the plurality of preprocessed 3D images.

9. The method according to claim 8 , wherein preprocessing the plurality of 3D images further comprises:

acquiring a reference coordinate system and original coordinate systems of the plurality of 3D images; and

transforming the original coordinate systems of the plurality of 3D images into the reference coordinate system.

10. The method according to claim 8 , wherein preprocessing the plurality of 3D images further comprises:

acquiring a reference voxel spacing and original voxel spacings of the plurality of 3D images; and

transforming the original voxel spacings of the plurality of 3D images into the reference voxel spacing.

11. The method according to claim 8 , wherein preprocessing the plurality of 3D images further comprises:

acquiring a reference dimension and original dimensions of the plurality of 3D images; and

transforming the original dimensions of the plurality of 3D images into the reference dimension.

12. A computing device, comprising:

one or more processors; and

memory storing one or more programs for execution by the one or more processors, the one or more programs comprising instructions for:

acquiring a plurality of three-dimensional (3D) images of a target body part of a patient, the plurality of 3D images including a first 3D image set having a first magnetic resonant imaging (MRI) modality of a plurality of MRI modalities and a second 3D image set having a second MRI modality of the plurality of MRI modalities, wherein the first MRI modality is different from the second MRI modality and the plurality of MRI modalities are selected from the group consisting of, Magnetic Resonance Angiography (MRA), diffusion weighted image (DWI), apparent diffusion coefficient (ADC), fat suppression (FS) imaging, and dynamic contrast-enhanced (DCE) imaging;

registering the first 3D image set having the first MRI modality with the second 3D image set having the second MRI modality so that images in the first 3D image set and images in the second 3D image set have the same image dimensions and the same coordinate system;

after the registering, extracting first image features from the first 3D image set having the first MRI modality and extracting second image features from the second 3D image set having the second MRI modality, wherein the first image features and the second image features are different image features;

fusing the extracted first and second image features to obtain fused features having multi-modality information based on the extracted first and second image features;

determining voxel types corresponding to voxels in the 3D images according to the fused features, each voxel type corresponding to a respective pathological type;

selecting, from the 3D images, target voxels having a preset voxel type corresponding to a predefined pathological type;

obtaining position information of the target voxels; and

localizing a target region within the target body part based on the position information of the target voxels.

13. The computing device according to claim 12 , wherein the instructions for determining the voxel types further comprise instructions for:

for each of the voxels in the 3D images:

determining a respective fused feature corresponding the voxel;

calculating a fused feature probability corresponding to the respective fused feature;

calculating a probability that the voxel corresponds to a respective voxel type of the voxel types based on the calculated fused feature probability; and

determining the preset voxel type to which the voxel corresponds based on the calculated probability.

14. The computing device according to claim 12 , wherein the instructions for fusing the extracted first and second image features further comprise instructions for:

acquiring preset feature weights corresponding to the first and second image features; and

weighting the image features of the plurality of 3D images based on the preset feature weights.

15. The computing device according to claim 14 , wherein the instructions for weighting the image features of the plurality of 3D images based on the preset feature weights further comprise instructions for:

based on the position information of the target voxels, determining a first plurality of image features corresponding to voxels having the same position in the plurality of 3D images;

weighting the first plurality of image features with the preset feature weights to obtain a plurality of weighted image features; and

accumulating the plurality of weighted image features to obtain fused features of the voxels.

16. The computing device according to claim 12 , wherein the instructions for localizing the target region further comprise instructions for:

selecting a target 3D image from the plurality of 3D images; and

localizing the target region on the target 3D image based on the position information of the target voxels.

17. A non-transitory computer-readable storage medium storing instructions that, when executed by one or more processors of a computing device, cause the one or more processors to perform operations comprising:

acquiring a plurality of three-dimensional (3D) images of a target body part of a patient, the plurality of 3D images including a first 3D image set having a first magnetic resonant imaging (MRI) modality of a plurality of MRI modalities and a second 3D image set having a second MRI modality of the plurality of MRI modalities, wherein the first MRI modality is different from the second MRI modality and the plurality of MRI modalities are selected from the group consisting of TI Weighted MRI, T2 Weighted MRI, Magnetic Resonance Angiography (MRA), diffusion weighted image (DWI), apparent diffusion coefficient (ADC), fat suppression (FS) imaging, and dynamic contrast-enhanced (DCE) imaging;

registering the first 3D image set having the first MRI modality with the second 3D image set having the second MRI modality so that images in the first 3D image set and images in the second 3D image set have the same image dimensions and the same coordinate system;

after the registering, extracting first image features from the first 3D image set having the first MRI modality and extracting second image features from the second 3D image set having the second MRI modality, wherein the first image features and the second image features are different image features;

fusing the extracted first and second image features of the plurality of 3D images to obtain fused features having multi-modality information based on the extracted first and second image features;

determining voxel types corresponding to voxels in the 3D images according to the fused features, each voxel type corresponding to a respective pathological type;

selecting, from the 3D images, target voxels having a preset voxel type corresponding to a predefined pathological type;

obtaining position information of the target voxels; and

localizing a target region within the target body part based on the position information of the target voxels.

18. The non-transitory computer-readable storage medium according to claim 17 , wherein the localizing the target region further comprises:

selecting a target 3D image from the plurality of 3D images; and

localizing the target region on the target 3D image based on the position information of the target voxels.

19. The non-transitory computer-readable storage medium according to claim 18 , wherein localizing the target region further comprises:

determining corresponding voxels on the target 3D image based on the position information of the target voxels to obtain the corresponding voxels; and

setting a voxel value of each of the corresponding voxels to a preset value to identify the target region.

20. The non-transitory computer-readable storage medium according to claim 18 , wherein localizing the target region further comprises:

calculating a mean of the position information of all of the target voxels to obtain position information of a central point of the target region; and

localizing the central point of the target region on the target 3D image based on the position information of the central point.