Water/fat image identification method and device, and harmonization method and device

An MRI water/fat image identification method, comprising:

• obtaining, via a processor, a water/fat image pair calculated by a Dixon method, and marking the pair as a first image and a second image;
• obtaining, via the processor, an in-phase image corresponding to the water/fat image pair, and using a grayscale histogram of the in-phase image to determine an image mask;
• subjecting the image mask to image segmentation, to obtain a region image mask corresponding to each anatomical region;
• selecting, via the processor, one of the region image masks as a reference region image mask;
• using the reference region image mask to extract a first reference foreground image from the first image and a second reference foreground image from the second image;
• calculating, via the processor, grayscale histograms for the first reference foreground image and the second reference foreground image, respectively, to obtain a first reference grayscale histogram and a second reference grayscale histogram;
• subtracting, via the processor, the second reference grayscale histogram from the first reference grayscale histogram, to obtain a reference water/fat relationship graph;
• selecting, via the processor, one of the remaining region image masks as a current region image mask;
• using the current region image mask to extract a first current foreground image from the first image and a second current foreground image from the second image;
• calculating, via the processor, grayscale histograms for the first current foreground image and the second current foreground image, respectively, to obtain a first current grayscale histogram and a second current grayscale histogram;
• subtracting, via the processor, the second current grayscale histogram from the first current grayscale histogram, to obtain a current water/fat relationship graph;
• judging, via the processor, whether the reference water/fat relationship graph and the current water/fat relationship graph have the same pattern characteristics;
• keeping the positions of the first current foreground image and the second current foreground image unchanged if the reference water/fat relationship graph and the current water/fat relationship graph are judged to have the same pattern characteristics;
• replacing, via the processor, the first current foreground image and the second current foreground image if the reference water/fat relationship graph and the current water/fat relationship graph are not judged to have the same pattern characteristics;
• judging, via the processor, whether there is any unprocessed region image mask remaining;
• returning, via the processor, to the selecting of one of the remaining region image masks if it is judged that any unprocessed region image mask remains;
• searching, via the processor, if it is judged that any unprocessed region image mask does not remain, for a relatively highest peak and relatively lowest trough in the reference water/fat relationship graph obtained;
• judging, via the processor, the positional relationship of the highest peak and the lowest trough in the reference water/fat relationship graph;
• determining, via the processor, if the relatively highest peak is judged to be located behind the relatively lowest trough, the first image as being a water image and the second image as being a fat image;
• and
• determining, via the processor, if the relatively highest peak is not judged to be located behind the relatively lowest trough, the second image as being a water image and the first image as being a fat image.

An MRI water/fat image harmonization method, comprising:

• obtaining, via a processor, a water/fat image pair calculated by a Dixon method, and marking the pair as a first image and a second image;
• obtaining, via the processor, an in-phase image corresponding to the water/fat image pair, and using a grayscale histogram of the in-phase image to determine an image mask;
• subjecting the image mask to image segmentation, to obtain a region image mask corresponding to each anatomical region;
• selecting, via the processor, one of the region image masks as a reference region image mask;
• using the reference region image mask to extract a first reference foreground image from the first image and a second reference foreground image from the second image;
• calculating, via the processor, grayscale histograms for the first reference foreground image and the second reference foreground image, respectively, to obtain a first reference grayscale histogram and a second reference grayscale histogram;
• subtracting, via the processor, the second reference grayscale histogram from the first reference grayscale histogram, to obtain a reference water/fat relationship graph;
• selecting, via the processor, one of the remaining region image masks as a current region image mask;
• using the current region image mask to extract a first current foreground image from the first image and a second current foreground image from the second image;
• calculating, via the processor, grayscale histograms for the first current foreground image and the second current foreground image, respectively, to obtain a first current grayscale histogram and a second current grayscale histogram;
• subtracting, via the processor, the second current grayscale histogram from the first current grayscale histogram, to obtain a current water/fat relationship graph;
• judging, via the processor, whether the reference water/fat relationship graph and the current water/fat relationship graph have the same pattern characteristics;
• keeping, via the processor, if the reference water/fat relationship graph and the current water/fat relationship graph are judged to have the same pattern characteristics, the positions of the first current foreground image and the second current foreground image unchanged;
• replacing, via the processor, if the reference water/fat relationship graph and the current water/fat relationship graph are not judged to have the same pattern characteristics, the first current foreground image and the second current foreground image;
• judging, via the processor, whether there is any unprocessed region image mask remaining;
• returning, via the processor, to the selecting of one of the remaining region image masks if it is judged that there is an unprocessed region image mask remaining;
• and
• finishing, via the processor, if it is judged that there are no unprocessed region image masks remaining.

An MRI water/fat image identification device, comprising:

• a processor configured to:
• calculate, via a first histogram calculation module, grayscale histograms for a first image and a second image calculated by a Dixon method, respectively, to obtain a first grayscale histogram and a second grayscale histogram, wherein the first image and the second image form a water/fat image pair;
• subtract, via a first relationship curve calculation module, the second grayscale histogram from the first grayscale histogram, to obtain a water/fat relationship graph;
• and
• search, via an image type determination module, a relatively highest peak and relatively lowest trough in the water/fat relationship graph, and configured to judge the positional relationship of the relatively highest peak and the relatively lowest trough in the water/fat relationship graph, and if the relatively highest peak is judged to be located behind the relatively lowest trough, the image type determination module is configured to determine the first image as being a water image and the second image as being a fat image;
• and if the relatively highest peak is judged not to be located behind the relatively lowest trough, the image type determination module is configured to determine the second image as being a water image and the first image as being a fat image.

An MRI water/fat image identification device, comprising:

• a processor configured to:
• use, via an image mask generation module, a grayscale histogram obtained for an in-phase image corresponding to a water/fat image pair calculated by a Dixon method to determine an image mask, wherein the water/fat image pair is formed by a first image and a second image;
• subject, via an image segmentation module, the image mask to image segmentation to obtain a region image mask corresponding to each anatomical region;
• select, via a reference region selection module, one of the region image masks as a reference region image mask;
• select, via a current region selection module, a region image mask from unprocessed region image masks as a current region image mask, in the case where there are unprocessed region image masks;
• use, via a foreground image extraction module, the reference region image mask to extract a first reference foreground image from the first image and a second reference foreground image from the second image and configured to use the current region image mask to extract a first current foreground image from the first image and a second current foreground image from the second image;
• calculate, via a second histogram calculation module, grayscale histograms for the first reference foreground image and the second reference foreground image, respectively, to obtain a first reference grayscale histogram and a second reference grayscale histogram, and configured to calculate grayscale histograms for the first current foreground image and the second current foreground image, respectively, to obtain a first current grayscale histogram and a second current grayscale histogram;
• subtract, via a second relationship curve calculation module, the second reference grayscale histogram from the first reference grayscale histogram, to obtain a reference water/fat relationship graph, and configured to subtract the second current grayscale histogram from the first current grayscale histogram, to obtain a current water/fat relationship graph;
• compare, via curve pattern comparison module, the reference water/fat relationship graph with the current water/fat relationship graph to determine whether they have the same pattern characteristics, and obtain a comparison result;
• replace, via a harmonization processing module, the first current foreground image and the second current foreground image in the case where the comparison result indicates that the reference water/fat relationship graph and the current water/fat relationship graph have different pattern characteristics and, in the case where the comparison result does not indicate that the reference water/fat relationship graph and the current water/fat relationship graph have different pattern characteristics, configured to keep the positions of the first current foreground image and the second current foreground image unchanged;
• and
• search, via an image type determination module, the relatively highest peak and relatively lowest trough in the reference water/fat relationship graph obtained by the second relationship curve calculation module once the harmonization processing module has completed harmonization processing of the first image and the second image, judging the positional relationship of the relatively highest peak and relatively lowest trough in the reference water/fat relationship graph, and if the relatively highest peak is located behind the relatively lowest trough, configured to determine the first image as being a water image and the second image as being a fat image, and if the relatively highest peak is not located behind the relatively lowest trough, configured to determine the second image as being a water image and the first image as being a fat image.

An MRI water/fat image harmonization device, comprising:

• a processor configured to:
• use, via an image mask generation module, a grayscale histogram obtained for an in-phase image corresponding to a water/fat image pair calculated by a Dixon method to determine an image mask, wherein the water/fat image pair is formed by a first image and a second image;
• subject, via an image segmentation module, the image mask to image segmentation to obtain a region image mask corresponding to each anatomical region;
• select, via a reference region selection module, one of the region image masks as a reference region image mask;
• select, via a current region selection module, a region image mask from unprocessed region image masks as a current region image mask, in the case where there are unprocessed region image masks;
• use, via a foreground image extraction module, the reference region image mask to extract a first reference foreground image from the first image and a second reference foreground image from the second image, and configured to use the current region image mask to extract a first current foreground image from the first image and a second current foreground image from the second image;
• calculate, via a second histogram calculation module, grayscale histograms for the first reference foreground image and the second reference foreground image, respectively, to obtain a first reference grayscale histogram and a second reference grayscale histogram, and configured to calculate grayscale histograms for the first current foreground image and the second current foreground image, respectively, to obtain a first current grayscale histogram and a second current grayscale histogram;
• subtract, via a second relationship curve calculation module, the second reference grayscale histogram from the first reference grayscale histogram, to obtain a reference water/fat relationship graph, and configured to subtract the second current grayscale histogram from the first current grayscale histogram, to obtain a current water/fat relationship graph;
• compare, via a curve pattern comparison module, the reference water/fat relationship graph with the current water/fat relationship graph to determine whether they have the same pattern characteristics, and configured to obtain a comparison result;
• and
• subject, via a harmonization processing module, the first current foreground image and the second current foreground image to position replacement in the first image and the second image in the case where the comparison result indicates that the reference water/fat relationship graph and the current water/fat relationship graph have different pattern characteristics and, in the case where the comparison result does not indicate that the reference water/fat relationship graph and the current water/fat relationship graph have different pattern characteristics otherwise, configured to keep the positions of the first current foreground image and the second current foreground image in the first image and the second image unchanged.