Image processing device, image processing method, and image processing program for performing determination regarding diagnosis of lesion on basis of synthesized two-dimensional image and priority target region
An image processing device includes at least one processor. The processor detects a specific structural pattern indicating a lesion candidate structure for a breast in a series of a plurality of projection images obtained by performing tomosynthesis imaging on the breast or in a plurality of tomographic images obtained from the plurality of projection images, synthesizes the plurality of tomographic images to generate a synthesized two-dimensional image, specifies a priority target region, in which the specific structural pattern is present, in the synthesized two-dimensional image, and performs determination regarding a diagnosis of a lesion on the basis of the synthesized two-dimensional image and the priority target region.
1 . An image processing device comprising at least one processor,
wherein the processor:
detects a specific structural pattern indicating a lesion candidate structure for a breast in a series of a plurality of projection images obtained by performing tomosynthesis imaging on the breast or in a plurality of tomographic images obtained from the plurality of projection images,
synthesizes the plurality of tomographic images to generate a synthesized two-dimensional image,
specifies a priority target region, in which the specific structural pattern is present, in the synthesized two-dimensional image,
performs determination regarding a diagnosis of a lesion on the basis of the synthesized two-dimensional image and the priority target region,
generates a likelihood map indicating a probability that the lesion is malignant for the entire synthesized two-dimensional image,
generates a weight map in which a weight for the priority target region is larger than a weight for another region in the synthesized two-dimensional image,
performs the determination regarding the diagnosis of the lesion by multiplying a likelihood in the likelihood map by the weight in the weight map for each corresponding region, and
makes a number of performing the determination regarding the diagnosis of the lesion on the priority target region larger than that on the other region in the synthesized two-dimensional image.
2 . The image processing device according to claim 1 ,
wherein the processor focuses the determination regarding the diagnosis of the lesion more on the priority target region than on the other region in the synthesized two-dimensional image.
3 . The image processing device according to claim 1 ,
wherein the processor extracts the priority target region from the synthesized two-dimensional image and performs the determination regarding the diagnosis of the lesion on the extracted priority target region.
4 . The image processing device according to claim 3 ,
wherein the processor extracts the priority target region on the basis of a condition corresponding to a type of the specific structural pattern.
5 . The image processing device according to claim 1 ,
wherein the processor detects the specific structural pattern for a type of the specific structural pattern.
6 . The image processing device according to claim 1 ,
wherein the processor specifies a type of the specific structural pattern and specifies the priority target region for a specified type.
7 . The image processing device according to claim 1 ,
wherein the processor specifies a type of the specific structural pattern and performs the determination regarding the diagnosis of the lesion on the basis of the specified type and the priority target region.
8 . The image processing device according to claim 7 ,
wherein the processor determines whether the lesion is benign or malignant as the determination regarding the diagnosis of the lesion.
9 . The image processing device according to claim 7 ,
wherein the processor determines whether or not the specific structural pattern is a lesion as the determination regarding the diagnosis of the lesion.
10 . The image processing device according to claim 7 ,
wherein the processor determines whether or not the lesion is malignant as the determination regarding the diagnosis of the lesion.
11 . The image processing device according to claim 7 ,
wherein the processor determines whether the specific structural pattern is a benign lesion, a malignant lesion, or a structure other than a lesion as the determination regarding the diagnosis of the lesion.
12 . The image processing device according to claim 7 ,
wherein the processor determines a degree of malignancy as the determination regarding the diagnosis of the lesion.
13 . The image processing device according to claim 1 ,
wherein the processor specifies a type of the specific structural pattern using a plurality of detectors that are provided for multiple types of specific structural patterns and outputs, as a detection result, information indicating the specific structural pattern from the plurality of projection images or from the plurality of tomographic images.
14 . The image processing device according to claim 1 ,
wherein the processor detects the specific structural pattern using a detector generated by performing machine learning on a machine learning model with a geometrical structural pattern, a detector generated by performing machine learning on a mathematical model with simulation image data, or a detector generated by performing machine learning on a machine learning model using a radiographic image of the breast as training data.
15 . The image processing device according to claim 1 ,
wherein, in a case in which a size of the breast included in the plurality of tomographic images is different from a size of the breast included in the synthesized two-dimensional image, the processor performs a process of making the sizes equal to each other to specify the priority target region, in which the specific structural pattern is present, in the synthesized two-dimensional image.
16 . The image processing device according to claim 1 ,
wherein a process of specifying the priority target region, in which the specific structural pattern is present, in the synthesized two-dimensional image is incorporated into a process of synthesizing the plurality of tomographic images to generate the synthesized two-dimensional image.
17 . The image processing device according to claim 1 , wherein the processor sets a window, which scans the synthesized two-dimensional image, or a slide width of the window in the priority target region to be smaller than that in the other region in the synthesized two-dimensional image.
18 . The image processing device according to claim 1 , wherein the processor:
generates a plurality of first mask images in which a position of the specific structural pattern is shown,
generates a second mask image indicating the priority target region, by use of the plurality of first mask images, and
performs the determination regarding the diagnosis of the lesion on the basis of the synthesized two-dimensional image and the second mask image.
19 . An image processing method executed by a computer, the image processing method comprising:
detecting a specific structural pattern indicating a lesion candidate structure for a breast in a series of a plurality of projection images obtained by performing tomosynthesis imaging on the breast or in a plurality of tomographic images obtained from the plurality of projection images;
synthesizing the plurality of tomographic images to generate a synthesized two-dimensional image;
specifying a priority target region, in which the specific structural pattern is present, in the synthesized two-dimensional image;
performing determination regarding a diagnosis of a lesion on the basis of the synthesized two-dimensional image and the priority target region;
generating a likelihood map indicating a probability that the lesion is malignant for the entire synthesized two-dimensional image,
generating a weight map in which a weight for the priority target region is larger than a weight for another region in the synthesized two-dimensional image,
performing the determination regarding the diagnosis of the lesion by multiplying a likelihood in the likelihood map by the weight in the weight map for each corresponding region, and
making a number of performing the determination regarding the diagnosis of the lesion on the priority target region larger than that on the other region in the synthesized two-dimensional image.
20 . A non-transitory computer-readable storage medium storing an image processing program that causes a computer to execute a process comprising:
detecting a specific structural pattern indicating a lesion candidate structure for a breast in a series of a plurality of projection images obtained by performing tomosynthesis imaging on the breast or in a plurality of tomographic images obtained from the plurality of projection images;
synthesizing the plurality of tomographic images to generate a synthesized two-dimensional image;
specifying a priority target region, in which the specific structural pattern is present, in the synthesized two-dimensional image;
performing determination regarding a diagnosis of a lesion on the basis of the synthesized two-dimensional image and the priority target region;
generating a likelihood map indicating a probability that the lesion is malignant for the entire synthesized two-dimensional image,
generating a weight map in which a weight for the priority target region is larger than a weight for another region in the synthesized two-dimensional image,
performing the determination regarding the diagnosis of the lesion by multiplying a likelihood in the likelihood map by the weight in the weight map for each corresponding region, and
making a number of performing the determination regarding the diagnosis of the lesion on the priority target region larger than that on the other region in the synthesized two-dimensional image.