Image diagnosis support apparatus, image diagnosis support program, and medical image acquisition apparatus including the same
The most appropriate image for a diagnostic target among a plurality of images is selected and accurate diagnosis support information is presented regardless of the type of a selected image, a modality, or the like. An image diagnosis support apparatus includes: a diagnostic information generation unit that generates diagnostic information based on a plurality of medical images; a reliability calculation unit that evaluates an image quality and calculates an image reliability for each of the plurality of medical images; and a degree-of-contribution calculation unit that calculates a degree of contribution of each of the plurality of medical images to the diagnostic information using an internal parameter indicating a degree of appropriateness of each medical image for a diagnostic target and the reliability calculated by the reliability calculation unit. An image for detection used by the diagnostic information generation unit is generated based on the degree of contribution.
1. An image diagnosis support apparatus, comprising:
a memory coupled a processor, the memory storing instructions that when executed configure the processor to:
train a machine learning algorithm using a plurality of training images to detect a tissue,
evaluate an image reliability for each of a plurality of medical images,
calculate a degree of contribution of each of the plurality of medical images to a diagnostic target based on a sum or a product of the image reliability and a degree of appropriateness of each medical image for a diagnostic target tissue, the degree of appropriateness being a predetermined value stored in a table, the degree of appropriateness selected from the table based on a type of each medical image and the diagnostic target tissue,
adjust a contrast of one or more of the plurality of medical images to be a same contrast of the training images using one or more conversion functions,
generate an image for detection by subjecting the adjusted images to weighted addition based on the calculated degree of contribution of each of the plurality of medical images, and
input the generated image for detection to the trained machine learning algorithm to detect a presence or an absence of the diagnostic target tissue in the generated image for detection as a detection result, which indicates the presence or absence of the diagnostic target tissue.
2. The image diagnosis support apparatus according to claim 1 ,
wherein the processor is configured to divide an image into a plurality of regions or portions and calculate a reliability for each region or each portion.
3. The image diagnosis support apparatus according to claim 1 ,
wherein the processor is configured to calculate the reliability based on an artifact included in the medical image or a signal noise ratio of the medical image.
4. The image diagnosis support apparatus according to claim 1 ,
wherein the processor is configured to:
select an image among the plurality of medical images having a greatest weight, and
calculate a pixel value for a part of the selected image estimated to include the diagnostic target tissue.
5. A non-transitory computer-readable medium storing an image diagnosis support program causing a computer to execute steps comprising:
training a machine learning algorithm using a plurality of training images to detect a tissue;
calculating an image reliability for each of a plurality of medical images;
calculating a degree of contribution of each of the plurality of medical images to a diagnostic target based on a sum or a product of the image reliability and a degree of appropriateness of each medical image for a diagnostic target tissue, the degree of appropriateness being a predetermined value stored in a table, the degree of appropriateness selected from the table based on a type of each medical image and the diagnostic target tissue;
adjust a contrast of one or more of the plurality of medical images to be a same contrast of the training images using one or more conversion functions,
generate an image for detection by subjecting the adjusted images to weighted addition based on the calculated degree of contribution of each of the plurality of medical images,
input the generated image for detection to the trained machine learning algorithm to detect a presence or an absence of the diagnostic target tissue in the generated image for detection, detection result, which indicates the presence or absence of the diagnostic target tissue.
6. The non-transitory computer-readable medium according to claim 5 , the steps further comprising:
selecting an image among the plurality of medical images having a greatest weight; and
calculating a pixel value for a part of the selected image estimated to include the diagnostic target tissue.
7. A medical image acquisition apparatus, comprising:
a memory coupled a processor, the memory storing instructions that when executed configure the processor to:
train a machine learning algorithm using a plurality of training images to detect a tissue,
evaluate an image quality and calculates an image reliability for each of a plurality of medical images,
calculate a degree of contribution of each of the plurality of medical images to a diagnostic target based on a sum or a product of the image reliability and a degree of appropriateness of each medical image for a diagnostic target tissue, the degree of appropriateness being a predetermined value stored in a table, the degree of appropriateness selected from the table based on a type of each medical image and the diagnostic target tissue,
adjust a contrast of one or more of the plurality of medical images to be a same contrast of training images using one or more conversion functions,
generate an image for detection by subjecting the adjusted images to weighted addition based on the calculated degree of contribution of each of the plurality of medical images, and
input the generated image for detection to the trained machine learning algorithm to detect a presence or an absence of the diagnostic target tissue in the generated image for detection, as a detection result, which indicates the presence or absence of the diagnostic target tissue.
8. The medical image acquisition apparatus according to claim 7 ,
wherein the imaging unit is a magnetic resonance imaging unit that collects a nuclear magnetic resonance signal of a subject and acquires a medical image using the nuclear magnetic resonance signal, and
wherein the plurality of medical images include at least one of a T1 weighted image, a T2 weighted image, a proton density weighted image, a FLAIR image, a magnetic susceptibility weighted image, and a diffusion weighted image as types of images.
9. The medical image acquisition apparatus according to claim 7 ,
wherein the plurality of medical images are brain images, and
the diagnostic target generation unit generates information regarding a cerebrovascular disease as the diagnostic information.
10. The medical image acquisition apparatus according to claim 7 ,
wherein the processor is configured to:
select an image among the plurality of medical images having a greatest weight, and
calculate a pixel value for a part of the selected image estimated to include the diagnostic target tissue.