Method for learning threshold value

Patent metadata

Field	Value
Publication number	US-12249066-B2
Application number	US-201917622493-A
Country	US
Kind code	B2
Filing date	Jul 12, 2019
Priority date	Jul 12, 2019
Publication date	Mar 11, 2025
Grant date	Mar 11, 2025

Abstract

The present invention has been made in view of the foregoing, and an object thereof is to provide a method for learning a threshold that can provide a more objective threshold applied to pixels in the mammography image.
The present invention provides a method for learning a threshold value applied to pixels in a mammography image comprising: an acquiring step; and a learning step, wherein in the acquiring step, the mammography image is acquired, in the learning step, a relationship between the mammography image and a mammary gland pixel estimation threshold is learned, the mammary gland pixel estimation threshold is a threshold value used to calculate a mammary gland pixel area of each pixel of a mammary gland region in the mammography image, and the mammary gland pixel area is a value indicating a degree of a mammary gland pixel-likeness of the pixel in the mammography image.

First claim

The invention claimed is:

1. A method for learning a threshold value applied to pixels in a mammography image comprising:
an acquiring step; a histogram generation step; and a learning step, wherein
in the acquiring step, the mammography image is acquired,
in the histogram generation step, first to third histograms are generated,
the first histogram is a histogram of a pixel value of each pixel in the mammography image,
the second histogram is a histogram of a pixel value of each pixel in a mammary gland region in the mammography image, and
the third histogram is a histogram of a mammary gland region probability of each pixel in the mammography image,
the mammary gland region probability indicates a probability that each pixel in the mammography image is in the mammary gland region,

in the learning step, a relationship between the first to third histograms and a mammary gland pixel estimation threshold is learned,
the mammary gland pixel estimation threshold is a threshold value used to calculate a mammary gland pixel area of each pixel of the mammary gland region in the mammography image, and
the mammary gland pixel area is a value indicating a degree of a mammary gland pixel-likeness of the pixel in the mammography image.

2. The method of claim 1 further comprising a mammary gland density acquiring step, wherein
in the acquiring step, a correct mammary gland density in the mammary gland region is further acquired,
in the mammary gland density acquiring step, a calculated mammary gland density in the mammary gland region is acquired based on the mammography image and the mammary gland pixel estimation threshold,
in the learning step, the relationship is learned based on the correct mammary gland density, and the calculated mammary gland density or the mammary gland pixel estimation threshold output in the learning step.

3. The method of claim 2 , wherein
in the learning step, a relationship between the mammography image and a mammary gland pixel estimation tilt is further learned,
the mammary gland pixel estimation threshold is a threshold of a predetermined threshold function,
the mammary gland pixel estimation tilt is a tilt of the threshold function at the mammary gland pixel estimation threshold,
the threshold function is a function that associates a pixel value of each pixel in the mammary gland region with the mammary gland pixel area of each such pixel, and
in the mammary gland density acquiring step, the mammary gland pixel area of each pixel in the mammary gland region is calculated based on the threshold function, the mammary gland pixel estimation threshold, and the mammary gland pixel estimation tilt, and the calculated mammary gland density is acquired based on the sum of the mammary gland pixel area.

4. The method of claim 1 , wherein
the mammary gland region is a narrower region than an entire breast in the mammography image.

5. A method for learning a threshold value applied to pixels in a mammography image comprising:
an acquiring step; and a learning step, wherein
in the acquiring step, the mammography image is acquired,
in the learning step, a relationship between the mammography image and a mammary gland pixel estimation threshold is learned,
the mammary gland pixel estimation threshold is a threshold value used to calculate a mammary gland pixel area of each pixel of a mammary gland region in the mammography image, and
the mammary gland pixel area is a value indicating a degree of a mammary gland pixel-likeness of the pixel in the mammography image,

the method further comprising a mammary gland density acquiring step, wherein
in the acquiring step, a correct mammary gland density in the mammary gland region is further acquired,
in the mammary gland density acquiring step, a calculated mammary gland density in the mammary gland region is acquired based on the mammography image and the mammary gland pixel estimation threshold,
in the learning step, the relationship is learned based on the correct mammary gland density, and the calculated mammary gland density or the mammary gland pixel estimation threshold output in the learning step,
in the learning step, a relationship between the mammography image and a mammary gland pixel estimation tilt is further learned,
the mammary gland pixel estimation threshold is a threshold of a predetermined threshold function,
the mammary gland pixel estimation tilt is a tilt of the threshold function at the mammary gland pixel estimation threshold,
the threshold function is a function that associates a pixel value of each pixel in the mammary gland region with the mammary gland pixel area of each such pixel, and
in the mammary gland density acquiring step, the mammary gland pixel area of each pixel in the mammary gland region is calculated based on the threshold function, the mammary gland pixel estimation threshold, and the mammary gland pixel estimation tilt, and the calculated mammary gland density is acquired based on the sum of the mammary gland pixel area.

6. The method of claim 5 , wherein
the mammary gland region is a narrower region than an entire breast in the mammography image.

Assignees

• Eizo Corp

Inventors

• Kato Yu
• Higashi Masafumi
• Aoki Reo
• Ogaki Mamoru

Classifications

• G06T2207/30068

  Mammography; Breast · CPC title

• G06T2207/20081

  Training; Learning · CPC title

• G06V10/28

  Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns · CPC title

• G06V10/758

  Involving statistics of pixels or of feature values, e.g. histogram matching · CPC title

• G06V10/50

  by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis · CPC title