Systems and methods for consistent presentation of medical images using deep neural networks

Patent metadata

Field	Value
Publication number	US-11615508-B2
Application number	US-202016785283-A
Country	US
Kind code	B2
Filing date	Feb 7, 2020
Priority date	Feb 7, 2020
Publication date	Mar 28, 2023
Grant date	Mar 28, 2023

Abstract

A method for automatic selection of display settings for a medical image is provided. The method includes receiving a medical image, mapping the medical image to an appearance classification cell of an appearance classification matrix using a trained deep neural network, selecting a first WW and a first WC for the medical image based on the appearance classification and a target appearance classification, adjusting the first WW and the first WC based on user preferences to produce a second WW and a second WC, and displaying the medical image with the second WW and the second WC via a display device.

First claim

The invention claimed is:

1. A method comprising:
generating an appearance classification matrix;
generating a plurality of training data pairs using the appearance classification matrix, wherein the plurality of training data pairs comprise a plurality of training images and a plurality of ground truth appearance classification cells corresponding to the plurality of training images;
training a deep neural network using the plurality of training data pairs to produce a trained deep neural network;
receiving a medical image;
mapping the medical image to an appearance classification cell of the appearance classification matrix using the trained deep neural network;
selecting a first window-width (WW) and a first window-center (WC) for the medical image based on the appearance classification and a target appearance classification;
adjusting the first WW and the first WC based on user preferences to produce a second WW and a second WC; and
displaying the medical image with the second WW and the second WC via a display device;
wherein generating the appearance classification matrix comprises:
receiving a plurality of medical images;
adjusting a first plurality of WWs and a first plurality of WCs of the plurality of medical images to produce a first plurality of appearance normalized medical images;
averaging the first plurality of WWs to produce an average WW;
averaging the first plurality of WCs to produce an average WC;
assigning the average WW and the average WC to the target appearance classification of the appearance classification matrix; and
generating a plurality of appearance classification cells around the target appearance classification based on the average WW and the average WC.

2. The method of claim 1 , wherein generating the plurality of training data pairs using the appearance classification matrix comprises:
selecting an image;
pairing the image to a ground truth appearance classification cell of the appearance classification matrix;
adjusting a WW and a WC of the image based on the ground truth appearance classification cell to produce a training image matching appearance characteristics of the ground truth appearance classification cell; and
storing the training image and the ground truth appearance classification cell corresponding to the training image as a training data pair in non-transitory memory.

3. The method of claim 1 , wherein selecting the first WW and the first WC for the medical image based on the appearance classification and the target appearance classification comprises iteratively selecting WW values and WC values until the appearance classification of the medical image matches the target appearance classification.

4. The method of claim 1 , wherein the appearance classification cell includes a WW difference between the appearance classification cell and the target appearance classification, and wherein the appearance classification cell further includes a WC difference between the appearance classification cell and the target appearance classification.

5. The method of claim 4 , wherein selecting the first WW and the first WC for the medical image based on the appearance classification and the target appearance classification comprises:
selecting the first WW based on the WW difference; and
selecting the first WC based on the WC difference.

6. A method comprising:
generating an appearance classification matrix;
generating a plurality of training data pairs using the appearance classification matrix, wherein the plurality of training data pairs comprise a plurality of training images and a plurality of ground truth appearance classification cells corresponding to the plurality of training images;
training a deep neural network using the plurality of training data pairs to produce a trained deep neural network;
receiving a medical image;
mapping the medical image to an appearance classification cell of the appearance classification matrix using the trained deep neural network;
selecting a first window-width (WW) and a first window-center (WC) for the medical image based on the appearance classification and a target appearance classification;
adjusting the first WW and the first WC based on user preferences to produce a second WW and a second WC; and
displaying the medical image with the second WW and the second WC via a display device;
wherein generating the plurality of appearance classification cells around the target appearance classification based on the average WW and the average WC comprises:
determining a position of a first appearance classification cell relative to the target appearance classification;
adjusting the average WW based on the position to produce a first appearance classification WW; and
adjusting the average WC based on the position to produce a first appearance classification WC.

7. The method of claim 6 , wherein generating the plurality of training data pairs using the appearance classification matrix comprises:
selecting an image;
pairing the image to a ground truth appearance classification cell of the appearance classification matrix;
adjusting a WW and a WC of the image based on the ground truth appearance classification cell to produce a training image matching appearance characteristics of the ground truth appearance classification cell; and
storing the training image and the ground truth appearance classification cell corresponding to the training image as a training data pair in non-transitory memory.

8. The method of claim 6 , wherein selecting the first WW and the first WC for the medical image based on the appearance classification and the target appearance classification comprises iteratively selecting WW values and WC values until the appearance classification of the medical image matches the target appearance classification.

9. The method of claim 6 , wherein the appearance classification cell includes a WW difference between the appearance classification cell and the target appearance classification, and wherein the appearance classification cell further includes a WC difference between the appearance classification cell and the target appearance classification.

10. The method of claim 9 , wherein selecting the first WW and the first WC for the medical image based on the appearance classification and the target appearance classification comprises:
selecting the first WW based on the WW difference; and
selecting the first WC based on the WC difference.

Assignees

• Ge Prec Healthcare Llc

Inventors

• Vera-Gonzalez German Guillermo
• Akram Najib
• Xue Ping
• Zhang Fengchao
• Rao Gireesha Chinthamani
• Wanek Justin

Classifications

• G06T12/30

  Image post-processing, e.g. metal artefact correction · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G16H50/20

  for computer-aided diagnosis, e.g. based on medical expert systems · CPC title

• G06T7/0012 Primary

  Biomedical image inspection · CPC title