IP Library Granted Patent US 12712077
Granted Patent B2
US 12712077 · App. 17/718,898 · Granted Aug 18, 2026

Information processing method, medical image diagnostic apparatus, and information processing system

Inventors: Yujie Lu (Vernon Hills, IL); Liang Cai (Vernon Hills, IL); Ting Xia (Vernon Hills, IL); Jian Zhou (Vernon Hills, IL); Zhou Yu (Vernon Hills, IL)
Assignee: CANON KABUSHIKI KAISHA
G16H50/20A61B6/032A61B6/12A61B6/5258G06N3/08G06T7/0012G06T7/11G06T7/62G06T12/30G16H30/40G06T2207/10081G06T2207/20081G06T2207/20084G06T2207/30052
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12712077
App. No.
17/718,898
Granted
Aug 18, 2026
Kind
B2
Abstract

A method of processing information acquired by imaging performed by a medical image diagnostic apparatus, the method including but not limited to at least one of (A) acquiring a training image volume including at least one three-dimensional object having an embedded three-dimensional feature having a first cross-sectional area in a first three-dimensional plane; selecting a second cross-sectional area in a second three-dimensional plane containing the embedded three-dimensional feature, wherein the second cross-sectional area is larger than the first cross-sectional area; and training an untrained neural network with an image of the second cross-sectional area generated from the training image volume; and (B) acquiring a first set of training data; determining a first distribution of tissue density information from the first set of training data; generating from the first set of training data a second set of training data by performing at least one of a tissue-density shifting process and a tissue-density scaling process; and training an untrained neural network with the first and second sets of training data to obtain a trained neural network.

Claims (19)

1 . An information processing method for information acquired by imaging performed by a medical image diagnostic apparatus, the information processing method comprising:

acquiring a training image volume including at least one three-dimensional object having an embedded three-dimensional feature having a first cross-sectional area in a first three-dimensional plane;

determining, by rotating the first three-dimensional plane, a second three-dimensional plane different from the first three-dimensional plane, and containing the embedded three-dimensional feature so that a second cross-sectional area of the embedded three-dimensional feature in the second three-dimensional plane is larger than the first cross-sectional area;

generating, from the acquired training image volume, a first image along the first three-dimensional plane and a second image along the second three-dimensional plane; and

training an untrained neural network for each part of a plurality of body parts with the first and second images generated from the acquired training image volume.

2 . The method according to claim 1 , wherein the first three-dimensional plane is orthogonal to the second three-dimensional plane.

3 . The method according to claim 1 , wherein the second three-dimensional plane is determined to provide a maximum cross-sectional area of the embedded three-dimensional feature.

4 . The method according to claim 1 , wherein the embedded three-dimensional feature is a stent.

5 . The method according to claim 1 , wherein the training image volume comprises image data reconstructed from CT projection data.

6 . An apparatus for an information processing method for information acquired by imaging performed by a medical image diagnostic apparatus, comprising:

processing circuitry configured to:

acquire a training image volume including at least one three-dimensional object having an embedded three-dimensional feature having a first cross-sectional area in a first three-dimensional plane;

determine, by rotating the first three-dimensional plane, a second three-dimensional plane different from the first three-dimensional plane, and containing the embedded three-dimensional feature so that a second cross-sectional area of the embedded three-dimensional feature in the second three-dimensional plane is larger than the first cross-sectional area;

generate, from the acquired training image volume, a first image along the first three-dimensional plane and a second image along the second three-dimensional plane; and

train an untrained neural network for each part of a plurality of body parts with the first and second images generated from the acquired training image volume.

7 . The apparatus according to claim 6 , wherein the first three-dimensional plane is orthogonal to the second three-dimensional plane.

8 . The apparatus according to claim 6 , wherein the second three-dimensional plane is determined to provide a maximum cross-sectional area of the embedded three-dimensional feature.

9 . The apparatus according to claim 6 , wherein the embedded three-dimensional feature is a stent.

10 . The apparatus according to claim 6 , wherein the training image volume comprises image data reconstructed from CT projection data.