# US 12,593,956 B2 · Method for building image reading model based on capsule endoscope, device, and medium

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 12,593,956 B2
- Document ID: US12593956B2
- Status: Granted
- Kind: B2
- Application number: 18/551,297
- Filed: 2022-03-15
- Published: 2024-06-13
- Granted: 2026-04-07
- Earliest claimed: 2021-03-19
- Assignees: Ankon Technologies Co., Ltd; Anx IP Holding Pte. Ltd.
- Applicants: Ankon Technologies Co., Ltd; Anx IP Holding Pte. Ltd.
- Inventors: Tianyi Yangdai
- Web page: https://patentgrep.com/patent/US12593956B2

## CPC

- A61B 1/00009: {of image signals during a use of endoscope}
- A61B 1/041: {Capsule endoscopes for imaging}
- A61B 1/045: Control thereof
- G06T 7/0014: {using an image reference approach}
- G06T 7/11: Region-based segmentation
- G06T 7/75: {involving models}
- G06T 15/06: Ray-tracing
- G06T 17/00: Three-dimensional [3D] modelling for computer graphics
- A61B 1/000094: {extracting biological structures}
- A61B 1/000096: {using artificial intelligence}
- A61B 1/00194: {adapted for three-dimensional imaging}
- G06T 2207/20021: Dividing image into blocks, subimages or windows
- G06T 2207/30092: Stomach; Gastric
- G06T 2207/30168: Image quality inspection
- G06T 2210/21: Collision detection, intersection
- G06T 2210/41: Medical
- G06T 2210/56: Particle system, point based geometry or rendering
- G06V 10/764: using classification, e.g. of video objects
- G06V 20/70: Labelling scene content, e.g. deriving syntactic or semantic representations
- A61B: DIAGNOSIS; SURGERY; IDENTIFICATION (usefulness limited to only animals A61D)
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

The present invention provides a method for building an image reading model based on a capsule endoscope, a device and a medium. The method includes: driving the capsule endoscope to move within a working area, sequentially recording position coordinates and field of view orientations of the capsule endoscope when it reaches each positioning point at a predetermined first frequency, and driving the capsule endoscope to sequentially capture and record images at a predetermined second frequency; constructing a 3D model corresponding to the outer contour of the working area based on the recorded position coordinates of the capsule endoscope at each positioning point; and mapping the recorded images onto the 3D model to create an image reading model. By mapping the obtained images onto a 3D model of the working area, to enhance the visualization effect of the examination, facilitate observation, save image reading time, and increase the detection efficiency.

## Claims

### Claim 1

A method for building an image reading model based on a capsule endoscope, comprising:
driving the capsule endoscope to move within a working area, sequentially recording position coordinates and field of view orientations of the capsule endoscope at each positioning point at a predetermined first frequency, and driving the capsule endoscope to sequentially capture images at a predetermined second frequency and record the images;
constructing a 3D model corresponding to an outline of the working area based on the recorded position coordinates of the capsule endoscope at each location point;
mapping the recorded images onto the 3D model to create the image reading model;
wherein constructing a 3D model corresponding to the outer contour of the working area based on the recorded position coordinates of the capsule endoscope at each positioning point comprises:
obtaining all position coordinates of the capsule endoscope to form a raw point cloud data;
generating a 3D model corresponding to the outer contour of the working area after applying Gaussian filtering, voxelization, voxel shell extraction, smoothing filtering, and surface reconstruction sequentially to the raw point cloud data;
Ω
⁡
(
p
)
=
{
1
,
p
⁢
is
⁢
on
⁢
the
⁢
outer
⁢
contour
⁢
of
⁢
the
⁢
working
⁢
area
0
,
p
⁢
is
⁢
inside
⁢
or
⁢
outside
⁢
the
⁢
circle
⁢
of
the
⁢
outer
⁢
contour
⁢
of
⁢
the
⁢
working
⁢
area
Wherein the 3D model is represented by Ω(p).

### Claim 2 (depends on 1)

The method of claim 1, wherein mapping the recorded images onto the 3D model to create an image reading model comprises:
dividing the 3D model into a plurality of sub-areas according to the structure of the working area;
mapping the recorded images to each sub-area to create a set of sub-area images, with each image mapped to a unique sub-area;
merging the set of the sub-area images on the 3D model to form the image reading model.

### Claim 3 (depends on 2)

The method of claim 2, wherein the step mapping the recorded images to each sub-area to create a set of sub-area images comprises:
iterating through each image and obtaining the positioning point with a closest capture time to a current image;
planning a virtual ray using the position coordinates of the obtained positioning point as a starting point and a corresponding field of view orientation as an extending direction, and obtaining an intersection point between the virtual ray and the 3D model;
obtaining the sub-area to which the position coordinates of current intersection point belong, and mapping the current image to the sub-area to which the position coordinates of current intersection point belong to form the set of the sub-area images.

### Claim 4 (depends on 1)

The method of claim 1, wherein the method further comprises: setting the first frequency to be higher than the second frequency.

### Claim 5 (depends on 1)

The method of claim 1, wherein, applying an interpolation filtering over a time sequence to supplement missing positioning points based on existing positioning points.

### Claim 6 (depends on 3)

The method of claim 3, further comprises:
constructing a cross-verification set;
verifying the images in the set of the each sub-area images, and/or verifying quality of the images in each set of the sub-area images;
if an image does not belong to a current set, and/or if an image quality score of the image is below a preset value, moving the current image to the cross-verification set.

### Claim 7 (depends on 3)

The method of claim 3, further comprises:
identifying and labeling an attribute for each image in each set of the sub-area images;
grouping the images with similar labels in each set of the sub-area images into one group;
generating a mapping identifier separately for each group of images with similar labels on the image reading model.

### Claim 8

An electronic device, comprising
a memory and a processor, wherein
the memory stores a computer program that runs on the processor, and
the processor executes
for building an image reading model based on a capsule endoscope, wherein the method comprises:
driving the capsule endoscope to move within a working area, sequentially recording position coordinates and field of view orientations of the capsule endoscope at each positioning point at a predetermined first frequency, and driving the capsule endoscope to sequentially capture images at a predetermined second frequency and record the images;
constructing a 3D model corresponding to an outline of the working area based on the recorded position coordinates of the capsule endoscope at each location point;
mapping the recorded images onto the 3D model to create the image reading model;
wherein constructing a 3D model corresponding to the outer contour of the working area based on the recorded position coordinates of the capsule endoscope at each positioning point comprises:
obtaining all position coordinates of the capsule endoscope to form a raw point cloud data;
generating a 3D model corresponding to the outer contour of the working area after applying Gaussian filtering, voxelization, voxel shell extraction, smoothing filtering, and surface reconstruction sequentially to the raw point cloud data;
Ω
⁡
(
p
)
=
{
1
,
p
⁢
is
⁢
on
⁢
the
⁢
outer
⁢
contour
⁢
of
⁢
the
⁢
working
⁢
area
0
,
p
⁢
is
⁢
inside
⁢
or
⁢
outside
⁢
the
⁢
circle
⁢
of
⁢
the
⁢
outer
⁢
contour
⁢
of
⁢
the
⁢
working
⁢
area
Wherein the 3D model is represented by Ω(p).