# US 2012/0249959 A1 · Apparatus and method for non-invasive diabetic retinopathy detection and monitoring

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

- Number: US 2012/0249959 A1
- Document ID: US20120249959A1
- Status: Published application
- Kind: A1
- Application number: 13/436,991
- Filed: 2012-04-01
- Published: 2012-10-04
- Earliest claimed: 2011-03-31
- Assignees: The Hong Kong Polytechnic University
- Inventors: Jia You; Qin Li; Zhenhua Guo; Qiang Sun; Chun Li
- Web page: https://patentgrep.com/patent/US20120249959A1

## CPC

- A61B 3/12: for looking at the eye fundus, e.g. ophthalmoscopes (A61B3/13 takes precedence)
- G06V 40/18: Eye characteristics, e.g. of the iris
- G06V 40/193: {Preprocessing; Feature extraction}
- A61B: DIAGNOSIS; SURGERY; IDENTIFICATION (usefulness limited to only animals A61D)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

A fundus camera using infrared light sources, which included an imaging optical path and an optical path for focusing and positioning and two optical paths share a common set of retina objective lens, and a computer-assisted method for retinal vessel segmentation in general and diabetic retinopathy (DR) image segmentation in particular. The method is primarily based on Multiscale Production of the Matched Filter (MPMF) scheme, which together the fundus camera, is useful for non-invasive diabetic retinopathy detection and monitoring.

## Claims

### Claim 1

A fundus camera, comprising an imaging optical path and an optical path for focusing and positioning, wherein said two optical paths share a common set of retina objective lens.

### Claim 2 (depends on 1)

The fundus camera of claim 1, wherein an infrared light source is used to minimize the disturbance to retina caused by the change of light intensity.

### Claim 3 (depends on 2)

The fundus camera of claim 2, wherein said optical path for focusing and positioning comprises objective lens, zoom lens, relay lens, cross-hair marker plate, annular aperture, condenser, and near-infrared light source.

### Claim 4 (depends on 3)

The fundus camera of claim 3, wherein said imaging optical path comprises retina objective lens, zoom lens, imaging lens, and imaging CCD, where said objective lens and zoom lens are shared with said optical path for focusing and positioning.

### Claim 5 (depends on 2)

The fundus camera of claim 2, further comprising a compensation lens group for compensating focal length variations of near- or far-sighted eyes.

### Claim 6 (depends on 5)

The fundus camera of claim 5, further comprising an ocular lens group for observing objects directly.

### Claim 7

A method of extracting multiple features at different scales of an image, comprising a process of combining Gabor filtering with scale production.

### Claim 8 (depends on 7)

The method of claim 7, wherein the Gabor filtering equation is modified to the following equation:
g
φ
,
s

(
x
,
y
)
=
exp

{
-
π

(
x
′


2
(
s


σ
x
)
2
+
y
′2
(
s


σ
y
)
2
)
}

cos

(
2

π


fx
′
)
.

### Claim 9 (depends on 7)

The method of claim 7, comprising steps of (a) coarse level detection based on multiscale correlation filtering and (b) fine level detection based on an adaptive thresholding scheme.

### Claim 10 (depends on 9)

The method of claim 9, where in step (a) a multiscale correlation filtering is used for detection of small sized red lesions in retinal images to a candidate pool and in step (b) an adaptive thresholding scheme is used to remove false red lesions from said candidate pool.

### Claim 11 (depends on 7)

The method of claim 7, comprising a first process of multiscale matched filtering and scale multiplication in an image enhancement step; and a second process of double thresholding in a classification step.

### Claim 12 (depends on 11)

The method of claim 11, wherein vessel structures in a retinal image are enhanced by multiplying the responses of matched filters at a number of scales and vessel pixels are then detected and fused in a scale production domain.

### Claim 13

A method of segmentation of color retinal images for DR diagnosis, comprising the steps of (a) erasing vessels, (b) extracting bright objects by edge detection, and (c) erasing bright objects.

### Claim 14 (depends on 13)

The method of claim 13, further comprising steps of extracting red objects by matched filtering and distinguishing vessels from red lesions using shape information, wherein said step of erasing bright object is by nearest neighbor interpolating.

### Claim 15 (depends on 13)

The method of claim 13, further comprising steps of extracting optic disk by vessel convergence and PCA, extracting bright lesion by subtracting optic disk from bright objects, and extracting fovea by selecting the candidate having a particular distance to optic disk.

### Claim 16 (depends on 15)

The method of claim 15, further comprising a step of extracting fovea candidate by 2D Gaussian filtering.
