System, method and device for automatic noninvasive screening for diabetes and pre-diabetes

• US 9,364,147 B2
• Filed: 02/11/2014
• Issued: 06/14/2016
• Est. Priority Date: 02/11/2013
• Status: Active Grant

• Current Assignee: Lifelens LLC
• Original Assignee: Lifelens LLC
• Inventors: Wakizaka, Jason, To, Wilson, Souaiaia, Tade
• Primary Examiner(s): LAURITZEN, AMANDA L
• Application Number:
  US14/178,163
• Publication Number:
  US 20140228668A1
• Time in Patent Office: 854 Days
• US Class Current: 1/1
• CPC Class Codes: A61B 3/12 for looking at the eye fundus, e.g. ophthalmoscopes A61B3/13 takes precedence
• CPC Class Codes: A61B 3/14 Arrangements specially adapted for eye photography apparatus or arrangements for taking photographs per se G03B
• CPC Class Codes: A61B 5/02007 Evaluating blood vessel condition, e.g. elasticity, compliance
• CPC Class Codes: A61B 5/0261 using optical means, e.g. infrared light
• CPC Class Codes: A61B 5/489 Blood vessels
• CPC Class Codes: A61B 5/7275 Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
• CPC Class Codes: G06T 7/0012 Biomedical image inspection
• CPC Class Codes: G06T 7/10 Segmentation; Edge detection motion-based segmentation G06T7/215
• CPC Class Codes: G16H 30/20 for handling medical images, e.g. DICOM, HL7 or PACS
• CPC Class Codes: G16H 30/40 for processing medical images, e.g. editing
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: G16H 50/30 for calculating health indices; for individual health risk assessment
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients

Abstract

A system for an automatic noninvasive screening for diabetes and pre-diabetes using a device to take at least one image of a patient'"'"'s eye, executing non-transitory instructions executable on a processor for analyzing the image and displaying an indication if the patient has diabetes.

Claims

1. A method for an automatic noninvasive screening for diabetes and pre-diabetes using at least one image, the method comprising the steps of:

a) capturing a color image of a patient'"'"'s eye;

b) storing the image in a storage for processing;

c) converting the color image to a grayscale image;

d) applying a box blur filter to the grayscale image;

e) applying a noise reduction filter the box blur image;

f) normalizing the noise reduced image;

g) increasing the range in the normalized, noise reduced image between white, identifying conjunctiva, and black, identifying blood vessel, pixels on the normalized image;

h) applying a Gaussian Matched filter to the range increased image;

i) scoring each pixel of the Gaussian Matched image on a likelihood of being in a blood vessel;

j) calculating, for each of the scored pixels, an optimal orthogonal angle;

k) rank segmenting each of the optimal orthogonal angles;

l) identifying blood vessel candidates from the segmented rankings using a threshold;

m) calculating a midpoint for each segmented ranking;

n) calculating midpoints for each segmented rank;

o) calculating blood vessel diameters associated with each midpoint;

p) chaining each identified midpoint to the other identified midpoints;

q) calculating a line that connects and traverses the blood vessel through the chained midpoints creating a line; and

r) performing feature analysis on the blood vessel using statistics and the blood vessel chains to identify and measure features in the micro circulation to identified diabetic conditions.

2. The method of claim 1, where the step of normalization increases the dynamic range of the image, exaggerating the difference in pixel intensity between the pixels in a blood vessel and pixels not in a blood vessel.

3. The method of claim 1, where enhancing the image using a Gaussian Matched Filter comprises the steps of:

a) analyzing, for every point P(x,y), in the image a line of pixels of some length centered on P(x,y);

b) scoring how closely the pixel intensities match a Gaussian curve, with P(x,y) acting as a mean;

c) repeating the scoring for a number of equidistant radials, centered on P(x,y); and

d) storing a highest score and an angle that produced the highest score for every point in the image.

4. The method of claim 1, where the midpoint chaining uses an O'"'"'Brien Midpoint Chaining algorithm to chain adjacent midpoints together to create a line that traverses a center of a blood vessel.

5. The method of claim 1, where only a light intensity of each pixel is stored, rather than color data for the step of converting the color image to grayscale.

6. The method of claim 5, where calculating the light intensity of each pixel is done using the formula:

Gray Value=0.114B+0.587G+0.299R, where;

B=blue channel;

G=green channel;

R=red channel.

7. The method of claim 1, the image noise is reduced by applying a box blur filter.

8. The method of claim 7, where the box blur filter comprises the steps of:

a) creating a new image, where each pixel in the new image is equal to the average value of the neighboring pixels in the original image;

b) defining neighboring pixels as a box with length k centered on the pixel; and

c) repeating step b) for a number of iterations.

9. The method of claim 8, where the length k is in the range of 1 pixel to the number of pixels in the image.

10. The method of claim 8, where the length k is five pixels.

11. The method of claim 1, where normalization comprises the steps of:

a) analyzing each pixel for a white value and a black value;

b) setting the whitest pixel to white, and the darkest pixel to black; and

c) scaling all other pixels accordingly.

12. The method of claim 11, where the step of scaling uses the formula pixel(x,y)=[(pixel(x,y)−

min)I(max−

min)]×

255, where pixel(x,y)=an individual pixel, where min is a smallest pixel intensity and max is a largest pixel intensity, respectively.

13. The method of claim 12, where the scaling is applied to an entire image, and where min is the smallest pixel intensity and max is the largest pixel intensity, respectively, in the entire image.

14. The method of claim 12, where the image is split into smaller local bins, and where min is the smallest pixel intensity and max is the largest pixel intensity, respectively, in the local bin.

15. The method of claim 3, where the score is calculated using twelve radials.

16. The method of claim 1, where the feature analysis, the method comprising the steps of:

a) identifying vessel tortuosity;

b) calculating an average, maximum, minimum, and variance of diameter for each chain to determine abnormal vessel diameter and uneven vessel thickness;

c) analyzing chains that terminate at a point not near the edge of the image;

d) identifying terminated chains as damaged blood vessels;

e) analyzing the chains identified as damaged vessels for a significant increase in diameter, a circular shape or both an increase in diameter and a circular shape at the end of the chain;

f) identifying the chains of damaged vessels as a microaneurysm;

g) analyzing chains identified as a damaged vessel for distended blood vessel, where the chains have a significant decrease in diameter at the end of the chain;

h) identifying the distended blood vessels;

i) dividing the image into a number of equal size sections;

j) analyzing the divided image for abnormal vessel distribution;

k) analyzing the divided image for ischemic sites; and

l) analyzing the image for “

Boxcar”

blood flow phenomena.

17. The method of claim 16, where the circular shape is detected with a Hough transform.

18. The method of claim 16 further comprising calculating an average diameter and variance over all chains.

19. The method of claim 16 further comprising the step of analyzing the original color image for significant areas of yellow-brown color indicating hemosiderin deposits.

20. The method of claim 16 further comprises the step of calculating vessel tortuosity.

21. The method of claim 20, where blood vessel tortuosity is calculated as being equal to a curvature of the chain divided by a length of the chain.

22. The method of claim 20, where blood vessel tortuosity is calculated as being equal to a sum of dot products of a sum of lengths.