Methods for tissue classification in cervical imagery

1. A rule-based unsupervised process for classifying cervical tissue into a diagnostic class, comprising:

collecting at least two digital color images of a patient cervix, one image of the patient cervix before the application of acetic acid to cervical tissue, and one image of the patient cervix after the application of acetic acid to cervical tissue;

determining, based on the collected digital color images, variables for a cervical tissue classification procedure, the variables including:

the size of the texture region, an opacity parameter, the size of acetowhite regions, the size of acetowhite region near os region, the size of coarse mosaics, the size of fine mosaics, the number of coarse punctations, the number of fine punctations, and the size of atypical vessels,

wherein the opacity parameter is expressed by the following formula:

Opacity

=

1

(

2

n

-

1

)

⁢

Ω

[

∑

i

,

j

⁢

(

f

k

*

⁡

(

i

,

j

)

-

g

k

⁡

(

i

,

j

)

)

p

*

r

⁡

(

i

,

j

)

]

1

p

,

where n is the number of bits of the image, f* k is the registered pre-acetic acid image and g k is the post-acetic acid image both at k band (k=1, 2, 3), r is the most opaque region extracted from the clustering algorithm in binary form, Ω is the number of foreground pixels in the opaque region r, and p is the norm metric;

serially applying, using a computer, a set of classifier rules to said cervical tissue, based on the determined variables, according to a diagnosis scheme, the classifier rules including:

determining the size of a texture region relative to the area of the cervix, determining an opacity parameter relative to one or more thresholds, and optionally including one or more classifier rules selected from the group consisting of:

determining the size of acetowhite regions relative to area of the cervix, determining the size of acetowhite regions near os region relative to area of the cervix, determining the number of coarse and fine punctations, determining the size of coarse and fine mosaics relative to area of the cervix, and determining the size of atypical blood vessels relative to a threshold; and

obtaining, based on the serial application of classifier rules according to the diagnosis scheme, a patient diagnosis classification of no evidence of disease (NED), low-grade dysplasia, high-grade dysplasia, or cancer.

2. A process according to claim 1 , wherein the rules of the diagnosis scheme include:

when said texture region is greater than approximately 1% of the area of the cervix, and the opacity parameter is less than approximately 4% of the pixel value or greater than approximately 12% of the pixel value, a diagnosis of no evidence of disease is obtained.

3. A process according to claim 1 , wherein the rules of the diagnosis scheme include:

when said texture region is greater than approximately 1% of the area of the cervix, said opacity parameter is between approximately 4% of the pixel value and approximately 12% of the pixel value, the size of the acetowhite region is greater than 7% of the area of the cervix, and the size of the acetowhite region near the os is greater than 0% of the area of the cervix, a diagnosis of high grade dysplasia is obtained.

4. A process according to claim 1 , wherein the rules of the diagnosis scheme include:

when said texture region is greater than approximately 1% of the area of the cervix, said opacity parameter is between approximately 4% of the pixel value and, approximately 12% of the pixel value, the size of the acetowhite region is less than 7% of the area of the cervix, and there is no acetowhite region near the os, a diagnosis of low grade dysplasia is obtained.

5. A process according to claim 1 , wherein the rules of the diagnosis scheme include:

when said texture region is less than approximately 1% of the area of the cervix, the opacity parameter is less than 5% of the pixel value, the number of coarse punctations is less than 2, and there are no coarse mosaic vessels, a diagnosis of low grade dysplasia is obtained.

6. A process according to claim 1 , wherein the rules of the diagnosis scheme include:

when said texture region is less than approximately 1% of the area of the cervix, the opacity parameter is greater than 10% of the pixel value, and the size of the acetowhite region near the os is greater than 15% of the area of the cervix, a diagnosis of cancer is obtained.

7. A process according to claim 1 , wherein the rules of the diagnosis scheme include:

when said texture region is less than approximately 1% of the area of the cervix, the opacity parameter is greater than 8% of the pixel value, the size of the acetowhite region near the os is greater than 0.5% of the area of the cervix, the number of coarse punctations is greater than 100, there are no coarse mosaic vessels, and the size of atypical vessels is greater than 0.5 millimeters, a diagnosis of cancer is obtained.

8. A process according to claim 1 , wherein the rules of the diagnosis scheme include:

when said texture region is less than approximately 1% of the area of the cervix, the opacity parameter is greater than 5% of the pixel value, the size of the acetowhite region near the os is less than 0.5% of the area of the cervix, the number of coarse punctation vessels is less than 100, coarse mosaic vessels are not present, and the size of atypical blood vessels is less than 0.5 mm, a diagnosis of high grade dysplasia is obtained.

9. A process according to any one of claims 1 to 8 , further comprising:

determining a confidence level of the diagnosis based on at least on the age of the patient; and

reducing the confidence level by between approximately 10% and approximately 20% if the patient is below 25 years old or over 60 years old.

10. A rule-based unsupervised process for classifying cervical tissue into a class, comprising:

collecting at least two digital color images of a patient cervix, one image of the patient cervix before the application of acetic acid to cervical tissue, and one image of the patient cervix after the application of acetic acid to cervical tissue;

determining, based on the collected digital color images, variables for a cervical tissue classification procedure, the variables including:

the size of the texture region, an opacity parameter, the size of acetowhite regions, the size of acetowhite region near os region, the size of coarse mosaics, the size of fine mosaics, the number of coarse punctations, the number of fine punctations, and the size of atypical vessels;

serially applying, using a computer, a set of classifier rules to said cervical tissue, based on the determined variables, according to a diagnosis scheme; and

obtaining a patient diagnosis classification of no evidence of disease (NED), low-grade dysplasia, high-grade dysplasia, or cancer, based on the application of classifier rules according to the diagnosis scheme, wherein the schema consists of the following classifier rules:

when said texture region is greater than approximately 1% of the area of the cervix, and the opacity parameter is less than approximately 4% of the pixel value or greater than approximately 12% of the pixel value, a diagnosis of no evidence of disease is obtained;

when said texture region is greater than approximately 1% of the area of the cervix, said opacity parameter is between approximately 4% of the pixel value and approximately 12% of the pixel value, the size of the acetowhite region is greater than 7% of the area of the cervix, and if the size of the acetowhite region near the os is greater than 0% of the area of the cervix, a diagnosis of high grade dysplasia is obtained;

when said texture region is less than approximately 1% of the area of the cervix, the opacity parameter is greater than 5% of the pixel value, the size of the acetowhite region near the os is less than 0.5% of the area of the cervix, the number of coarse punctation vessels is less than 100, coarse mosaic vessels are not present, and the size of atypical blood vessels is less than 0.5 mm, then a diagnosis of high grade dysplasia is obtained;

when said texture region is greater than approximately 1% of the area of the cervix, said opacity parameter is between approximately 4% of the pixel value and approximately 12% of the pixel value, the size of the acetowhite region is less than 7% of the area of the cervix, and there is no acetowhite region near the os, a diagnosis of low grade dysplasia is obtained;

when if when said texture region is less than approximately 1% of the area of the cervix, the opacity parameter is less than 5% of the pixel value, the number of coarse punctations is less than 2, and there are no coarse mosaic vessels, a diagnosis of low grade dysplasia is obtained;

when said texture region is less than approximately 1% of the area of the cervix, the opacity parameter is greater than 10% of the pixel value, the size of the acetowhite region near the os is greater than 15% of the area of the cervix, a diagnosis of cancer is obtained; and

when said texture region is less than approximately 1% of the area of the cervix, the opacity parameter is greater than 8% of the pixel value, the size of the acetowhite region near the os is greater than 0.5% of the area of the cervix, the number of coarse punctations is greater than 100, there are no coarse mosaic vessels, and the size of abnormal vessels is greater than 0.5 millimeters, a diagnosis of cancer is obtained.

11. A process according to claim 10 , further comprising:

determining a confidence level of the diagnosis based on at least on the age of the patient; and

reducing the confidence level by between approximately 10% and approximately 20% if the patient is below 25 years old or over 60 years old.