Automated detection of choroidal neovascularization (CNV)

1 . A method comprising:

receiving optical coherence tomography (OCT) imaging data for a patient eye;

collapsing the OCT imaging data about a region of interest, to form an OCT input, comprising:

flattening one or more OCT images of the OCT imaging data about a retinal pigment epithelium (RPE) layer of the patient eye;

cropping a number of pixels about the RPE layer;

wherein cropping the number of pixels about the RPE layer comprises cropping a first number of pixels above the RPE layer and a second number of pixels below the RPE layer; and

selecting a predetermined number of central B-scans as the OCT input;

analyzing the OCT input for a presence of choroidal neovascularization (CNV) disease; and

generating an output that indicates the detection of the presence of CNV disease in the patient eye using the OCT input;

wherein the output comprises a margin of error or an uncertainty number.

2 . The method of claim 1 , wherein detecting the presence of CNV disease comprises detecting retinal fluid in the patient eye using the OCT input, wherein the retinal fluid comprises at least one of intraretinal fluid, subretinal fluid, or subretinal pigment epithelial fluid.

3 . The method of claim 2 , wherein the output further comprises a value indicating the amount of retinal fluid in the patient eye.

4 . The method of claim 1 , wherein the detecting the presence of CNV disease is performed using a machine learning system including a recalibration module.

5 . The method of claim 1 , wherein the detecting the presence of CNV disease is performed via a squeeze and excitation module embedded on a machine learning system.

6 . The method of claim 1 , wherein the predetermined number of central B-scans is less than 30.

7 . The method of claim 1 , wherein the output comprises a probability value indicating the probability of the presence of CNV disease.

8 . The method of claim 1 , wherein the first number of pixels is more than twice the second number of pixels.

9 . The method of claim 1 , wherein the first number of pixels is three times more than the second number of pixels.

10 . A system, comprising:

a non-transitory memory; and

a hardware processor coupled with the non-transitory memory and configured to read instructions from the non-transitory memory to cause the system to perform operations comprising:

receiving optical coherence tomography (OCT) imaging data for a patient eye;

collapsing the OCT imaging data about a region of interest, to form an OCT input, comprising:

flattening one or more OCT images of the OCT imaging data about a retinal pigment epithelium (RPE) layer of the patient eye;

cropping a number of pixels about the RPE layer;

wherein cropping the number of pixels about the RPE layer comprises cropping a first number of pixels above the RPE layer and a second number of pixels below the RPE layer; and

selecting a predetermined number of central B-scans as the OCT input;

analyzing the OCT input for a presence of CNV disease; and

generating an output that indicates the detection of the presence of CNV disease in the patient eye using the OCT input;

wherein the output comprises a probability value indicating the probability of the presence of CNV disease and a margin of error or an uncertainty number.

11 . The system of claim 10 , wherein detecting the presence of CNV disease comprises detecting retinal fluid in the patient eye using the OCT input, wherein the retinal fluid comprises at least one of intraretinal fluid, subretinal fluid, or subretinal pigment epithelial fluid.

12 . The system of claim 10 , wherein the detecting the presence of CNV disease is performed using a machine learning system including a recalibration module.

13 . The system of claim 10 , wherein the detecting the presence of CNV disease is performed via a squeeze and excitation module embedded on a machine learning system.

14 . The system of claim 10 , wherein the first number of pixels is more than twice the second number of pixels.

15 . The system of claim 10 , wherein the first number of pixels is three times more than the second number of pixels.

16 . A non-transitory computer-readable medium having stored thereon computer-readable instructions executable to cause a computer system to perform operations comprising:

receiving optical coherence tomography (OCT) imaging data for a patient eye;

collapsing the OCT imaging data about a region of interest, to form an OCT input comprising:

flattening one or more OCT images of the OCT imaging data about a retinal pigment epithelium (RPE) layer of the patient eye;

cropping a number of pixels about the RPE layer;

wherein cropping the number of pixels about the RPE layer comprises cropping a first number of pixels above the RPE layer and a second number of pixels below the RPE layer; and

selecting a predetermined number of central B-scans as the OCT input;

analyzing the OCT input for a presence of CNV disease; and

generating an output that indicates the detection of the presence of CNV disease in the patient eye using the OCT input;

wherein the output comprises a probability value indicating the probability of the presence of CNV disease and a margin of error or an uncertainty number.

17 . The non-transitory computer-readable medium of claim 16 , wherein detecting the presence of CNV disease comprises detecting retinal fluid in the patient eye using the OCT input, wherein the retinal fluid comprises at least one of intraretinal fluid, subretinal fluid, or subretinal pigment epithelial fluid.

18 . The non-transitory computer-readable medium of claim 16 , wherein the detecting the presence of CNV disease is performed via a squeeze and excitation module embedded on a machine learning system.

19 . The non-transitory computer-readable medium of claim 16 , wherein the first number of pixels is more than twice the second number of pixels.

20 . The non-transitory computer-readable medium of claim 16 , wherein the first number of pixels is three times more than the second number of pixels.