Image based screening system for prediction of individual at risk of late age-related macular degeneration (AMD)

1. A method for identifying individuals at risk of progression from intermediate stage AMD to late stage AMD comprising:

receiving retinal image data and socio-demographic parameters; and

performing AMD suspect screening comprising

performing image data segmentation on the retinal image_to prepare a segmented image having prominent regions;

performing elastic registration to the segmented image to prepare a registered image comprising the locations of an AMD pathology;

generating a matrix from the registered image;

training a deep convolution neural network with matrix data and socio-demographic parameters; and

generating a fuzzy weighted score regarding risk of developing late AMD using the deep convolution neural network;

wherein performing elastic registration to the segmented image to prepare a registered image comprising the locations of the AMD pathology comprises macular area mapping; and

wherein macular area mapping comprises

identification of the optic disc;

identification of major vessel-segments about the optic disc boundary; and

identification of the retinal raphe.

2. The method of claim 1 , wherein receiving retinal image data comprises receiving color fundus (CF) and red free (RF) image data.

3. The method of claim 1 , wherein receiving retinal image data comprises receiving color fundus (CF) image data.

4. The method of claim 1 , wherein receiving retinal image data comprises receiving red free (RF) image data.

5. The method of claim 1 , wherein macular area mapping further comprises detection of the macular center and selection of a 6000 micron diameter region about the macular center.

6. The method of claim 1 , wherein generating the fuzzy weighted score comprises use of one or more of random forest, decision stump, support vector machine, or artificial neural network classifiers to predict the risk of progression from intermediate stage AMD to late stage AMD.

7. The method of claim 1 , further comprising, prior to generating a matrix from the registered image, generating a normalized image including prominent features comprising performing one or more of median filtering, image normalization and Gabor filtering.

8. The method of claim 1 , wherein training a deep convolution neural network with the matrix data and socio-demographic parameters comprises

training the neural network to classify a pixel as RPD/bright lesion or background;

mapping each of a plurality of RPD/pathology prominent regions;

analyzing the shape and size of one of the plurality of RPD/pathology prominent regions to determine if the region is soft or hard drusen.

9. The method of claim 1 , further comprising:

if the AMD suspect screening provides a score indicative of progression from intermediate stage AMD to late stage AMD, performing an AMD incidence prediction comprising:

performing machine-learning on retinal image data comprising analyzing the shape of one of the prominent regions and analyzing the size of one of the prominent regions to distinguish drusen from background image data;

performing a graph-based method on the retinal image data for drusen quantification comprising normalizing the image data, detecting seed points based on local peak intensity, and detecting an edge around each seed point corresponding to the edge of the drusen;

merging the information from the machine-learning and graph-based methods to determine the drusen regions;

providing a prediction score;

if the prediction score indicates late stage AMD providing a treatment regimen; and

if the AMD suspect screening provides a score not indicative of progression from intermediate stage AMD to late stage AMD, recommending a repetition of the AMD suspect screening at a future time.

10. The method of claim 9 , wherein detecting an edge around each seed point comprises

applying Dijkstra's shortest path algorithm and

analyzing color, intensity, and texture analysis with Gabor filter bank response.

11. A system for identifying individuals at risk of progression from intermediate stage AMD to late stage AMD, the system comprising a computer system having one or more processors comprising a server and a remote device said computer system configured by machine-readable instructions to:

receive at the server de-identified encrypted retinal image data from the remote device and socio-demographic parameters; and perform AMD suspect screening by performing the method of claim 1 ; and

transmit information from the server to the remote device regarding AMD stage, risk of progression from intermediate stage AMD to late stage AMD, and recommendation to visit an ophthalmologist immediately or at a certain time-frame.

12. The system of claim 11 , wherein the information transmitted from the server to the remote device comprises an image that is cropped in the center area for a region of interest based on image center, and wherein the image is split into a plurality of parts for transmission from the server to the remote device.

13. The system of claim 11 , further comprising:

a re-identification and decryption module at the remote device configured by machine-readable instructions to produce a report of individual's AMD stage with an image, and an individual's personal and pathological information.

14. The system of claim 11 , wherein the retinal image data comprises color fundus (CF) and red-free (RF) image data.

15. The system of claim 11 , wherein the retinal image data comprises color fundus (CF) image data.

16. The system of claim 11 , wherein the retinal image data comprises red-free (RF) image data.

17. A method for identifying individuals at risk of progression from intermediate stage AMD to late stage AMD comprising:

receiving retinal image data and socio-demographic parameters; and

performing AMD suspect screening comprising

performing image data segmentation on the retinal image to prepare a segmented image having prominent regions;

performing elastic registration to the segmented image to prepare a registered image comprising the locations of an AMD pathology;

generating a matrix from the registered image;

training a deep convolution neural network with matrix data and socio-demographic parameters; and

generating a fuzzy weighted score regarding risk of developing late AMD using the deep convolution neural network;

wherein training a deep convolution neural network with the matrix data and socio-demographic parameters comprises

training the neural network to classify a pixel as RPD/bright lesion or background;

mapping each of a plurality of RPD/pathology prominent regions; and

analyzing the shape and size of one of the plurality of RPD/pathology prominent regions to determine if the region is soft or hard drusen.