RETINAL CAMERA WITH SELECTIVE ILLUMINATION BANDS
A technique for imaging a retina with a retinal camera includes illuminating the retina with a plurality of distinct illumination bands that are substantially exclusive of green visible light while substantially not illuminating the retina with the green visible light. A first retinal image is acquired while illuminating the retina with the distinct illumination bands and substantially not illuminating the retina with the green visible light.
1 . A method of imaging a retina with a retinal camera, the method comprising:
illuminating the retina with a plurality of distinct illumination bands that are substantially exclusive of green visible light while substantially not illuminating the retina with the green visible light; and
acquiring a first retinal image while illuminating the retina with the distinct illumination bands and substantially not illuminating the retina with the green visible light.
2 . The method of claim 1 , further comprising:
sequentially acquiring a burst of retinal images including the first retinal image while illuminating the retina with the distinct illumination bands and substantially not illuminating the retina with the green visible light.
3 . The method of claim 2 , further comprising:
after acquiring the burst of retinal images substantially devoid of illumination with the green visible light, flashing the retina with white light, including the green visible light; and
acquiring a full color image while illuminating the retina with the white light to acquire a dataset including a plurality of multiband images acquired substantially without the green visible light and the full color image acquired with the green visible light.
4 . The method of claim 3 , wherein the burst of retinal images is acquired during an alignment period when the retinal camera is not yet deemed to have achieved alignment with the retina and the full color image is acquired once alignment is deemed to have been achieved.
5 . The method of claim 4 , wherein the retina is illuminated by the plurality of distinct illumination bands substantially devoid of the green light for more than twice the time period of flashing the retina with the white light including the green visible light.
6 . The method of claim 2 , wherein sequentially acquiring the burst of retinal images while illuminating the retina with the distinct illumination bands and not illuminating the retina with the green visible light, comprises:
sequentially illuminating different combinations of the distinct illumination bands that are substantially devoid of the green visible light between instances of image acquisitions within the burst of retinal images.
7 . The method of claim 2 , wherein the plurality of distinct illumination bands comprises five distinct illumination bands substantially devoid of the green visible light.
8 . The method of claim 7 , wherein the five distinct illumination bands comprise:
a first band including a 400 nm wavelength;
a second band including a 700 nm wavelength;
a third band including a 800 nm wavelength;
a fourth band including a 900 nm wavelength; and
a fifth band including a 950 nm wavelength.
9 . The method of claim 1 , further comprising:
converting the first retinal image acquired with the distinct illumination bands to an RGB false color image using a neural network trained using full color RGB images associated with multiband images acquired with the distinct illumination bands substantially devoid of the green visible light.
10 . The method of claim 1 , further comprising:
classifying, with a machine learning (ML) classifier, whether the first retinal image is symptomatic of one or more of a plurality of diseases, the ML classifier including a neural network trained with labeled image data acquired with the plurality of distinct illumination bands substantially devoid of the green visible light.
11 . At least one non-transitory machine-readable medium having instructions stored thereon that, in response to execution, cause a retinal camera to perform operations comprising:
illuminating a retina with a plurality of distinct illumination bands that are substantially exclusive of green visible light while substantially not illuminating the retina with the green visible light; and
acquiring a first retinal image while illuminating the retina with the distinct illumination bands and substantially not illuminating the retina with the green visible light.
12 . The at least one non-transitory machine-readable medium of claim 11 , wherein the operations further comprise:
sequentially acquiring a burst of retinal images including the first retinal image while illuminating the retina with the distinct illumination bands and substantially not illuminating the retina with the green visible light.
13 . The at least one non-transitory machine-readable medium of claim 12 , wherein the operations further comprise:
after acquiring the burst of retinal images substantially devoid of illumination with the green visible light, flashing the retina with white light, including the green visible light; and
acquiring a full color image while illuminating the retina with the white light to acquire a dataset including a plurality of multiband images acquired substantially without the green visible light and the full color image acquired with the green visible light.
14 . The at least one non-transitory machine-readable medium of claim 13 , wherein the burst of retinal images is acquired during an alignment period when the retinal camera is not yet deemed to have achieved alignment with the retina and the full color image is acquired once alignment is deemed to have been achieved.
15 . The at least one non-transitory machine-readable medium of claim 14 , wherein the retina is illuminated by the plurality of distinct illumination bands substantially devoid of the green light for more than twice the time period of flashing the retina with the white light including the green visible light.
16 . The at least one non-transitory machine-readable medium of claim 12 , wherein sequentially acquiring the burst of retinal images while illuminating the retina with the distinct illumination bands and not illuminating the retina with the green visible light, comprises:
sequentially illuminating different combinations of the distinct illumination bands that are substantially devoid of the green visible light between instances of image acquisitions within the burst of retinal images.
17 . The at least one non-transitory machine-readable medium of claim 12 , wherein the plurality of distinct illumination bands comprises five distinct illumination bands substantially devoid of the green visible light.
18 . The at least one non-transitory machine-readable medium of claim 17 , wherein the five distinct illumination bands comprise:
a first band including a 400 nm wavelength;
a second band including a 700 nm wavelength;
a third band including a 800 nm wavelength;
a fourth band including a 900 nm wavelength; and
a fifth band including a 950 nm wavelength.
19 . The at least one non-transitory machine-readable medium of claim 11 , further comprising:
converting the first retinal image acquired with the distinct illumination bands to an RGB false color image using a neural network trained using full color RGB images associated with multiband images acquired with the distinct illumination bands substantially devoid of the green visible light.
20 . The at least one non-transitory machine-readable medium of claim 11 , further comprising:
classifying, with a machine learning (ML) classifier, whether the first retinal image is symptomatic of one or more of a plurality of diseases, the ML classifier including a neural network trained with labeled image data acquired with the plurality of distinct illumination bands substantially devoid of the green visible light.