Image retention and stitching for minimal-flash eye disease diagnosis

Inventors

Clarida, Warren James • Amelon, Ryan Earl Rohret • Shah, Abhay • SUTHER, JACOB PATRICK • Niemeijer, Meindert • Abramoff, Michael David

Assignees

Digital Diagnostics Inc

Interested in licensing this patent?

MTEC can help explore whether this patent might be available for licensing for your application.

Publication Number

US-12322106-B2

Patent

Publication Date

2025-06-03

Expiration Date


Abstract

Systems and methods are provided herein for minimizing retinal exposure to flash during image gathering for diagnosis. In an embodiment, a system captures a plurality of retinal images of different retinal regions. The system determines that a first portion of a first image does not meet a criterion while a second portion of the first image does meet the criterion, identifies a portion of the retina depicted in the first portion that does not meet the criterion, and determines whether the portion of the retina is depicted in a third portion of a second image and whether the third portion meets the criterion. Responsive to determining that the third portion meets the criterion, the system performs the diagnosis. Responsive to determining that the portion of the retina is not depicted in the second image, the system captures an additional image of the retinal region.

Core Innovation

A method for minimizing exposure of a retina to flash during image gathering for diagnosis captures a first retinal image of a first retinal region by flashing the first retinal region and a second retinal image of a second retinal region by flashing the second retinal region, such that the first and second retinal images together form a plurality of retinal images. The method determines whether to again flash the first retinal region by determining that a first portion of the retina depicted in the first retinal image does not meet a quality criterion.

The method checks whether that first portion is depicted in a second portion of the second retinal image, and then determines whether the depiction of the first portion in the second portion meets the quality criterion. Responsive to determining that the depiction in the second portion meets the quality criterion, the method determines not to again flash the first retinal region, thereby using a usable depiction from another retinal image portion instead of re-illuminating.

The disclosed workflow further includes image sufficiency and quality criteria including over-exposure, under-exposure, shadows, artifacts, and blurriness. When the quality determination indicates a portion is not usable, the workflow evaluates whether the same retinal portion is depicted in another captured image with acceptable quality; if so, diagnosis is performed using stitched or combined portions when sufficient, and low-quality areas may be treated as discounted when forming analyses used for diagnosis confidence.

Claims Coverage

The independent claims cover a method, a computer program product, and a system that minimize flash exposure by selectively avoiding re-flashing a retinal region when a depicted retinal portion meets a quality criterion, and then using the resulting plurality of retinal images with a fully autonomous machine learning model to output a likelihood of a disease condition. Across the independent claims, the core inventive features total three main inventive feature groups.

Selective re-flashing decision based on cross-image depiction quality

Determining whether to again flash the first retinal region by determining that a first portion of the retina depicted in the first retinal image does not meet a quality criterion; responsive to determining that the first portion does not meet the quality criterion, determining whether the first portion is depicted in a second portion of the second retinal image; responsive to determining that the first portion is depicted in the second portion of the second retinal image, determining whether a depiction of the first portion in the second portion meets the quality criterion; and responsive to determining that the depiction in the second portion meets the quality criterion, determining not to again flash the first retinal region.

Fully autonomous machine learning likelihood output from plurality of retinal images

Passing the plurality of retinal images to a fully autonomous machine learning model that outputs a likelihood of a disease condition based on the plurality of retinal images.

Two-region flash image gathering forming a plurality of retinal images

Capturing a first retinal image of a first retinal region by flashing the first retinal region and capturing a second retinal image of a second retinal region by flashing the second retinal region, the first retinal image and the second retinal image together forming a plurality of retinal images.

Across the independent method, program product, and system claims, the inventive coverage centers on capturing first and second retinal region images to form a plurality of retinal images, evaluating a first retinal portion that fails a quality criterion to see whether the same portion is depicted in a second retinal image portion that meets the quality criterion, and then selectively determining not to again flash the first retinal region when the second depiction meets the criterion. The resulting plurality of retinal images are then passed to a fully autonomous machine learning model to output a likelihood of a disease condition.

Stated Advantages

Minimizing exposure of a retina to flash during image gathering for diagnosis.

Documented Applications

Flash-based retinal image gathering for diagnosis using a fully autonomous machine learning model to output a likelihood of a disease condition.

Diagnosis workflows based on overlapping retinal regions where usable depiction of a retinal portion from another retinal image portion can be used instead of re-flashing.

JOIN OUR MAILING LIST

Stay Connected with MTEC

Keep up with active and upcoming solicitations, MTEC news and other valuable information.