Retinal Image Averaging with Offset-Based Texture Preservation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Retinal imaging techniques face challenges in suppressing noise while preserving anatomical texture, particularly in low-power imaging scenarios, leading to reduced signal-to-noise ratio and difficulty in identifying boundaries between strongly scattering structures.
Innovation Solution
A method involving the processing of a sequence of retinal images to determine offsets between reference and comparison images, selecting images with offsets below a threshold to generate an averaged image that retains texture, and optionally training a machine learning algorithm using these processed images to filter noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If multiple image frames are recorded and averaged to suppress noise, then noise is reduced, but valuable low-contrast anatomical features and texture are suppressed or lost
Solution Approach 1:
The patent applies different processing treatments to different components of the image: noise is averaged out in homogeneous regions while texture features are preserved in regions where they are detected. This local differentiation allows noise suppression without losing anatomical texture information.
Solution Approach 2:
The patent performs preliminary registration and alignment of multiple image frames before averaging, ensuring that anatomical structures are properly aligned. This preliminary action prevents misalignment artifacts from being mistaken for texture and allows for more effective noise suppression.
2Object-affected harmful factors
If low-power illumination sources are used for patient safety, then patient safety is improved, but signal-to-noise ratio deteriorates
Solution Approach 1:
The patent combines multiple low-signal image frames through registration and averaging to produce a single high-signal image. By merging the information from multiple frames, the signal-to-noise ratio is improved while maintaining the low-power illumination constraint for patient safety.
Solution Approach 2:
The patent captures multiple continuous image frames in sequence, maintaining continuous illumination at low power levels. This continuous capture allows for accumulation of signal information over time without requiring high instantaneous power that would compromise patient safety.
3Object-affected harmful factors
If conventional averaging techniques are used to suppress noise, then noise is reduced, but image registration accuracy deteriorates due to misalignment
Solution Approach 1:
The patent implements an iterative registration process where each image frame is registered to a reference frame, and the registration quality is continuously evaluated. This feedback mechanism allows for correction of misalignment issues and ensures high registration accuracy before the averaging step is applied.
Solution Approach 2:
The patent employs dynamic registration parameters that can adapt to different types of motion and distortion in the image sequence. This dynamic approach allows the registration process to handle varying alignment conditions and maintain accuracy across diverse imaging scenarios.
Data Source
AI summary
A computer-implemented method of processing a sequence of images of a region of a retina to generate an averaged image of the region, comprising: determining, for each combination of a reference image selected from the sequence of images and a respective comparison image being an image from remaining images in the sequence, a respective offset between the reference image and the respective comparison image; comparing each offset with an threshold to determine whether the offset is smaller than the threshold; selecting the respective comparison image in each combination for which the respective offset has been determined to be smaller than the threshold; and using the selected comparison images to generate the averaged image of the region, wherein the threshold is such that the averaged image shows more texture features in the region of the retina than a reference averaged image generated from the images in the sequence of images.