Retinal image quality assessment, error identification and automatic quality correction

Abstract

Automatically determining image quality of a machine generated image may generate a local saliency map of the image to obtain a set of unsupervised features. The image is run through a trained convolutional neural network (CNN) to extract a set of supervised features from a fully connected layer of the CNN, the image convolved with a set of learned kernels from the CNN to obtain a complementary set of supervised features. The set of unsupervised features and the complementary set of supervised features are combined, and a first decision on gradability of the image is predicted. A second decision on gradability of the image is predicted based on the set of supervised features. Whether the image is gradable is determined based on a weighted combination of the first decision and the second decision.

Description

FIELD[0001]The present application relates generally to computers and computer imaging, and more particularly to automatic image quality assessment and correction.BACKGROUND[0002]Retinal image quality assessment (IQA) is a step in screening systems for diseases like diabetic retinopathy (DR), glaucoma and age related macular degeneration (AMD) which require rapid and accurate evaluation. For example, color funds retinal image assessment is used to diagnose such diseases. Digital fundus photography of the retina is an effective non-invasive examination medium of many retinal conditions with the potential to reduce workload of ophthalmologists and increase the cost effectiveness of screening systems. Medical image quality assessment has not be much explored since many studies report a significant percentage of acquired study images to be of insufficient quality for an automated assessment. Poor quality images have to be discarded. Existing approaches to IQA use hand crafted features w...

AI Technical Summary

Benefits of technology

This patent describes a computer-based system and method for determining image quality. The system receives an image from a machine and uses both unsupervised and supervised learning to extract features from the image. These features are then combined to predict the gradability of the image based on a weighted combination of different decision scores. The method may provide a more accurate and reliable way to determine image quality.

Problems solved by technology

Medical image quality assessment has not be much explored since many studies report a significant percentage of acquired study images to be of insufficient quality for an automated assessment.

Poor quality images have to be discarded.

Existing approaches to IQA use hand crafted features which are not inclusive and do not generalize well to new datasets.

Images