Retina vessel measurement

Abstract

Disclosed is a method for training a neural network to quantify the vessel caliber of retina fundus images. The method involves receiving a plurality of fundus images; pre-processing the fundus images to normalize images features of the fundus images; and training a multi-layer neural network, the neural network comprising a convolutional unit, multiple dense blocks alternating with transition units for down-sampling image features determined by the neural network, and a fully-connected unit, wherein each dense block comprises a series of cAdd units packed with multiple convolutions, and each transition layer comprises a convolution with pooling.

Description

technical field [0001] The present invention relates to a deep learning system for automatic retinal vascular measurements from fundus photographs. Background technique [0002] Clinical studies have shown that changes in retinal vascular structure are early warnings of underlying cardiovascular disease (CVD) and other conditions, such as dementia and diabetes. This is because the condition of retinal arterioles and venules mirrors the condition of blood vessels in other parts of the body. [0003] Currently, scoring of retinal photographs by human evaluators is challenged by implementation issues, evaluator availability and training, and long-term financial sustainability. Deep learning systems (DLS) have been proposed as an option for large-scale analysis of retinal images. DLS utilizes artificial intelligence and representation-learning methods to process natural raw data to identify complex structures in high-dimensional information. In contrast to traditional pattern...

AI Technical Summary

Problems solved by technology

There are also problems when looking to train a neural network to ensure that errors are backpropagated through the neural network - such as the vanishing gradient problem

Images