Predicting blindness

Seeing what ophthalmologists can't see, DELL Technologies




Voxeleron LLC, a company specialising in the development of Machine Vision and Machine Learning software.


AI, CNN, Dell Precision workstations and NVIDIA GPUs.

Need addressed and benefit provided

Today, incurable age-related macular degeneration (AMD) affects nearly 200 million people worldwide, making it the leading cause of blindness in people aged 60 and older. The disease has two stages: an early dry stage and an advanced wet stage when vision loss and blindness can occur rapidly. Currently, ophthalmologists use optical coherence tomography (OCT) to generate 3D retinal images to diagnose and monitor the disease through regular examinations to detect changes.

The goal was to help ophthalmologists predict the likelihood of a patient moving from the dry to the wet stage. To achieve this, 3D data sets need to be processed in a deep learning convolutional neural network model. By applying artificial intelligence models and using Dell Precision 7920 Tower workstations with NVIDIA Quadro GV100 GPUs, it was possible to save up to three months of time running the models.


Voxeleron is expanding the diagnostic horizons of ophthalmology with image analysis based on artificial intelligence (AI) models trained on Dell Precision workstations with NVIDIA GPUs. Voxeleron, aims to improve ophthalmologists’ ability to predict the likelihood that patients with dry-stage AMD will progress to the wet stage. The 3D retinal images may contain useful clues that can be discerned by artificial intelligence (AI) in the form of a deep learning convolutional neural network (CNN) model.

In addition, doctors can potentially use their 3D retinal imaging tools to diagnose neurological disorders as well. While the retina may be located at the back of the eye, it is actually at the front of the brain and is a window to the central nervous system. It is changing the way AI can be applied, particularly to tasks that humans are not so good at, such as finding complex patterns in large data sets.