RetinaLyze v3.2.0

RetinaLyze v3.2.0 with major changes to the Glaucoma algorithm has just been released!

Posted on
September 4, 2018
in
Product updates

We are proud to present RetinaLyze v.3.2.0. The update brings two major Quality of Life upgrades and some small tweaks, which should improve the ease and speed of use of the Glaucoma algorithm in particular.

Simplification and automation of the RetinaLyze Glaucoma workflow

The RetinaLyze Glaucoma workflow in the WebApp has now been simplified and automated. Here is a short list of the changes in this update - read more in detail below.

The algorithm no longer requires the user to click on the center of the Optic Nerve Head after initiating the initiating the Glaucoma analysis.

The algorithm no longer requires the user to manually edit the border of the Optic Nerve Head. In 99% of all analyses, the algorithm will automatically detect and set the border of the Optic Nerve head without any user intervention.

The speed of the RetinaLyze Glaucoma analysis (from first click to receival of automated response) has been greatly improved.

The number of images rejected due to saturation of the image has been vastly reduced (across all fundus cameras). This reduces the number of times a user must recapture images notably.

Before each RetinaLyze Glaucoma analysis, fundus images will now automatically be checked to make sure that the image quality is good enough and that an Optic Nerve Head is present.

Support for multiple new fundus cameras has been added.

Removal of an entire step in the workflow

The RetinaLyze Glaucoma workflow has been simplified extensively. Before this update, you had to initiate the analysis, click on the center of the ONH and set the border of the ONH to finally get the result. We're happy to tell you, that you no longer have to click on the center of the ONH and set the border of the ONH (read more below). The process will now be as simple as follows:

  1. Click to Run Glaucoma analysis
  2. Confirm the border by clicking "Yes" or "No" (read more below).
  3. Get result

Automatic AI-based optic nerve head segmentation

One of the most requested changes by our users, was to automate the manual segmentation of Optic Nerve Head, which required the user to set/edit the border of the Optic Nerve Head manually in some cases. We're happy to tell you, that this process has now been fully automated. In 99% of all analyses, the algorithm will automatically detect and set the border of the Optic Nerve head correctly without any user intervention.

The new version of the AI (neural network) has been trained using over 40,000 images. Results are much better and even cases that had issues with the current version are now being segmented correctly. We have been working hard to make sure that the segmentation works well with the Glaucoma algorithm and that it produces better or equivalent results to our current version. It now produces a consistent segmentation that is far better than our current segmentation method. Also, and not least important, the speed of the whole process has also been improved and is much faster than previously. Like all our services, we will keep improving this algorithm over time.

The area to be analysed is slightly adjusted inside the apparent edge, i.e. where the inner limit of the Elschnig Scleral Ring lies. If you still feel that the border set by the AI is incorrect, simply capture a new image and re-run the analysis.

The RetinaLyze Glaucoma algorithm now automatically detects the border of the Optic Nerve Head requiring no intervention in 99% of the cases.

New image quality and saturation test

We have introduced a new image quality and saturation test, which makes sure that the images fed to the Glaucoma algorithm are of acceptable image quality and will therefore give accurate results. The number of images rejected due to saturation of the image has also been vastly reduced (across all fundus cameras). This reduces the number of times a user must recapture images notably - yielding a speedier automated screening.

Before each RetinaLyze Glaucoma analysis, fundus images will now automatically be checked to make sure that the image quality is good enough and that an Optic Nerve Head is present in the image.

Smaller improvements

Results are now also displayed on the Detailed-page

Results of the automated analyses and manual assessments were traditionally only displayed on the Overview-page leading to users jumping back and forth between the Overview and Detailed pages to see the image and the result.

The image and the results are now shown on the Detailed-page as well to ease the job of delivering the results to the client/patient and taking action.

Results are now also displayed on the Detailed-page

Support for more cameras in the RetinaLyze Glaucoma algorithm

We have expanded the list of supported cameras for the RetinaLyze Glaucoma algorithm quite a bit since the launch (see the entire list below).
If you're interested in RetinaLyze Glaucoma, but your fundus camera is not on the list, please feel free to get in touch and we can build support for your specific camera without any cost to you.

A continuously updated list of supported cameras can be found here: https://help.retinalyze.com/general-articles/technical-details/imageinput-data-compatibility

Ganesh Ram

CCO and Co-founder of @RetinaLyze.
Productivity and UX-geek.
Passionate about making an impact.