The Alexandra Institute // Cases // Brainreader

Brainreader

Case

Brainreader

Faster product gave a competitive advantage

Neuroreader is a patented method that tries to solve one of the major diagnostic challenges in the western world - a safe, simple, non-invasive diagnosis of Alzheimer's disease at an early stage.

The small start-up company Brainreader offers a product Neuroreader that uses MRI scans to predict Alzheimer's disease. They wanted to optimise the speed it takes to load the 3D image of a patient's brain.

Brainreader’s product Neuroreader is a patented method that tries to solve one of the major diagnostic challenges in the western world - a safe, simple, non-invasive diagnosis of Alzheimer's disease at an early stage. Such a solution will have significant societal impact.

Based on MRI scans of the brain, the product can locate various structures in the brain, preferably the right and left hippocampus, and quantify the volume of those structures. This provides valuable information that based on an advanced statistical model can be used as input for diagnosing Alzheimer's disease.

With our competences in massive data, we made it possible for Brainreader to optimise the speed of the analysis of scanned images from an hour to 16 seconds. Now, Brainreader is a market leader in a high-demand market.

From an early stage in Brainreader’s development we have been involved as technology supplier. The collaboration began as a project under the auspices of Caretech, where the objective was to reduce the execution time of Brainreader’s technological solution at the time. For a long time, we have subsequently handled both the operation of computation servers and the technology development of more robust and accurate techniques for locating the relevant brain structures.

Read more in the article New Neuroreader Software Speeds Detection of Changes in Brain Volume; Vital to Diagnosis of Degenerative Brain Diseases.

Locating and sizing of the structures in the brain is an example of a computation task that is based on a so-called segmentation of MR images. Such a segmentation is a difficult problem and there are several approaches to the problem. In Neuroreader, the segmentation is done using a so-called atlas consisting of a MR image with hand-tracing of the relevant structures. By a so-called registration between the atlas image and a new MR image a mathematical transformation is computed that makes the atlas image look like the new MR image as much as possible. This transformation can then be used to transfer the traced structures of the atlas to the new image.

The Alexandra Institute has developed a new segmentation platform that expands this segmentation algorithm to use not just one atlas, but many different atlas images. This allows for an enhanced robustness to variations in anatomy between patients. In addition, the new platform uses e.g. the latest research methods for dividing the brain into different types of tissue as well as advanced image processing techniques for the refinement of the segmentation result.

We collaborated with Brainreader in this project.

We helped Brainreader to compute large amounts of data to give them a better product.

Can we help you?

 

 

Contact
Principal Visual Computing Specialist
Visual Computing Lab
+45 40 24 80 03
Karsten Østergaard Noe See profile
.