San Francisco science hack day project
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.

Gray Matter: Science Hack Day SF 2017 Project

Participants: Valentina Borghesani, Fabian Pedregosa, Ines Bahri

There is a rare kind of dementia that affects language more than memory, i.e., patients don't forget things, they lose knowledge of the meaning of words.

The data

We have data on ~100 of these patients: brain atrophy maps and behavioral scores from a series of neuropsychological testing.

We know clinicians tend to cluster the patients in two sub-groups: primarily left vs primarily right atrophy.

These are the two classes identified by the Neurologist at time=0

These are the two classes identified by the Neurologist at time=1

Our project

We want to cluster the patients in a data-driven way by looking at the atrophy in specific regions of interest of the brain. Then we compared:

  • our clustering with the one made by neurologists
  • our clustering at time=0 with the resuls of the same process at time=1 (second visit for each patient)
  • the behavioral performance across the 3 clusters we obtained

Our Results

Data-driven clustering suggests actually 3 clusters at time=0:

It looks like patients in blue have the worse kind of atrophy (primarily on the left side), and behaviorally they present with the most advanced stage of dementia.

Even patients in green appear to have predominantly left atrophy, yet not extended to frontal areas. Behaviorally, they are really bad at naming things.

Finally, patients in red, behaviorally really bad with social skills, show diffuse atrophy involving mostly the right side of the brain.

And it's all confirmed (of course atrophy advanced!) at time=1