gerard sanroma

gerard sanroma

machine learning, neuroimaging and computer vision

Posts by Tags

ANTs

FreeSurfer

bullseye parcellation of the cerebral white matter with FreeSurfer

8 minute read

Published:

In this post I will explain the nits and grits on how to create a bullseye parcellation of the cerebral white matter using (part of) the FreeSurfer outputs and commands. The software package recreating the parcellations according to this post is freely-available in this github repository. So, let’s get started!

offscreen rendering with pysurfer

3 minute read

Published:

Pysurfer is a Python package to display brain cortical surfaces with color overlays. In its most common configuration, it needs a working graphics card and physical display to generate the graphics via OpenGL. Therefore, you need to do some tweaking if you want to use in a remote (headless) server. In this post, I explain how to set up offscreen rendering with Pysurfer. A docker file reproducing all the steps in detail is available here.

career

transitioning from academia to industry

13 minute read

Published:

I am very excited to soon start an industry position after the end of my post-doc period. In this post I want to share my experience and give tips to people in a similar situation thinking as well on changing to industry.

machine learning

can we predict cognitive abilities from brain imaging data using conventional (unpenalized) regression ?

8 minute read

Published:

This post summarizes the talk that I gave to the study technicians of the Rhineland Study explaining the challenges in predicting cognitive abilities from high-dimensional brain imaging data using conventional (unpenalized) regression, but the ideas here explained apply generally to the modeling of relationships between any input and output data in high dimensions. Conventional regression techniques are not useful in high-dimensional settings. I show how machine learning provides an explanation for this issue and proposes a solution.

neuroimaging

bullseye parcellation of the cerebral white matter with FreeSurfer

8 minute read

Published:

In this post I will explain the nits and grits on how to create a bullseye parcellation of the cerebral white matter using (part of) the FreeSurfer outputs and commands. The software package recreating the parcellations according to this post is freely-available in this github repository. So, let’s get started!

can we predict cognitive abilities from brain imaging data using conventional (unpenalized) regression ?

8 minute read

Published:

This post summarizes the talk that I gave to the study technicians of the Rhineland Study explaining the challenges in predicting cognitive abilities from high-dimensional brain imaging data using conventional (unpenalized) regression, but the ideas here explained apply generally to the modeling of relationships between any input and output data in high dimensions. Conventional regression techniques are not useful in high-dimensional settings. I show how machine learning provides an explanation for this issue and proposes a solution.

offscreen rendering with pysurfer

3 minute read

Published:

Pysurfer is a Python package to display brain cortical surfaces with color overlays. In its most common configuration, it needs a working graphics card and physical display to generate the graphics via OpenGL. Therefore, you need to do some tweaking if you want to use in a remote (headless) server. In this post, I explain how to set up offscreen rendering with Pysurfer. A docker file reproducing all the steps in detail is available here.

visualization

offscreen rendering with pysurfer

3 minute read

Published:

Pysurfer is a Python package to display brain cortical surfaces with color overlays. In its most common configuration, it needs a working graphics card and physical display to generate the graphics via OpenGL. Therefore, you need to do some tweaking if you want to use in a remote (headless) server. In this post, I explain how to set up offscreen rendering with Pysurfer. A docker file reproducing all the steps in detail is available here.