Thinking Inside the ChBOX -- Delivering Practical Computing Easily to Neuroimaging Reseachers

2255 

Abstract Submission 

Rudolph Pienaar¹, Jennings Zhang², Raghuram Banda³, Ellen Grant⁴

¹Boston Children's Hospital, Harvard Medical School, Boston, MA, ²Boston Children's Hospital, Boston, MA, ³Red Hat, Inc, Raleigh, NC, ⁴Division of Newborn Medicine, Boston Children’s Hospital and Harvard Medical School, Boston, MA

Rudolph Pienaar  
Boston Children's Hospital, Harvard Medical School
Boston, MA

Jennings Zhang  
Boston Children's Hospital
Boston, MA

Raghuram Banda  
Red Hat, Inc
Raleigh, NC

Ellen Grant  
Division of Newborn Medicine, Boston Children’s Hospital and Harvard Medical School
Boston, MA

ChBOX is a novel concept in research: make the setup of a computer suitable for research coding drop dead simple. Simply insert a USB, reboot, and research. Think of it as a Live Linux distribution for neuroscience. By booting off the USB, into "Live" mode, any existing computer can be simply repurposed as a ChRIS system -- without changing the existing OS. Alternatively, the USB can install ChRIS natively on the computer. ChRIS is an open source infrastructure allowing researchers an easy on ramp to cloud-based computing. Built on pervasive containerization, ChRIS is developer, not infrastructure, focused. In many cases need only a thin python wrapper to bring existing algorithms into ChRIS and the cloud.

Neuroimaging researchers are fundamentally reliant on computing, not only as users but also as creators. As the computing needs of the field grow, the startup "cost" to doing research is more and complex. Contributing to research programming requires a fair amount of setup often including python, jupyter notebook, containerization, etc. Moreover, the ever increasing needs of AI requiring researchers to run their code on cloud clusters and to become familiar with cloud-computing concepts such as containerization and clusters such as kubernetes and Openshift.

Unsurprisingly, many researchers are less interested in mastering these technologies, leaving a pressing need for simple turnkey solutions. This is especially pressing in developing communities where researchers might have little experience in configuring their own neuroimaging platforms.

In this work we present progress on "ChRIS in a Box" or simply, "ChBOX". As the name suggests, "ChBOX" delivers the ChRIS Research Integration System to any computing device. ChRIS lets researchers work with folders and files when they code, not volumes and REST APIs. Behind the scenes, ChRIS interfaces with cloud technologies and translates down-to-earth workstation-based approaches to in the sky cloudiness without needing the developer to do so, thus greatly reducing the time to deployment and usage of research software. Once in ChRIS, the platform can interface with cloud backends seamlessly allowing researchers to run their applications on many environments without needing to learn new cloud or cluster technologies.

"ChBOX" is a Linux-based bootstrapping system based off Ansible. When inserted into a new computer and booted, a user can choose to provision the machine with a ChRIS instance. ChBOX will install a basic Fedora Linux setup on the computer and then install ChRIS. The provisioning of the system is all performed using "software as a platform" with github as the default source provider. By simply using a CLI update command, the ChRIS on the Box will update to latest versions of all available research plugins (from a central ChRIS repository -- tools such as FreeSurfer, image anonymization, image conversions, etc).

In addition to installing ChRIS and components, ChBOX will also provision the machine with basic developer tools such as the open source version of Visual Studio Code, latest python versions, and a system login account. Infrastructure such as podman and podman-desktop (for containerization) is also provided.

Once prepared a user can easily add new analysis programs to their ChBOX by simply adding new information to a reference github and rerunning the ChBOX update, or they can connect to any other ChBOX via the web and synchronize applications. They can of course also develop new code on-box and register this to their ChBOX ChRIS, making it thus available to other ChBOX instances if desired.

We hope to demonstrate provisioning ChBOX on a new computer and then show how to further update and install neuroimaging tools live. We also hope to show how solutions such as ChBOX can enhance collaboration and also democratize research by affording simple, easy, and powerful infrastructure based on open source and open science principles to any party engaged in the field.

Databasing and Data Sharing

Workflows ²

Informatics Other ¹

Computing

Data analysis

Open Data

Open-Source Code

Open-Source Software

Workflows