From e894aacfa1e13c7a6ede63886b010ad338d69502 Mon Sep 17 00:00:00 2001 From: Tristan Glatard Date: Tue, 10 May 2022 07:50:41 -0400 Subject: [PATCH] Comments in intro --- paper/sea-neuro/paper-neuro.tex | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) diff --git a/paper/sea-neuro/paper-neuro.tex b/paper/sea-neuro/paper-neuro.tex index e0a4751..eb388ff 100644 --- a/paper/sea-neuro/paper-neuro.tex +++ b/paper/sea-neuro/paper-neuro.tex @@ -7,6 +7,10 @@ \usepackage{graphicx} \usepackage{caption} \usepackage{subcaption} +\usepackage{xspace} + +\newcommand{\TG}[1]{\color{red}\textsc{From Tristan}: #1\xspace\color{black}} + \linenumbers @@ -55,11 +59,11 @@ \section{Introduction}\label{sec:sea_neuro:introduction} The recent explosion in publicly available neuroimaging data has lead to new data - management challenges, from storage infrastructure application processing times. To meet the storage, accessibility and - security demands of neuroimaging data, large datasets have been stored in object stores provided by cloud storage providers. + management challenges, from storage infrastructure application processing times \TG{missing word?}. To meet the storage, accessibility and + security demands of neuroimaging data, large datasets have been stored in object stores provided by cloud storage providers \TG{mentioning clouds so early in the paper may give the impression that the paper will focus on them.}. Standardized metadata formats, such as the Brain Imaging Data Standards (BIDS)\cite{bids}, have been implemented to facilitate the sharing of the datasets. Tools such as DataLad~\cite{datalad} have been developed to provide versioning and provenance capture of data. - Furthermore, recent developments in neuroimaging pipelines have addressed computation time limitations by adopting machine-learning approaches. + Furthermore, recent developments in neuroimaging pipelines have addressed computation time limitations by adopting machine-learning approaches \TG{cite fastsurfer, hd-bet}. While all these solutions to Big Data-related data management exist, certain aspects, such as processing-related data-transfer overheads have received limited attention. @@ -98,7 +102,7 @@ Newer neuroimaging applications leveraging popular neuroimaging pipeline engines also do not benefit from processing-related data management. Although engines such as Nipype~\cite{nipype} and - Joblib~\cite{joblib} do not prohibit the use of data-management + Joblib~\cite{joblib} \TG{joblib is multithreading, dask or ray. I wouldn't call it an engine. You should use another example.} do not prohibit the use of data-management strategies, they do not facilitate the integration of these strategies into their resulting workflow. To give neuroimaging applications data management capabilities, the applications must interact with a file system or library that enable the strategies.