Neuroimaging datasets: organize the context, not just the files
Use a structured review process to understand acquisition details, participant boundaries, derivatives, and dataset handoffs.
Read the guide ↗DATA & DECISIONS
Data Science Neuro Browser
Connect exploratory analysis with a documented question, meaningful labels, and a reproducible account of how the evidence was prepared.
TOPIC GUIDE / 03 · READ, QUESTION, EXPLORE
Start with a data question
Before selecting an algorithm, describe the unit of analysis and the intended use of the result. A row might represent a document, a session, or a participant; mixing those levels can make an analysis difficult to interpret. Keep missing values, exclusions, and transformations in a readable record. Preserve the original material separately from working copies.
Keep decisions inspectable
Create a simple data dictionary and note why each field exists. Use a held-aside review collection for changes that affect classification. When a source is removed or corrected, make its downstream use traceable. These habits support a clearer analysis even without machine learning and give reviewers a concrete starting point when they disagree with a result.
Three questions to take with you
01What does one record represent?
02Which transformations changed the data?
03Can another reviewer reproduce the selection?
Reference for this topic: Brain Imaging Data Structure introduction. Read the original context alongside the guide.