How calcium imaging analysis actually works — bench scientists

Consent

Consent * I consent to my responses being used for this research.

Section 1 — You and your work

B1. Your position * PhD student Postdoc Research assistant / technician Research software engineer Core facility staff Other

B2. How long have you worked with calcium imaging data? Less than 1 year 1–2 years 3–5 years 5+ years

B3. In a typical MONTH, roughly how many hours do you personally spend on calcium imaging DATA ANALYSIS (not acquisition)? * key hours per month

B4. Which tools do you personally use? (select all) * key Suite2p CaImAn Inscopix IDPS EZcalcium Custom MATLAB (the lab's own) Custom Python (the lab's own) Scripts I wrote myself Imaris / Aivia Other

B5. After preprocessing (motion correction + segmentation), what do you do next, and in what? * key

Please be specific about file formats and tools — this is one of the most useful answers in the survey.

B6. How much of your analysis code is…

Item	0–10%	11–25%	26–50%	51–75%	76–100%
Written by me from scratch	Written by me from scratch — 0–10%	Written by me from scratch — 11–25%	Written by me from scratch — 26–50%	Written by me from scratch — 51–75%	Written by me from scratch — 76–100%
Inherited from a previous lab member	Inherited from a previous lab member — 0–10%	Inherited from a previous lab member — 11–25%	Inherited from a previous lab member — 26–50%	Inherited from a previous lab member — 51–75%	Inherited from a previous lab member — 76–100%
Copied from a tutorial / paper / GitHub	Copied from a tutorial / paper / GitHub — 0–10%	Copied from a tutorial / paper / GitHub — 11–25%	Copied from a tutorial / paper / GitHub — 26–50%	Copied from a tutorial / paper / GitHub — 51–75%	Copied from a tutorial / paper / GitHub — 76–100%
A packaged tool used as-is	A packaged tool used as-is — 0–10%	A packaged tool used as-is — 11–25%	A packaged tool used as-is — 26–50%	A packaged tool used as-is — 51–75%	A packaged tool used as-is — 76–100%

0–10%

11–25%

26–50%

51–75%

76–100%

Written by me from scratch

Inherited from a previous lab member

Copied from a tutorial / paper / GitHub

A packaged tool used as-is

Section 2 — Where the time actually goes

B7. Think about the LAST TIME you analysed a calcium imaging dataset. What took longer than it should have? * key

B8. How often does each of these happen to you? *

Item	Never	Rarely	Sometimes	Often
Re-running an analysis because a parameter was wrong	Re-running an analysis because a parameter was wrong — Never	Re-running an analysis because a parameter was wrong — Rarely	Re-running an analysis because a parameter was wrong — Sometimes	Re-running an analysis because a parameter was wrong — Often
Not being sure whether a result is real or an artefact	Not being sure whether a result is real or an artefact — Never	Not being sure whether a result is real or an artefact — Rarely	Not being sure whether a result is real or an artefact — Sometimes	Not being sure whether a result is real or an artefact — Often
Being unable to reproduce something I did a few months ago	Being unable to reproduce something I did a few months ago — Never	Being unable to reproduce something I did a few months ago — Rarely	Being unable to reproduce something I did a few months ago — Sometimes	Being unable to reproduce something I did a few months ago — Often
Waiting on a long computation	Waiting on a long computation — Never	Waiting on a long computation — Rarely	Waiting on a long computation — Sometimes	Waiting on a long computation — Often
Getting a different answer than a labmate on the same data	Getting a different answer than a labmate on the same data — Never	Getting a different answer than a labmate on the same data — Rarely	Getting a different answer than a labmate on the same data — Sometimes	Getting a different answer than a labmate on the same data — Often
Manually curating ROIs	Manually curating ROIs — Never	Manually curating ROIs — Rarely	Manually curating ROIs — Sometimes	Manually curating ROIs — Often
Rewriting code that surely already exists somewhere	Rewriting code that surely already exists somewhere — Never	Rewriting code that surely already exists somewhere — Rarely	Rewriting code that surely already exists somewhere — Sometimes	Rewriting code that surely already exists somewhere — Often
Discovering a problem only after showing results to my PI	Discovering a problem only after showing results to my PI — Never	Discovering a problem only after showing results to my PI — Rarely	Discovering a problem only after showing results to my PI — Sometimes	Discovering a problem only after showing results to my PI — Often

Never

Rarely

Sometimes

Often

Re-running an analysis because a parameter was wrong

Not being sure whether a result is real or an artefact

Being unable to reproduce something I did a few months ago

Waiting on a long computation

Getting a different answer than a labmate on the same data

Manually curating ROIs

Rewriting code that surely already exists somewhere

Discovering a problem only after showing results to my PI

B9. Roughly how much of your analysis time goes on CHECKING that the analysis is correct, as opposed to doing it? * key 0–10% 11–25% 26–50% 51–75% More than 75%

B10. Have you ever presented or submitted a result that later turned out to be wrong because of an analysis problem? Yes No Not that I know of Prefer not to say

Section 3 — Reaction to the concept

A system that takes data you have ALREADY PREPROCESSED (Suite2p/CaImAn output, or an NWB file) and works through the rest of the analysis: checking the data is what you think it is, assessing quality, inferring spikes, characterising population activity, making figures, and drafting a report.

It shows you every line of code it runs, pauses for your approval between stages, and records everything so it can be re-run exactly later.

It does NOT do motion correction or segmentation — it picks up after those.

B11. Honestly, what is your first reaction? *

B12. How useful would each be to you personally? *

Item	Not useful	Somewhat useful	Very useful	Would change how I work
It shows the code it ran, so I can check it	It shows the code it ran, so I can check it — Not useful	It shows the code it ran, so I can check it — Somewhat useful	It shows the code it ran, so I can check it — Very useful	It shows the code it ran, so I can check it — Would change how I work
It tells me when data quality is too poor to conclude anything	It tells me when data quality is too poor to conclude anything — Not useful	It tells me when data quality is too poor to conclude anything — Somewhat useful	It tells me when data quality is too poor to conclude anything — Very useful	It tells me when data quality is too poor to conclude anything — Would change how I work
It catches when a file isn't what I assumed	It catches when a file isn't what I assumed — Not useful	It catches when a file isn't what I assumed — Somewhat useful	It catches when a file isn't what I assumed — Very useful	It catches when a file isn't what I assumed — Would change how I work
It drafts a methods section and figures	It drafts a methods section and figures — Not useful	It drafts a methods section and figures — Somewhat useful	It drafts a methods section and figures — Very useful	It drafts a methods section and figures — Would change how I work
I can re-run the exact analysis months later	I can re-run the exact analysis months later — Not useful	I can re-run the exact analysis months later — Somewhat useful	I can re-run the exact analysis months later — Very useful	I can re-run the exact analysis months later — Would change how I work
Everyone in the lab gets the same analysis	Everyone in the lab gets the same analysis — Not useful	Everyone in the lab gets the same analysis — Somewhat useful	Everyone in the lab gets the same analysis — Very useful	Everyone in the lab gets the same analysis — Would change how I work
It suggests analyses I hadn't thought of	It suggests analyses I hadn't thought of — Not useful	It suggests analyses I hadn't thought of — Somewhat useful	It suggests analyses I hadn't thought of — Very useful	It suggests analyses I hadn't thought of — Would change how I work

Not useful

Somewhat useful

Very useful

Would change how I work

It shows the code it ran, so I can check it

It tells me when data quality is too poor to conclude anything

It catches when a file isn't what I assumed

It drafts a methods section and figures

I can re-run the exact analysis months later

Everyone in the lab gets the same analysis

It suggests analyses I hadn't thought of

B13. What would STOP you using something like this? (select all) * key I don't trust automated analysis for results I would publish I would have to check its work anyway, so it saves nothing My data isn't in a format it would read Learning it costs more than it saves My PI wouldn't accept results from it I enjoy / need to do this myself — it is my skill Concern about my data leaving the lab Nothing — I would try it Other

B14. If it produced a figure for your paper, what would you need before trusting it? (select all) * key Read the code Compare against my own manual analysis on the same data Published validation Approve every step myself Test it on data where I know the answer Nothing would convince me Other

B15. Would you rather it be… * Fully automatic — give me the answer Automatic but shows its work and asks before each stage A set of building blocks I assemble myself Not sure

Section 4 — Practicalities

B16. Where does your raw and processed data live? (select all) * key Lab workstation Institutional cluster / HPC Institutional network storage Cloud (S3 / Google / Dropbox) External hard drives Public archive (DANDI etc.)

B17. Roughly how large is a typical SINGLE imaging dataset for you? * key Less than 1 GB 1–10 GB 10–100 GB More than 100 GB

B18. Would your institution's rules allow you to upload research data to an external service? * key Yes Yes, with approval No Don't know

B19. Do you currently use any AI tools in your research work? (select all) ChatGPT / Claude / Gemini for code For writing Copilot or similar in the editor A domain-specific research tool None Prefer not to say

B20. Anything else we should know?