name: cover layout: true class: indos-cover no-logo no-counter --- count: false .indos-title[
# Federated Journal Club ## How to read and review a neuroimaging paper **Mélanie Garcia, PhD** .meta[WG3 Training School · Madrid and online · September 30, 2026] ] --- name: section layout: true class: indos-section no-logo no-counter --- name: content layout: true .perma-sidebar[
INDoS WG3 Training School (Sep 30, 2026)
] --- class: vcenter # This afternoon .cards.steps.agenda[ .card[ .enum.dur[15′] **Theory:** how to read a paper ] .card[ .enum.dur[15′] **The web app:** presentation and registration ] .card.break[ .enum.dur[15′] Break ] .card[ .enum.dur[55′] **Individual reading** and critical annotation ] .card[ .enum.dur[20′] **Group discussion:** the paper section by section, comparing annotations ] .card[ .enum.dur[10′] **Group reports** and statistics ] ] --- class: vcenter # About me .cards.cols-3.story[ .card[ .kicker[PhD · 2019–2023] ### Deep learning and autism Brain sMRI and fMRI .gray-text[Trinity College Dublin, School of Medicine, Ireland] ] .card[ .kicker[Postdoc · 2023–2026] ### Machine learning in psychiatry sMRI, fMRI, EEG and clinical data .gray-text[Massachusetts General Hospital, Harvard Medical School, USA] ] .card[ .kicker[Now] ### Innovations Editor Aperture Neuro ] ] .boxed-content[ .gray-text[Neuroimaging · machine learning · psychiatry] ] ??? TODO: introduce yourself in a few sentences (~30 seconds). --- class: no-logo
THE
method
??? Are you ready? I am going to teach you THE method for reading any scientific paper. [pause] Actually, I am not. Having worked as a PhD candidate and postdoc in neuroimaging and psychiatry taught me the concept of neurodiversity, and made me realize how beautiful and diverse human minds are. --- # Imagine an apple .pull-left.center[ .apple[🍎] .figcap.invisible[Some of us *see* it] ] .pull-right.center.invisible[
apple
fruit
red
round
crunchy
.figcap[Some of us *know* it: as a concept, a graph, a diagram] ] .slide-footer.invisible[ .ref[*Aphantasia*: the inability to voluntarily form mental images.] ] ??? For instance, have you heard of aphantasia? It's when you can't really form mental images. If someone asks you to imagine an apple, you know what an apple is, but you don't see it in your head. --- count: false # Imagine an apple .pull-left.center[ .apple[🍎] .figcap[Some of us *see* it] ] .pull-right.center[
apple
fruit
red
round
crunchy
.figcap[Some of us *know* it: as a concept, a graph, a diagram] ] .slide-footer[ .ref[*Aphantasia*: the inability to voluntarily form mental images.] ] ??? My husband is like that. He thinks more in concepts, like graphs or diagrams. --- # Even one reader reads differently .boxed-content[
You
One paper
1st read · the images
2nd read · the stats
3rd read · the pipeline
4th read · could I redo it?
Your
understanding
.center.larger[Each time you read the same paper, you notice something new.] ] ??? DRAFT — adapt to your own words. Even on your own, you never read a paper the same way twice. The first time, your eye goes to the images; the next time, to the statistics; later, you follow the pipeline step by step, or wonder whether you could redo it. --- # ...and we all read differently .boxed-content[
One paper
A visual thinker
A conceptual thinker
A statistician
A pipeline builder
Federated
review
.center.larger[A **federated review** combines everyone's readings.] ] ??? So no, we are not all built the same way, and that's great! It means we all read differently, with different perspectives and different biases. And comparing them is exciting: that's how we explore and learn. That's why the idea of our following activity will be to leverage this variety of views, with the aim of building federated reviews of articles: reviews built by combining everyone's readings. --- # How to read a paper: your turn .pull-left.guide-steps[ .cards.steps[ .card[ .enum.dur[10′] **Read the guide** from the Journal Club organizers. ] .card[ .enum.dur[5′] .step-body[ **Discuss with your neighbour:** * What is your overall impression of the guide? * Which of the approaches seems most useful to you, and why? * How does it compare with the way you usually read papers? ] ] ] ] .pull-right.center[
.qr-link[[indos-costaction.eu/journal-club/reading.html](https://www.indos-costaction.eu/journal-club/reading.html)] ] ??? DRAFT — adapt to your own words. If you feel a bit lost or unsure how to approach a paper, don't worry. The organizers wrote a short guide on how to read a paper: scan the QR code or open the link. Take ten minutes to read it. Then turn to your neighbour, or pair up in the chat if you're online, and talk for five minutes: what is your overall impression of the guide, which of the approaches seems most useful to you and why, and how does it compare with the way you usually read papers? If one of the methods resonates with you, today is a good opportunity to try it out. --- exclude: true # Three ways in .cards.cols-3[ .card[ ### Read in passes Triage first: title, abstract, figures, conclusions. Then read properly. Then read as if you had to re-run the study. ] .card[ ### Read methods as a pipeline For each step: what it does, what it assumes, which parameters were chosen or left at default. ] .card[ ### Question the figures What is plotted, with what sample size? Does it support the claim in the text, or a weaker one? ] ] .boxed-content[
If Methods and Results feel hard: that's normal. Reading papers is a skill that grows with practice. More in the Journal Club guide: [indos-costaction.eu/journal-club/reading.html](https://www.indos-costaction.eu/journal-club/reading.html) ] .slide-footer[ .ref[Keshav (2007) [doi:10.1145/1273445.1273458](https://doi.org/10.1145/1273445.1273458) · Hoskins et al. (2007) [doi:10.1534/genetics.107.071183](https://doi.org/10.1534/genetics.107.071183) · Hubbard & Dunbar (2017) [doi:10.1371/journal.pone.0189753](https://doi.org/10.1371/journal.pone.0189753)] ] ??? BACKUP SLIDE: hidden with `exclude: true`. Delete that line to show it again. DRAFT — adapt to your own words. About 3 minutes. Three approaches from the Journal Club's guide, written by the organizers. Take whichever speaks to you. First, read in passes rather than top to bottom. Srinivasan Keshav's three-pass method: a quick triage (title, abstract, figures, conclusions), then a proper read where you note what you don't understand, then a last pass where you imagine re-running the study yourself. Second, since most papers in the pool are about preprocessing, read the methods like a pipeline: for each step, what it does, what it assumes, and which parameters were chosen deliberately versus left at default. Third, question the figures, which is the idea behind the CREATE method: what is actually plotted, with how many subjects, and does it support the claim in the text, or only a weaker version of it? And if you find Methods and Results the hardest part: a survey of 260 students and researchers found exactly that for less experienced readers. It is a skill that grows with practice, which is what today is for. The full guide is at the link. --- template: section .section-mark[
# The activity How the Federated Journal Club works, and what we do in the next two hours ] ??? Let's see in more detail how the activity will go. --- # Four questions for today .boxed-content.larger[ No single right way to read a paper, but the same four questions, so we can compare our readings. ] .cards.cols-2.questions[ .card.q1[ ### What is claimed? ] .card.q2.hide[ ### What evidence supports it? ] .card.q3.hide[ ### What is missing or unclear? ] .card.q4.hide[ ### Could I reproduce it? ] ] ??? By now you've probably understood that there is no single right way to read a paper. But for today's activity, I would like you to answer a set of questions to guide all of us during our time practicing reviewing. When reading and annotating the paper, please try to find answers to these four questions: --- count: false # Four questions for today .boxed-content.larger[ No single right way to read a paper, but the same four questions, so we can compare our readings. ] .cards.cols-2.questions[ .card.q1[ ### What is claimed? ] .card.q2[ ### What evidence supports it? ] .card.q3.hide[ ### What is missing or unclear? ] .card.q4.hide[ ### Could I reproduce it? ] ] --- count: false # Four questions for today .boxed-content.larger[ No single right way to read a paper, but the same four questions, so we can compare our readings. ] .cards.cols-2.questions[ .card.q1[ ### What is claimed? ] .card.q2[ ### What evidence supports it? ] .card.q3[ ### What is missing or unclear? ] .card.q4.hide[ ### Could I reproduce it? ] ] --- count: false # Four questions for today .boxed-content.larger[ No single right way to read a paper, but the same four questions, so we can compare our readings. ] .cards.cols-2.questions[ .card.q1[ ### What is claimed? ] .card.q2[ ### What evidence supports it? ] .card.q3[ ### What is missing or unclear? ] .card.q4[ ### Could I reproduce it? ] ] ??? Sharing this framework will enable us to compare annotations across individuals and groups more easily. Now, the tool we'll use for the activity. --- # The web app: newgaze .cards.steps.webapp[ .card[ .enum[1] Go to **[newgaze.com](https://newgaze.com)** ] .card[ .enum[2] Enter the access code: **[CODE]** ] ] ??? TODO: replace [CODE] with the preview access code. Live demo. --- # 55′ · Individual reading and annotation .pull-left.larger[ 1. Answer the **self-assessment questions**. 2. Open the **paper assigned to you**. You may read more if you have time after this one. 3. Annotate with the **four questions** in mind, and leave one longer comment at the end: your overall take. .gray-text[Stuck? Raise your hand, or write in the chat if you are online.] ] .pull-right[ .cards.cols-1.review[ .card[ **Anchor each comment** to the exact passage, step or figure it is about. ] .card[ **Cover the whole paper**: Methods and Results too, not just the introduction. ] .card[ **Say what you didn't understand.** A precise "I don't get why…" is one of the most useful comments. ] ] .admonition.warning[
**No AI.** Read and annotate the paper yourself. ] ] ??? DRAFT — adapt to your own words. You now have 55 minutes on your own. In newgaze, first answer the self-assessment questions, then open the paper assigned to you. If you finish early, you may read another one. A few tips for good annotations: attach each comment to the exact passage it is about; cover the whole paper, not just the introduction; and don't hesitate to say what you didn't understand. A precise "I don't get why…" is one of the most useful comments you can leave. And please, no AI: the point is your own reading. TODO: check the no-AI rule still applies with newgaze. --- class: vcenter # 20′ · Group discussion .boxed-content.larger[ In groups, go through the paper **section by section** and compare annotations. * The **least experienced** person speaks first about their annotations. * Where did you notice the same thing? Where did you disagree, and why? * What did one person catch that the others missed? The least experienced person also **reports the group's outcomes**. ] ??? TODO: say how groups are formed (same paper? same modality?). Why the least experienced person first: once a senior person has spoken, others tend to align with them. Starting with the least experienced keeps every reading on the table, and reporting back gives them the floor. --- # 10′ · Group reports and statistics .boxed-content.larger[ Each group, about two minutes: ] .cards.cols-2.questions[ .card.q1[ ### Claim: in one sentence ] .card.q2[ ### Evidence: the strongest piece ] .card.q3[ ### Missing: the biggest gap ] .card.q4[ ### Reproducible? Yes, partly or no ] ] ??? Then we look at the numbers from the whole session (next two slides). --- template: section .section-mark[
# Federated Journal Club # Global Outcomes Activity on September 30, 2026 ] --- class: vcenter # Summary of the experiment .boxed-content[ A hybrid session, in person in Madrid and online, two hours on September 30, 2026. Each participant was assigned one of **7 articles**, pulled from the [Journal Club pool](https://www.indos-costaction.eu/journal-club/), and could read and review others once that review was done. Groups then discussed and reported on each article. .cards.cols-4.stats[ .card.q1[ .num[30] participants answered the self-assessment ] .card.q2[ .num[148] annotations ] .card.q3[ .num[15] completed reviews ] .card.q4[ .num[7] articles ] ] ] ??? DRAFT — adapt to your own words, and check the figures against your data. Here is what we did together today, in numbers. Thirty of you answered the self-assessment questions, you left 148 annotations, and 15 reviews were completed, across 7 articles. --- # What did we annotate? .pull-left.fjc-left[
] .pull-right.fjc-right[ .cards.cols-1.review[ .card.q1[ About **half** of the annotations carry **no type**. ] .card.q2[ Among typed ones: **questions**, **positive points**, **unclear passages** and **missing information** come first. ] ] ] ??? DRAFT — adapt to your own words, and check the figures against your data. Now, what kind of comments did we leave? Each annotation could be tagged with a type. First observation: about half of them were left without a type. Among the typed ones, the most common were questions, positive points, passages that needed clarification, and missing information. --- class: vcenter # How long did it take?
.slide-footer[ .ref[All data anonymized. Groups are small (n = 1 to 9 per category): read the plots as a description, not a test.] ] ??? DRAFT — adapt to your own words, and check the figures against your data. This shows how long each of the 15 completed reviews took, split by career stage, by how good a time it was for reading, and by how familiar you were with the topic. Most reviews took between about half an hour and an hour and a quarter; the median is around 40 minutes. A few were completed in only a few minutes. The groups are small, so these are descriptions, not findings. --- template: section .section-mark[
# Thank you Questions? · [indos-costaction.eu/journal-club](https://www.indos-costaction.eu/journal-club/) .fine[This presentation is based upon work from COST Action CA24161 “Improving Neuroimaging Data for Sharing” (INDoS), supported by COST (European Cooperation in Science and Technology), [www.cost.eu](https://www.cost.eu).] ]