We also quantify how costly delayed mitigation can be for severe diseases. Many thanks to Laura for her excellent first-author work, and to @violapriesemann.bsky.social for a wonderful collaboration.
Main results: with constant R₀, the optimal response is binary—either strict mitigation or none at all. With seasonality, mitigation is stricter in winter but infections peak in spring with a delayed single wave. Even with vaccination, optimal policies can produce transient waves.
Happy to announce a new preprint with
@violapriesemann.bsky.social
! We introduce an optimization framework to balance infection costs against mitigation costs during epidemics and pandemics, deriving optimal mitigation strategies rather than fixing policies ad hoc. arxiv.org/abs/2512.11454
We've arrived in San Diego for #SfN25! 🌴🧠 @maxeggl.bsky.social
Proudly representing @desantislab.bsky.social and super excited to attend the biggest neuroscience conference. Ready to learn, share, and connect! 🔬✨ #Neuroscience #SfN2025
Excited to be in San Diego to be showing off my work with @tatjanat.bsky.social at #SfN25 #Sfn2025! Come by on Monday and have a look if you are interested in synaptic plasticity and automated analysis of calcium images (PSTR152.15)!
Huge thanks to:
🔹 David Moratal - @upv.es
🔹 @plopezlarrubia.bsky.social - @iibm-csic-uam.bsky.social
🔹 Mohamed Selim - @uniofnottingham.bsky.social
🔹 From @neuroalc.bsky.social: Santiago Canals @canalslab.bsky.social, Silvia De Santis @desantislab.bsky.social & @maxeggl.bsky.social 🙌
However, this time I managed! I would like to thank
@silviadesantis.bsky.social @desantislab.bsky.social for the incredible support, which I think was exactly the difference that allowed me to succeed this year. Now on to more science and research at the intersection between neuroscience and ML!
With a bit of delay (it’s taken a while to process), I am happy to announce that I was awarded a #RyC2024 fellowship this year! One of the most prestigious research fellowships in Spain, I have been working towards this goal for a long time (with many rejections along the way)!
9/
💡 SpyDen bridges the gap between molecular-resolution imaging and user-friendly analysis.
If you’re doing fluorescence imaging of neurons—check it out.
Reproducible, customizable, and made for the community.
#OpenScience #Neuroinformatics #Microscopy
8/
📦 Get it here:
Code: github.com/meggl23/SpyDen
Compiled executables: gin.g-node.org/CompNeuroNet...
Documentation & tutorials included. This is an ongoing project so we’d love feedback from the community!
7/
🖥️ Works out of the box:
• For coders & non-coders
• Trained networks available
• No need to install complex dependencies
• Built for transparency and customization
6/
🧪 We validated SpyDen against expert annotations across diverse datasets and use cases.
It performs reliably, making it suitable for both exploratory research and reproducible pipelines.
5/
🧰 What can SpyDen do?
• Detect neurites & synapses
• Track fluorescent puncta over time
• Analyze intensity & localization
• Export data in standard formats
• All with a GUI and video tutorials for onboarding
4/
✅ Enter SpyDen:
A Python-based platform designed to address this with 3 core goals:
1️⃣ Easy to use for multiple tasks
2️⃣ Fully open-source, with open data formats
3️⃣ Editable annotations, robust across resolutions
3/
❌ Most workflows today:
• Multiple software packages
• Custom scripts
• Manual annotation
• Poor reproducibility
• Limited scalability
And many AI tools are not general-purpose, not open, or hard to modify.
2/
🔍 Motivation:
Studying learning and memory means understanding molecules inside axons, dendrites, and synapses.
Modern microscopy can detect single molecules—
…but analyzing those images? Still often manual, fragmented, or semi-automated.
🧵1/Just published: SpyDen, developed with @tatjanat.bsky.social @surbhitwagle.bsky.social, J. Filling, T. Chater & Y. Goda — an open-source Python tool for analyzing 2D microscopy time-series of neurons.
GUI-based, robust, and validated by experts.
🔗 shorturl.at/IVmyM
Read on for more: 👇
Our pop science article (in Spanish 🇪🇸) breaks it down for everyone: shorturl.at/esXTu
And check out the preprint too: shorturl.at/8agYv
@neuroalc.bsky.social
🧵
Why this matters:
✔️ Shorter, more comfortable scans
✔️ More patients can get high-quality imaging
✔️ Older, noisy datasets can now be rescued
✔️ Clinics can finally use advanced dw-MRI tools
All without needing to share or store sensitive patient data. (6/n)
We trained SBI to estimate key diffusion parameters from fewer MRI measurements — and it worked.
✨ 90% faster scans
✨ High accuracy preserved
✨ Robust even with noisy data
(5/n)
Enter: Simulation-Based Inference (SBI) — a new AI approach that flips the script.
Instead of training on huge amounts of real patient data, SBI learns from simulated data and can handle noisy, sparse measurements much more efficiently. (4/n)
Current methods need to oversample like crazy to get good-quality maps. That’s why advanced diffusion imaging is rarely used outside high-end research labs.
Patients? They get the basics—if they’re lucky. (3/n)
Diffusion-weighted MRI (dw-MRI) is a powerhouse in neuroimaging — revealing microstructural details of the brain without invasive procedures.
But there’s a catch: it’s slow and resource-heavy. Long scans = longer waitlists + fewer patients helped. (2/n)
Our article "Menos tiempo, mejor diagnóstico: cuando la IA alivia las listas de espera en resonancia magnética" was published @es.theconversation.com! It deals with the problem of making MRIs 10x faster without losing accuracy? Let’s dive in! @desantislab.bsky.social @silviadesantis.bsky.social
Incredible discussions and insights—already excited for next year’s edition! Interested in joining or learning more? Follow us for future updates, and stay tuned! 🚀 #Workshop #Neuroscience #AI
16 talented participants explored exciting frontiers at the intersection of neuroscience and AI. This year’s focus: dimensionality reduction and understanding high-dimensional data through dynamical systems. #DimensionalityReduction #DataScience
Just wrapped up the 3rd Bio-inspired Deep Learning Workshop, led by Angus Chadwick (@edinburgh-uni.bsky.social ), with co-organizers Laura Bernaez timon and Janko Petkovic, plus fantastic support from @isabelmaria-c.bsky.social and Arthur Pellegrino. Funded by @jherzstiftung.bsky.social. #NeuroAI 🧵
Don't miss out on this opportunity! Link to the application (deadline 21st of February): shorturl.at/s9RNI (flyer with more information here: rb.gy/21z6l9)
Exciting news! We got funding to organize "Bio-inspired Deep Learning" workshop near Mainz. This time on the topic of dimensionality reduction techniques and led by Angus Chadwick. Applications are now open for participants from both experimental/ computational backgrounds.