Anthony A. Gatti

Overlaid axial, sagittal, and coronal views of the mean control knee (score = 0.0) and the mean recurrent instability knee (score = 3.5). A higher score, associated with the recurrent instability cohort, reflects a combination of greater patellar tilt, height, and lateral displacement; greater external rotation of the tibia relative to the femur; more knee valgus; and more posterior offset of the tibia relative to the femur.

3D > 2D for patellar instability 🦵

Our new study in OJSM introduces 3D-PASS, a score that describes patellar instability severity using 3D anatomical data and is capable of predicting patient outcomes

Higher 3D-PASS <> worse predicted outcomes

🔗 doi.org/10.1177/2325...

@aossmjournals.bsky.social

We're moving closer to using 3D imaging to personalize operative vs nonoperative decisions for patients. 🏥🩹

Very grateful to collaborate with an amaing team including @aagatti.bsky.social, the JUPITER Study Group, Scott Delp, and Seth Sherman!

YouTube video by Mobilize Center Webinar: Integrating Bone and Joint Geometry into Musculoskeletal Models, Part 1 of 2

Watch the @mobilizecenter.bsky.social/Restore Center webinar with @aclouthier.bsky.social and Erin Lee entitled “Integrating Bone and Joint Geometry into Musculoskeletal Models”

Research presentation: youtu.be/QtqDx3_uK9I
Tutorial: youtu.be/wqxQl3Dm9lc

This is a point missed by most Canadians. We need to stop worrying about what the US is doing and reflect on what we can be doing.

Is this less of a trick if the goal is to see the SMD on elite athletes? It’s only biasing it/misleading if we are interpreting the effect size for the population of all people?

This ☝🏽 💯

If 🇨🇦 created a thriving research environment by investing in tri-council & ditching the PJT focused approach that kills investment in long term programs - we would ATTRACT international talent vs bribing them to move here; to only leave later when they see there’s no sustainable funding!

I can't* fathom why the top picture, and not the bottom picture, is the standard diagram for an autoencoder.

The whole idea of an autoencoder is that you complete a round trip and seek cycle consistency—why lay out the network linearly?

“Everyone knows” what an autoencoder is… but there's an important complementary picture missing from most introductory material.

In short: we emphasize how autoencoders are implemented—but not always what they represent (and some of the implications of that representation).🧵

We are starting to see some nuanced discussions of what it means to work with advanced AI in its current state

In this case, GPT-5 Pro was able to do novel math, but only when guided by a math professor (though the paper also noted the speed of advance since GPT-4)

The reflection is worth reading.

We're thrilled to introduce ATHENA: Automatically Tracking Hands Expertly with No Annotations – our open-source, Python-based toolbox for 3D markerless hand tracking!

Paper: www.biorxiv.org/content/10....

Oh so it does have phd level intelligence

Methods like NeRF and Gaussian Splats model the world as radioactive fog, rendered using alpha blending. This produces great results.. but are volumes the only way to get there?🤔 Our new SIGGRAPH'25 paper directly reconstructs surfaces without heuristics or regularizers.

French fries just a surrogate for fast food in this case?

Rest in peace, boss. @beierlab.bsky.social -- your stoic and positive support helped me in too many ways to count over the years.

Biggest human imaging study scans 100,000th UK volunteer UK Biobank scientists say the human body can be studied in greater detail than ever thanks to people like Steve.

What an absolutely extraordinary milestone. Congrats @fmrib-steve.bsky.social @fmrib-karla.bsky.social and everyone else involved.

www.bbc.co.uk/news/article...

I love this type of approach - for imaging, biomechanics, or other domains…

Take high quality dense data, use physics simulators to generate synthetic sparse data, then train a model to recon dense data from sparse measurements.

Lots of potential pitfalls, but the concept is immensely powerful.

I’m torn on this.

I feel like it is valuable to be honest in the data and to improve it if things change/update.

But it just seems like a nightmare to keep up on the literature when not only is there a new paper, but the one you’ve read changed.

We write papers, but we build people.

You (yes, you, my fellow academic) too; please don’t forget that.

✨Some news✨: after finishing my postdoc, I’ll be starting my lab as an Assistant Professor at Georgia Tech. Join us in Atlanta to study how joints work and where they come from!

Something big is coming up tomorrow. Watch this space…
💪🏻 ⚡️ ⚽️ 💪🏻🦵🏃🏻‍♂️‍➡️👟🥋🏋🏻🚴‍♂️📈📊📚

Are there useful or interesting ways to use LLMs other than prompting them? I feel like compressing all the text in the world via a hierarchically structured statistical model is probably useful, but that we're using it in a way that is unlikely to do what we'd hope.

Since LLMs came out I’ve thought they are the perfect tool for systematic reviews. This is an excellent use case.

Are you interested in muscle #biomechanics and looking for #postdoc opportunities?

If you have skills in (statistical) shape modelling, advanced data analyses or 3D imaging/model reconstructions, get in touch with me!

I will be at #SEB2025 (Antwerp), #CNB2025 (Helsinki), and #ISB2025 (Stockholm).

Centre of Functional and Metabolic Mapping - Western University Western University, in vibrant London, Ontario, delivers an academic and student experience second to none.

Our Siemens 7T Collaboration Scientist is moving to support the Siemens 7T Terra.X Impulse at Stanford. Siemens is rehiring for our position, located in Canada's National Ultra-High Field MRI platform cfmm.uwo.ca in London, Ontario 🇨🇦. onehealthineers.wd3.myworkdayjobs.com/SHSJB/job/OA...

Animation technique simulates the motion of squishy objects MIT researchers developed a computationally efficient method that could enable artists to design realistic simulations of elastic objects, like bouncy or squishy characters, for animated movies or vid...

MIT News today features a paper with fun examples coming out this summer by our brilliant PhD student Leticia Mattos Da Silva, postdoc—and Columbia professor in a matter of days!— @silviasellan.bsky.social, and PhD student collaborator Natalia Pacheco-Tallaj. news.mit.edu/2025/animati... Flubber!

Image from article in Radiology: Artificial Intelligence

Radiology: Artificial Intelligence has joined the rapidly growing #Bluesky social media network https://pubs.rsna.org/page/ai/blog/our_presence_on_bluesky @radiology-ai.bsky.social @cekahn.bsky.social #ESPR2025 #PedsRad #radiology

You can never cover everything - totally fair.

Now makes me wonder about this approach for fitting a generative model to a new example. Eg optimizing latent vector to best recon new surface.

Awesome work.

I'm particularly interested in generative SDFs. This coarse-to-fine analogy seems similar to curriculum learning which has been proposed previously (arxiv.org/abs/2003.08593) ... but simpler.

Is this your intuition also?

bsky.app/profile/nmws...