(3/3) we used the benchmark dataset proposed in Ritsche & Sarto et al. (2025) osf.io/preprints/os... for reproducible and comparable algorithm performance analysis.

@mvfranchi.bsky.social @fabiosartophd.bsky.social @olivierseynnes.bsky.social

(2/3) Compared to a recent publication by Van der Zee et al. we show:

1. Improvement of new DL_Track_US models
2. Higher comparability between both approaches
3. Comparison of four different model types

💡Details in the news section of the docs ⤵
paulritsche.github.io/DL_Track_US/...

⏭️ Next up:

⬆️ increase comparability to MR assessed muscle volume by:

🔍 objectifying calibration methods
🐍 automate image analysis

UMUD is here to support muscle ultrasound research, but we can't do it alone...

We need YOU to:
🤝 Contribute datasets & annotations
🔬 Test & refine algorithms
💬 Share insights & ideas

Together, we can foster open science and reproducible research in ultrasound imaging 🤗

What else does UMUD include? 🤔

📊 Indexing (so far) 11 datasets with 75,569 images & 2,573 videos from 1,769 participants.
🔍 Detailed metadata for easy dataset exploration: muscle groups, devices, demographics, & more.
💡 Guidance on state-of-the-art automated analysis tools

I’m Paul 👋🏼 I study muscle architecture using ultrasonography and how it adapts during maturation, ageing or due to training. I try to automate the analysis of (EFOV) muscle ultrasonography because I got tired of clicking on images. I am currently working in 🇨🇭