Not sure about you Perry, but my hand has been up since 1996...

Congrats to @nephrin.bsky.social for knocking another thesis chapter out of the ballpark!

And just like 99% of all modern science, we couldn't have done this without federal support, so we thank NIH and DoD. 💰

Full paper is available here: authors.elsevier.com/a/1l7aP1SPTB... and as always the whole code including annotation and the training data is available on our GitHub or on Zenodo: zenodo.org/records/1527...

COBRA not only surpasses accuracy and reproducibility of existing AFM analysis methods by over an order of magnitude, but it also introduces a human-in-the-loop training segment that allows fine tuning for user-specific content. 🔍

Trained with over 5,000 indentation curves from 7 different cell types, COBRA rapidly processes AFM elastography data, or force-indentation curves, identifying contact point, removing noisy or anomalous curves, extracting biomechanical information including non-Hertzian depth-dependent modulus.

We built a deep learning algorithm, COBRA (comprises a COnvolutional and Bidirectional Recurrent neural network Architecture), to automatically process atomic force microscopy (AFM) data in real-time. 🔬

Thanks Kris Billiar for the shout out: www.sciencedirect.com/science/arti...

We are continuing to look into grant terminations at the NIH, and we want to hear from you.

Please reach out to us with your experiences: www.propublica.org/getinvolved/...