DeepMACT

Deep learning based Metastasis Analysis in Cleared Tissue

This online resource serves as guidance to implement DeepMACT; please refer to the publication for more details: Pan*, Schoppe*, Parra-Damas*, Cai, Todorov, Gondi, Neubeck, Böğürcü-Seidel, Seidel, Sleiman, Veltkamp, Förstera, Mai, Rong, Trompak, Ghasemi, Reimer, Coronel, Jeremias, Saur, Acker-Palmer, Acker, Garvalov, Menze, Zeidler, and Ertürk. Deep learning reveals cancer metastasis and therapeutic antibody targeting (2019) (available as a pre-print on BioRxiv)

The DeepMACT pipeline is an end-to-end procedure that enables highly automated detection and characterization of tumor micro-metastasis in large volumetric scans of whole cleared mice. This pipeline comprises two core technologies: DISCO tissue clearing and deep learning. Here, we provide additional material that may be beneficial to try out DeepMACT yourself:

• The DeepMACT handbook is intended to provide a step-by-step guideline to explain each implementation step in detail, with photos, videos, and code demos.
• You can download the full code, trained models, and (labeled) training data of a full-body scale mouse scan: DeepMACT demo package.
• (As mentioned in the publication and explained in detail in the handbook, we also provide a set of alternative algorithms, which are based on full 3D convolutions)
• Furthermore, please find a web-based version of DeepMACT on CodeOcean.com. This let's you run a fully functional demo of DeepMACT in your web browser without the need to install any software on your computer.
• Please click [here] to see videos of 3D animations of the DeepMACT output.
• Last but not least, please also check out our video tutorials on DISCO technologies.