MICCAI 2020

2 Open Source Deep Learning Platform 0 DAILY Mo n d a y functions, andmetrics that are specific to medical image analysis tasks . With release 0.3, MONAI has added 3 new evaluation metric functions, 11 new network layers and blocks, 6 new network architectures, and 14 new data transforms. All of this is accomplished within a framework that promotes reproducibility - which is fundamental to open science and of vital importance to medical applications. The openness of MONAI comes from its rapidly expanding community of contributorsandusers, itseaseofuse,and its documentation. MONAI has received over 1,400 stars on github as well as contributions from over 39 developers from around the world. The MONAI team organized a MICCAI endorsed, three-day bootcamp in September of 2020, and that bootcamp received over 560 attendance applications from 40 countries. In version 0.3, the MONAI repositories have been re-organized so that the core repository is more compact, tutorials are easier to find and follow, and a research repository has been created to host cutting-edge contributions that are companion to recent peer reviewed publications. The recordings from the bootcamp are also being posted on the MONAI website. When we asked MONAI and ITK developers, why do you develop open source deep learning software? The most consistent answer was “ impact ”. The developers wanted their efforts and research to have the biggest possible impact on themedical field, and they saw using and contributing to open source MONAI and ITK as the best way to have that impact. This close alignment of the primary goal of the developers and their open source projects means that these developers and these platforms are likely to be working together for a very long time. The developers also enjoyed the collaborations that came from producing open source, with those collaborations often improving their research as well as their code.