[Pycon] [new paper] "Giuseppe Di Bernardo" - Deep Learning for brain MRI segmentation: Big Data, AI and HPC meet together

[info a pycon.it]

Title: Deep Learning for brain MRI segmentation: Big Data, AI and HPC meet together
Duration: 45 (includes Q&A)
Q&A Session: 15
Language: en
Type: Talk

Abstract: With ever-increasing advancements in technology, neuroscientists are able to collect data in greater volumes and with finer resolution. There has been a growing interest in leveraging this vast volume of data across  levels of analysis, measurement techniques, and experimental paradigms to gain more insight into brain function. At multiple stages and levels of neuroscience investigation, ML holds great promise as an addition to the arsenal of analysis tools for discovering how the brain works. As quantitative analysis of brain MRI is routine for many neurological diseases and conditions, deep learning-based segmentation approaches for brain Magnetic Resonance Imaging (MRI) are gaining interest due to their self-learning and generalisation ability over large amounts of data.
On the other hand, High Performance Computing (HPC) and AI will increasingly 
intertwine as we transition to an exascale future using new computing, storage, and communications technologies. 
In this talk I will walk you through fundamentals of generating high-performance deep-learning models in TensorFlow platform using Python on large computing system (e.g NVIDIA® Tesla® GPUs powered by Tensor Cores), in order to infer and segment thousands of cell centroids out of the brain objects of interest.
>From a more technological perspective, although astonishing results have been achieved concerning the distribution of training  large convolutional neural networks on big data, to date the Python scientific ecosystem is still missing tools for an optimised and, above all, distributed inference of deep learning models. In this talk I will show you how a tiling-based inferencing approach could be a good solution to remedy the problem.
The talk is intended for intermediate PyData researchers and practitioners. Basic to intermediate level experience in image recognition/object detection deep learning applications is assumed. Overall, a good proficiency with the Python language 
and with scientific python libraries (e.g. numpy, TensorFlow, Keras) are required for the entire talk.

Tags: [u'ComputerVision', u'Machine Learning', u'GPUComputing', u'parallelization', u'data-analysis', u'bio-informatics', u'optimization', u'Artificial Intelligence']