Philip Noonan - Pawel Markiewicz

Title - Harnessing the Power of the GPU for Fast and Robust Motion Tracking, Image Reconstruction and Analysis.

Part 1: Philip Noonan

Subject motion can greatly decrease medical image quality if not accounted for. Many techniques exist in how to reduce the effects of motion such as head restraints, data driven motion correction, and external motion tracking systems. Here we explore the latter and demonstrate using consumer grade depth sensing cameras such as the Microsoft Kinect. We describe the Kinect and showcase how its 3D and colour videos can be used to implement state-of-the-art computer vision algorithms. Specifically, we will report its use as a tool for real-time tracking during brain PET, highlight the additional modifications made to the hardware to enable its use in brain PET/MRI, and finally describe the software to enable non-rigid whole body motion tracking.

Part 2: Pawel Markiewicz

I will present an advanced high-throughput (GPU) software infrastructure dedicated to amyloid imaging with state-of-the-art quantitative accuracy and precision estimation. The proposed open source software package, called NiftyPET, allows for full control of every single part of raw list-mode data processing, motion tracking and correction, image reconstruction and analysis, leading to precise estimates of regional or voxel-level image statistics. I will show how the innovative solutions used in this package are useful for advanced harmonisation of image reconstruction and analysis across all UK PET/MR centres and the Dementias Platform UK (DPUK) network