Using AVID for 3D Functional Region Of Interest (ROI) Creation Jeff Hoerle Friday Forum November 5 th, 2004
AVID - Outline Motivation Initial Concepts Solution and Demonstration Implementation User Study Overview Future Work
AVID - Motivation Where does the data used by AVID come from? Cognitive Neuroscience Simple Example xx x x x x x xxxx x xxx x = Functional Image Acquisition
AVID - Motivation Functional Volumes Event List (What and when) SPM tstatprofile Activation Map Activation Map – contains a statistical score (e.g. t- statistic) for each voxel in the functional volume that indicates the likelihood of activation due to a specific stimuli Generating an Activation Map
AVID - Motivation What is a region of interest (ROI)? - any sub-region in space that possesses some “interesting” feature which requires further investigation How is an ROI specified? - for a 3D volume, it typically involves a slice-by-slice investigation and selection of voxels - problem: time-consuming
AVID - Motivation Example: Functional ROI Drawing Tool
AVID - Motivation Can we use a 3D interface and interaction scheme to improve ROI creation? What is the best way to navigate and interact in 3D using 2D interaction devices (e.g. mouse, desktop monitors)?
AVID – Initial Concepts Scheme One (1)Specify a bounding polygon or circle (2)Adjust the endplanes of the projected volume (3)Accept or reject volume Step OneStep Two
AVID – Initial Concepts Scheme Two (1)Click any point in the 3D volume (2)Select a point along the ray cast through the volume to serve as the midpoint for the to-be-created sphere (3)Use mouse to modify the diameter and placement of the sphere (and possibly shape too – make the sphere an ellipsoid)
AVID – Initial Concepts Scheme Three – use a magic wand tool to select an entire cluster (connected component) of activation
AVID - Solution AVID uses a 3D magic wand Pros: - Takes advantage of inherent clustering in activation maps - Very simple, avoid multi-step process Cons: - What if the desired ROI consists of many small clusters of activation? - Region growth can extend beyond the desired anatomical region. What can be done to prevent this?
AVID Demonstration
AVID - Implementation Windows… but why? - User population (BIAC) almost exclusively consists of Windows users Visualization Toolkit (VTK) - Vastly sped the development of the 3D interface Written in C++ using MFC library - VTK written in C++, making C++ an obvious choice for AVID
AVID - Implementation VTK - developed by Kitware ( - open-source, object-oriented graphics library - written in C++ using OpenGL or Mesa as graphics API - can interface graphics API using Cocoa, Carbon, Win32 or X - has bindings written for Java, Tcl, Python
AVID - Implementation Human Heart Simulation Reservoir Model - GeoMap - Univ of Oslo Abdomen Segmentation – SINTEF CactusX
AVID – Implementation VTK Pros: - design intelligently, factory classes allow the same code can be compiled for Linux, Mac, Windows (ideally) - speeds development, has pre-built visualization algorithms VTK Cons: - VTK is HUGE and complex… over 900 classes with deep class hierarchies - doing things the “non-VTK way” can be (very) painful - not enough documentation
AVID - Implementation Collaboration Diagram Inheritance Diagram
AVID – User Study Intended to be used as a feedback tool to improve functionality and extend feature set NOT intended to provide a quality measurement of AVID based on some set of metrics Recruit five to seven end-users to participate in a 20 – 30 minute session
AVID – User Study How does user expectation align with AVID functionality? Do system defaults make sense? What is the most intuitive mechanism to add/remove voxels from an ROI? - Can we think of ROI creation in terms of file selection?
AVID – Future Work Incorporate feedback from user study into new releases of AVID Revisit the idea of 3D selection - will one of our existing schemes work? - are others out there? Develop a “true” volumetric rendering option Find the source of 3D rendering anomalies on VTK
AVID - Acknowledgements Rachael Brady – VTG Allen Song – BIAC Melanie Wright – Human Simulation Lab Josh Bizzell – BIAC