Haptics-Assisted 3D Lasso Drawing for Tracts-of-interest Selection in DTI Visualization Wenjin Zhou * Stephen Correia † David H. Laidlaw ‡ Brown University Figure 1: Our interface selecting the uncinate fasciculus (UF) using 3D lasso: the user places two lassos around the UF and singles out the structure with an AND operation. The final result is shown in purple. ABSTRACT We present a new haptics-assisted 3D lasso drawing interface for selecting tracts-of-interest (TOI) in diffusion tensor imaging (DTI) in fishtank virtual reality (VR). This interface brings TOI selection tasks into 3D stereo VR with higher-input bandwidth devices. In the system, the 6D input Phantom device lets the user make selec- tions by drawing a smooth 3D lasso with haptics constraint assis- tance directly in the 3D space. The hand tracker lets the user use hand gestures to rotate and zoom the model. Users also reported that making selections by drawing 3D lassos is easy, as it resembles pointing out structures to a collaborator with their fingers but has higher precision. Users also remarked that VR helped them appre- ciate the three-dimensional structure of the fiber tracts more eas- ily, and they gained more confidence in identifying the structures and the area they project in the brain. Working in the VR environ- ment reduces the navigation time of the TOI selection task, a key challenge in TOI selection tools. Users were able to segment out tortuous structures that are often time-consuming to select using traditional rigidly shaped volume-of-interest (VOI) fiber pathway selection tools. Index Terms: J.3 [LIFE AND MEDICAL SCIENCES]: Med- ical information systems—; I.3.8 [COMPUTER GRAPHICS]: Applications— 1 I NTRODUCTION Diffusion tensor imaging (DTI) is a magnetic resonance (MR) imaging technique that makes possible the non-invasive exploration of fibrous white matter (WM) structures within the human brain. Understanding the WM structure is crucial for studying such dis- eases as HIV, Alzheimer’s disease, and brain tumors. Current DTI visualization and analysis have focused on rendering fiber tracts and segmentation of WM trajectories into anatomically meaning- ful bundles [8]. The densely sampled fiber tracts representing WM structures tend to be visually cluttered, making it difficult for brain scientists to understand the data. Automatic segmentation methods * e-mail: wzhou@cs.brown.edu † e-mail: SCorreia@butler.org ‡ e-mail: dhl@cs.brown.edu impose a rigid, possibly inaccurate model of which WM pathways belong to which bundles. Instead, enabling brain scientists to in- teract with the visualization models and manually perform segmen- tation by selecting TOI is an important interaction on which many published clinical research studies using DTI tractography, such as [4], have relied. Most interfaces for selecting TOI interactively depend on devices such as mice and pens, which offer only two degrees of freedom for manipulation of 3D VOI, and they also use standard 2D monitor screens to display the 3D structure of the brain. Brain scientists have noted that they must constantly rotate the brain model in order to retain their mental picture of the 3D structure presented on the 2D screen. The extensive navigation time is also a key drawback in the current TOI selection tools. Here we present a new haptics- assisted interface that uses a 3D lasso for selecting TOI in 3D stereo VR. 2 RELATED WORK Segmentation and clustering of WM pathways by interactively se- lecting TOI has become a popular way for brain scientists to test their hypotheses of WM connectivity and functionality, and a num- ber of tools have been developed for interactively assembling TOI into anatomically meaningful fiber bundles. Sherbondy et al. [7] [2] defined volume-of-interest to let users interactively group WM pathways going through these regions. Blaas et al. [3] presented a real-time fiber bundle selection system based on multiple con- vex selection objects. However, in these methods, the shape of the volume the user can define is rather limited, making selection cum- bersome due to the curved and complex nature of WM pathways. CINCH, a 2D marking interface for selecting 3D TOI, has been de- veloped recently [1] to alleviate the difficulties involved in selecting complex 3D structures using only commodity input devices, which offer only two degrees of freedom. However, CINCH users still find navigating and locating the desired WM structure undesirably time-consuming. It is also reported that selection using 2D plane intersection, as in CINCH, requires extensive neurological knowl- edge because the user loses the 3D context while working in the 2D space. Here we describe a new haptics-assisted 3D lasso-drawing in- terface for selecting TOI in DTI. Our system brings TOI selection tasks into 3D fishtank VR with higher-input-bandwidth devices. Our system greatly reduced navigation time in the TOI selection task and enabled users to identify the WM structures in the brain