Motor learning has been actively studied to examine exactly how humans learn a motor skill by building sensorimotor maps that transform the sensory inputs into the motor output. One of the crucial problems involved in this transformation is the “degrees of freedom problem”, a problem of effectively eliminating the redundancy of transforming high-dimensional joint space motor commands to the three-dimensional operational (endpoint) space movement.

We hypothesized that the motor system exploits the redundant degrees of freedom to fine-tune the secondary factors of the motor cost, such as manipulability and dexterity. We believe the redundant kinematic factors should be concurrently and systematically assessed, rather than constrained, in motor learning studies. To accomplish this goal, we incorporated 3D motion capture and musculoskeletal models with standard motor learning studies.

we developed an automatic heartbeat detection method that can detect and track the heartbeats of immobilized, ventrally-positioned (VP) zebrafish larvae without requiring direct observation of the larva heart. Our zebrafish heart rate automatic method (Z-HRAM) detects and tracks localized zebrafish body deformation that is highly correlated with heart movement. multi-resolution dense optical flow-based motion tracking is used to calculate this deformation, and principle component analysis technique is applied to identify heartbeats. Here we present results of Z-HRAM on estimating heartbeat rate and beat-to-beat intervals from video recordings of larvae undergoing seizures, which were of low resolution and low frame rate (3 to 4 fps). Heartbeats detected from Z-HRAM were compared to those determined through frame-by-frame manual video inspection. Validation that the localized deformation corresponds to larva heartbeats was performed through comparison to recorded electrocardiogram data. The results show that Z-HRAM is a robust and fast tool for studying larva cardiac function in general laboratory conditions.

Problem:

We propose to use ultrasound imaging as an alternative modality to quantify extraocular muscle deformation. Ocular ultrasound is the most common imaging modality used by ophthalmologists to examine ocular anatomy and diseases. Its interactive and rapid imaging capabilities make it suitable to look at eye muscle dynamics.

In this project, we describe our experimental setup in acquiring high resolution ultrasound images and eye movement data. We also present preliminary results on estimating the extraocular muscle motion. Our long term goal is to examine whether there exists differential movement between the two layers in each rectus muscle.

This project proposes a novel automatic method to segment extraocular muscles and orbital structures. Instead of conventional segmentation at the pixel level, superpixels at the structure level were used as the basic image processing unit. A region adjacency graph was built based on the neighborhood relationship among superpixels. Using Normalized Cuts on the region adjacency graph, we refined the segmentation by using a variety of features derived from the classical shape cues, including contours and continuity.

To demonstrate the efficiency of the method, segmentation of Magnetic Resonance images of fives healthy subjects was performed and analyzed. Three region-based image segmentation evaluation metrics were applied to quantify the automatic segmentation accuracy against manual segmentation. Our novel method could produce accurate and reproducible eye muscle segmentation.

Spine surgeries are high risk operations which require the surgeons to have ample experiences. For young surgeons, effective and extensive training is critical.

This project presents a real time haptic spine surgical simulator that will be used to train residents, fellows and spine surgeons in a hospital training program. It provides a realistic environment for the trainees to practice spine surgeries and has the advantages of being interactive, low-cost, representative, and repeatable over conventional training approaches. Haptic Phantom offers the users force feedback, differentiating our system from other screen-based training systems. Computational efficiency was achieved by developing advanced graphical rendering methods. The volumetric data was classified into surface voxel cloud and inner voxel cloud by the adjacency graph which stored the relationship among voxels. To speed up the collision detection and real time rendering between the virtual surgical tools and the lumbar model, Octree-based algorithms and GPU technique were applied. To enhance the physical realism, three dimensional lumbar vertebrae models were reconstructed from CT images and associated with non-homogeneous bone density such that the rendered model best represents the spine anatomy and mechanics. We demonstrate system performance by conducting pedicle screw insertion.

Determining shape of the eyeball is important to diagnose eyeball disease like myopia. In this project, we present an automatic approach to precisely reconstruct three dimensional geometric shape of eyeball from MR Images. The model development pipeline involved image segmentation, registration, B-Spline surface fitting and subdivision surface fitting, neither of which required manual interaction. From the high resolution resultant models, geometric characteristics of the eyeball can be accurately quantified and analyzed. In addition to the eight metrics commonly used by existing studies, we proposed two novel metrics, Gaussian Curvature Analysis and Sphere Distance Deviation, to quantify the cornea shape and the whole eyeball surface respectively. The experiment results showed that the reconstructed eyeball models accurately represent the complex morphology of the eye. The ten metrics parameterize the eyeball among different subjects, which can potentially be used for eye disease diagnosis.

Cloud computing is attracting the attention of the scientific community. In this project, we develop a new cloud-based computing system in the Windows Azure platform that allows users to use the Zeolite Structure Predictor (ZSP) model through a Web browser. The ZSP is a novel machine learning approach for classifying zeolite crystals according to their framework type. The ZSP can categorize entries from the Inorganic Crystal Structure Database into 41 framework types. The novel automated system permits a user to calculate the vector of descriptors used by ZSP and to apply the model using the Random Forest™ algorithm for classifying the input zeolite entries. The workflow presented here integrates executables in Fortran and Python for number crunching with packages such as Weka for data analytics and Jmol for Web-based atomistic visualization in an interactive computing system accessed through the Web. The compute system is robust and easy to use. Communities of scientists, engineers, and students knowledgeable in Windows-based computing should find this new workflow attractive and easy to be implemented in scientific scenarios in which the developer needs to combine heterogeneous components.