Medical Imaging

medical_imaging.jpg

 

 

 

 

  

Work within the medical field concentrates on developing 3D surface and volumetric models for estimating, analyzing and classifying shapes within medical images.

Lab students Yunfeng Sui and Beibei Zhou accompanied Dr. Andrew Willis to this years SPIE Conference on Medical Imaging. They presented their work entitled "Improving Inter-fragmentary Alignment for Virtual 3D Reconstruction of Highly Fragmented Bone Fractures"
at the conference. This work is sponsored by the NIH under an R21 grant and seeks to quantify the severity of traumatic, i.e., highly fragmented, bone fractures involving many bone fragments. Such fractures often pose difficult problems for orthopeadic surgeons and have a higher instance of PTOA (Post-Traumatic Osteo-Arthritis). Our paper presented algorithms that semi-automatically aligns virtual models of bone fragments extracted from CT (computerized tomography) data. At the moment, we are validating results using surrogate bone material which can be measured accurately using a laser scanner and appears similar to human bones in CT images. Details are available in our paper.

The picture shows the three of us at the conference. Left to right: Yunfeng Sui, Dr. Andrew Willis and Beibei Zhou.

This joint project between the University of Iowa Biomechanical Orthopaedics Laboratory and the University of North Carolina at Charlotte Machine Vision Laboratory seeks to develop computer software to aid in the virtual reconstruction of comminuted bone fractures, i.e., bone fractures that involve many bone fragments. University of Iowa researchers will obtain bone fragment measurements from two sources: (1) actual fracture cases experienced by their orthopedic surgeons and (2) artificially generated fractures using a "drop tower" device and special synthetic bone surrogate materials. Images of the bones are recorded by 3D laser scanning or 3D Computerized Tomography (CT) which serves as source data for the reconstruction software. Software algorithms developed at the UNCC Machine Vision Lab process the fracture data to detect, match, and subsequently reconstruct the bone fragments to discover the unknown geometric structure of the bone prior to the fracture event. Such systems are crucial components for quantifying and classifying bone fractures; a critical component to defining effective treatment strategies. Surgeons may also use such systems for pre-operative planning and for experimenting with novel reconstruction strategies. Such tools promise to aid in the development of new approaches to this problem that will reduce the occurence of Post-Traumatic Ostreo Arthritis (PTOA), a debilitating condition that commonly occurs in patients with such injuries.

This new joint Centers of Research Translation (CORT) project is headed by Joseph Buckwalter M.D., Chair of the Dept. of Orthopaedics and Rehabilitation at the University of Iowa. This expansive $7.3M project seeks to develop new methods of forestalling post-traumatic osteoarthritis (PTOA) through a multi-disciplinary translational approach including basic science, bioengineering, imaging, and clinical research. The central theme is that joint injuries initiate a sequence of biologic events that lead to PTOA and that new treatments of joint injuries will minimize these deleterious events and promote joint healing. The specific aims are to: 1) advance understanding of the pathogenesis of PTOA, 2) develop and refine reliable quantitative measures of severity of joint injuries, including measures of structural damage and biologic response to joint injury, and 3) apply the advances in understanding of the pathogenesis of PTOA and assessment of joint injury to new methods of forestalling PTOA.

Andrew Willis from the UNC Charlotte Machine Vision Laboratory is collaborating with researchers at the University of Iowa as part of the Biomechanics and Imaging Core, one of four core research thrusts in the project that concentrates on the biomechanical aspects of PTOA and is operated from the University of Iowa's Orthopaedic Biomechanics Laboratory. The collaboration uses medical image processing, pattern recognition and computer vision techniques to estimate the size, shape, and location of the bone fragments from 3D CT data. From this data, researchers seek to infer a variety of clinically significant information that can aid in effective treatment such as the energy needed to  generate the observed fracture and the severity of the fracture.

Relevance to Public Health: Osteoarthritis (OA) is the most common joint disease and is among the most important causes of pain, disability, and economic loss. About 12% of OA arises following joint trauma. The risk of OA ranges from 20% to 50% or more following many common joint injuries; and, despite evolving surgical methods of treating joint injuries, this risk has not decreased appreciably in the last 20 years. The basic science, bioengineering and clinical research in this CORT will lead to new biologic and improved minimally invasive operative treatments of joint injury that will forestall OA and thereby improve the lives of hundreds of thousands of people.

There are four CORT projects: 1. Cartilage Extracellular Matrix Fragments and Trauma-Induced Chondrolysis, an in vitro study that will identify pathways responsible for propagation of cell damage following injury; 2. Acute versus Chronic Mechanical Damage in the Etiology of PTOA, an experimental study that will define the role of loading of injured joints in causing OA, and new methods for preventing OA in injured joints; 3. Validation and Application of MRI Biomarkers in Assessing Articular Cartilage Health, a clinical and experimental study of non-fracture cartilage injury that will help define the ability of non-invasive measures to assess the severity of cartilage damage, that will identify which synovial fluid markers of acute joint injury reflect that damage, and that will test the hypothesis that decreased loading accelerates restoration of injured joint surfaces; and 4. Quantifying Injury Severity to Assess the Risk for Post-Traumatic OA, a clinical study of intra-articular fractures that will examine the hypothesis that new quantitative measures of the severity of structural joint injury predict clinical outcomes. This project also will conduct a multi-center study of the severity of joint injury, in preparation for clinical trials of molecular interventions to minimize the risk of OA following joint injury. The four projects will be supported by an administrative-biostatistics core, a biomechanics-imaging core, and a tissue and experimental modeling core.

New NIH Project on Automatic Reconstruction of Bone Fractures Fundeddon_thad_and_tom_small.jpg

The NIH recently approved funding for a joint project between the Department of Orthopaedic Biomechanics at the University of Iowa (Iowa City, IA) and the Department of Electrical and Computer Engineering at the University of North Carolina at Charlotte (Charlotte, NC). The joint work is sponsored as part of the Biomechanics and Imaging Core from a National Institute of Health P-50 grant. Principal investigators from the University of Iowa include Thomas Brown and Donald Anderson and Andrew Willis serves as the principal investigator for the University of North Carolina at Charlotte subcontract.

    The  project seeks to reconstruct complex high-energy fractures from volumetric 3D Computerized Tomography (CT) images of the damaged limb. The project will take place from September 2007 to September 2012. The press release from the University of Iowa is available here.

 

This project is joint work with co-PIs Tom Brown and Donald Anderson from the University of Iowa and is a part of a larger P50 CORT grant from the National Institute of Health (NIH), There are two main goals for our research contribution to the project. They are :

  1. To further develop and refine a 3-D puzzle-solving computer algorithm for automated virtual reconstruction of comminuted extremity fractures, given surface and volumetric segmentations of a set of dispersed bony fragments, and given their associated (pre-injury) template surface and volumetric segmentation. The algorithm’s output will be a probabilistically-based estimate of the location (position and orientation) from which each of the measured fragments originated from within the template.
  2. To collaborate in developing methodologies to quantify the amount of tissue disruption occurring during injury, due to motion of fragments from their initial anatomic positions to their (displaced) post-injury positions apparent on CT.

Motivation: Severe limb trauma often generates highly fragmented bones. Accurate reconstruction of the original unbroken bone from the fragments is a key factor in generating favorable outcomes for injury rehabilitation.  Below is an image showing different CT scans of broken limbs, each representing a different level of damage. They range from the most simple breaks (7-15 fragments), to the most complex ones (15- >30 fragments)

fracture_montage2.png

Overview: The aim of the project is to design a 3D interactive system capable of semi-automatically aligning fractured bone fragments. The semi-automatic approach shows promise for improving on the geometric alignment of multiple broken fragments which are difficult for surgeons to manually align. Below is a result from the existing bone fragment reassembly system which takes in user input to divide the fracture surfaces for matching it to the right surface. To the right is the result of the reassembled bone.

interactive_fracturesurf_segment.png case0_3fragments_aligned0.png

Implementation: The implementation is done using Java. The Graphical interface used is ShaRP, a 3D servlet used for rendering and manipulating 3D objects.

The present system for aligning the fragments requires human interaction for fracture surface segmentation. To eliminate this step, the concept of spin images is used to automatically identify surfaces which are potential matches.

Updates: The progress of this project will be updated periodically as it gets to completion. Right now I am working on generating spin images for multiple points on the fragment surfaces to be able to match them accurately.