The Bernard M Gordon Center for Subsurface Sensing

The Bernard M. Gordon Center for Subsurface Sensing and Imaging Systems Overview of Image and Data Information Management R 3 -Thrust David Kaeli, NU This work was supported by in part by the Gordon Center for Subsurface Sensing and Imaging Systems, under the Engineering Research Centers Program of the National Science Foundation (Award # EEC-9986821).

Cen. SSIS Image and Data Information Management § Addressing key research barriers in computational efficiency, embedded computing and image/sensor data management § Exploiting hardware/software acceleration/parallelization and Grid resources to enable new discovery in SSI applications § Producing an image/data repository and softwareengineered Subsurface Sensing and Imaging Toolsets § Developing enabling tools targeting system-level projects • Near real-time reconstruction and visualization • FPGA-based phase unwrapping and registration MVT

System Level Impact of R 3 Activities Bio-Med L 3 Enviro-Civil S 2 S 3 S 1 S 4 S 5 Validating L 2 Test. BEDs R 2 L 1 Fundamental Science R 3 R 1 Image and Data Information Management

R 3 Researchers Faculty/Staff § George Chen – MGH § James Hale – Spelman § Shawn Hunt - UPRM § David Kaeli - NU § Miriam Leeser – NU § Steve Jiang – MGH § Luis Jimenez - UPRM § Waleed Meleis - NU § Wilson Rivera - UPRM § Samuel Rosario – UPRM § Nayda Santiago – UPRM § Greg Sharp - MGH § Charles Steward – RPI § Miguel Velez - UPRM § Furong Yang – NU Graduate students § Malak Alshawabkeh – NU § Fatemeh Azmandian - NU § Burak Erem – NU § Carolina Gerardino – UPRM § Evangelous Kanoulos - NU § Avi Kelman - RPI § Diego Rivera – NU § Maria E. Sanchez-Castro - UPRM § Dana Schaa – NU § Alejandra Umana - UPRM § Yijian Wang – NU § Gehua Yang - RPI Jaime § Yeckle – UPRM § Juemin Zhang - NU

R 3 Researchers Undergraduate students § Carolyn Andrews - NU § Kelli Crosby – NU § Nicholas Dedual - NU § Danielle Edwards – Spelman § Buffy Johnson - Spelman § Rafal Norton - NU § Jason Selwyn – NU

Goals for Year 7 § Focus on sustainability - Won sustaining proposals, developed new industrial ties, submitted new proposals § NSF STTR, NSF MRI, NSF IGERT, NIH, industry § Continued support of existing software toolboxes and new toolbox development § Integrated 4 -D visualization and image tagging (Northeastern and Spelman) § Support of the SSI textbook effort with complementing tools § Expanded work on existing Cen. SSIS R 3 projects § Grid Computing – Load balancing - funded by MGH for yrs. 6 and 7 § Hyperspectral Image Analysis – Acceleration/semi-automatic parallelization § FPGA acceleration – QR Decomposition and phase unwrap § Increased utility of the image database (additional images and users) § Provided image database user classes § Continued collaboration with the S 2 group on tumor tracking Toolboxes

Sustainability Project Posters § NSF STTR on automatic parallelization of high level languages with Interactive Supercomputing Corp. § Poster R 3 B P 5 - "Semi-Automated Parallelization Using Star-P" § Utilizing Cen. SSIS applications to evaluate/demonstrate the efficiency of these tools § NSF affiliated project on acceleration of sparse matrix problems § Poster R 3 B P 1 - "Characterizing the Relationship Between ILUType Preconditioners and the Storage Hierarchy" § New projects on utilizing GPU technology to accelerate image reconstruction – Submitted IGERT and talking to AMD/ATI and NVIDIA § Poster R 3 B P 2 - "Digital Tomosynthesis Mammography Parallelization Using Graphics Processors“ § Poster R 3 A P 4 – "Fast and Parallel Implementation of Image Processing Algorithm Using CUDA”

Cen. SSIS Toolbox Posters § Multiview Tomography Toolbox (MVT) MVT § C++ fddlib: January 2003 (v. 1. 0) § July 2003 (v. 1. 1) § MATLAB mvt: October 2004 § 164 downloads of these tools in the past 2 years § Being used in the SSI Textbook and NSF STTR project § Poster R 3 B P 5 - "Semi-Automated Parallelization Using Star-P" § Rensselaer Generalized Registration Library (RGRL) RGRL § Released September 2004 § Effort has spawned a startup company § Poster R 2 D P 5 - "New Applications and Enhancements in the Theory of the GDB - ICP Registration Algorithm" § Hyperspectral Image Analysis Toolbox (HIAT) HIAT § 749 downloads in past 3 years § The range of algorithms included in the toolbox have been funded by NASA, Do. D, ARMY, and NSF § New work highlighted in poster: § Poster R 3 B P 6 - "New Developments and Application of the MATLAB Hyperspectral Image Analysis Toolbox"

The Image Tagging Interface A 2 -D X-ray image is read, and then notes are entered into the radiologist’s report

The Image Tagging Interface Instead, a 3 -D image can be annotated by the radiologist with searchable tags, so the notes can be integrated with the image in a relational database system. <feature type="Ellipse" label="Tumor"> <ellipse> <x. Center>277</x. Center> <y. Center>278</y. Center> <x. Radius>44</x. Radius> <y. Radius>34</y. Radius> </ellipse> </feature> Poster R 3 B P 4 - "Online Image Tagging System"

4 -D Radiotheraphy – 3 D + Time § New visualization tools to allow physicians to rapidly segment and identify regions of interest in 3 D or 4 D § Leverages SCI-Run (Univ. of Utah) – a visual programming problem solving environment § The total time for treatment planning can be reduced by allowing more efficient analysis of size, shape, and movement of tumors § Present work focuses on enabling accurate physical measurements in our visualization environment Poster R 3 B P 3 - "Development of Visualization Tools for 4 D CT Datasets"

Parallelization and Acceleration § Parallelization work at NU and UPRM is focused on developing a general set of Grid level tools § Tomosynthesis and classification are being used as the proof-ofconcept applications § Tomosynthesis reconstruction paper in Proc. of the 2006 IEEE International Conference on Parallel and Distributed Computing, coauthored by NU/MGH Poster R 3 A P 3 – “Design of Workflow Scheduler" Poster R 3 A P 5 - "Utilizing High-Performance Computing to Investigate Performance and Sensitivity of an Inversion Model for Hyperspectral Remote Sensing of Shallow Coral Ecosystems” § FPGAs are being used at NU to provide real-time processing closer to the sensor § Discrete Cosine Transform acceleration with the S 1 project to accelerate phase unwrapping § Floating point division and square root based on Taylor series for the variable precision floating point library Poster R 3 A P 2 - "Phase Unwrapping Using Reconfigurable Hardware" Poster R 3 A P 1 – “Floating-Point Divide and Square Root for Efficient FPGA Implementation of Image and Signal Processing Algorithms"

Cen. SSIS Image Database System Poster R 3 B – P 7 "The Cen. SSIS Web-Accessible Image Database System" • Over 17, 000 image files online • Offered 3 online classes to the Cen. SSIS research community

Impact on System Level Projects R 3 Project Parallel/Grid/GPU Processing Hyperspectral toolbox Image database 4 -D visualization S Project S 3 S 4 S 1 S 2 Impact Accelerating tomosynthesis reconstruction < 5 minutes on a $500 platform Accelerating HIAT code by more than 25 X – reducing parallel programming time by 10 X Developing customized database interfaces to facilitate rapid data entry Improving radiation treatment planning