rept2

7/31/2019 rept2

1/58

Page 1 of 58

2Q NIST ATP report by Dr Karron. 5/1/2002 3:56 AM

C A S I D M T S O F T W A R EA N D H A R D W A R E

D E V E L O P M E N T E F F O R T RE S E A R C H P R O G R E S S RE P O R T

S E C O N D Q U A R T E R ( 2 Q )

CCCOOOMMMPPPUUUTTTEEERRRAAAIIIDDDEEEDDDSSSUUURRRGGGEEERRRYYYIIINNNCCCOOORRRPPPOOORRRAAATTTEEEDDD333 000 000 EEE AAA SSS TTT 333 333 RRR DDD SSS TTT RRR EEE EEE TTT ,,, SSS UUU III TTT EEE 444 NNN

NNN EEE WWW YYY OOO RRR KKK ,,, NNN EEE WWW YYY OOO RRR KKK 111 000 000 111 666 UUU SSS AAA http://www.casi.net

TTT EEE LLL EEE PPP HHH OOO NNN EEE ::: 111(((222111222)))666888666888777444888

FFF AAA XXX :::111(((222111222)))444444888000222666111

EEE --- MMM AAA III LLL ::: KKKAAA RRR RRR OOO NNN@@@CCC AAA SSS III...NNN EEE TTT

Date of Publication: 5/1/2002 3:56 AM.

Address: 300 E. 33rd

St., Suite 4N, New York, NY 10016 USA.

Telephone/Fax: (212) 686-8748, (212) 448-0261.

Research Technical Progress report 02 for

NIST ATP COOPERATIVE AGREEMENT

70NANB1H3050Prepared by D. B. Karron, Ph.D.

Principle Investigator

7/31/2019 rept2

2/58

Page 2 of 58


1 Abstract................................................................................................................ 1:3

2 Progress Metrics ................................................................................................... 2:4

2.1 GANTT Charts ............................................................................................. 2:5

2.2 Flow Charts................................................................................................. 2:10

2.3 Resource Charts .......................................................................................... 2:133 Collaborations and Partnerships........................................................................... 3:21

3.1 SGI............................................................................................................. 3:21

3.2 City University of New York (CUNY)......................................................... 3:21

3.3 NYU Medical Center (Department of Radiation Oncology).......................... 3:22

3.4 Columbia University (Departments of Medical Informatics and Radiology) .. 3:22

4 Accomplishments Narratives............................................................................... 4:24

4.1 Accomplishments........................................................................................ 4:24

4.2 Infrastructure, Hardware and Software Systems Report ................................ 4:24

4.2.1 Network Hardware Overview: ............................................................. 4:25

4.2.2 Network Architecture: ......................................................................... 4:25

4.2.3 Requirements: ..................................................................................... 4:26

4.2.4 Shortcomings and Solutions:................................................................ 4:26

4.3 Guide to CASI software under development:................................................ 4:27

4.4 Graph Generation (Dr James Cox). .............................................................. 4:27

4.5 Wolberg Accomplishments.......................................................................... 4:28

4.6 Screen Shots ............................................................................................... 4:32

5 CASI Business Activities (Gurfein) ..................................................................... 5:415.1 ITAC .......................................................................................................... 5:41

5.2 InQTel Proposal.......................................................................................... 5:41

5.3 SGI............................................................................................................. 5:41

5.4 NYSTAR.................................................................................................... 5:42

6 Appendix............................................................................................................ 6:43

6.1 SGI Developers Contract ............................................................................. 6:43

6.2 InQTel Proposal.......................................................................................... 6:49

6.2.1 Proposal .............................................................................................. 6:49

6.2.1.1 II. ATP Grant .................................................................................. 6:49

6.3 ITAC Proposal............................................................................................ 6:53

6.3.1 Proposal .............................................................................................. 6:53

6.3.2 II. ATP Grant ...................................................................................... 6:53

6.4 Copy of JMIV Paper Acceptance Letter ....................................................... 6:58

7/31/2019 rept2

3/58

Page 3 of 58


6.4.1.1 JMIV Paper (As revised per JMIV reviewers).......Error! Bookmark notdefined.

1 ABSTRACT

Digital Morse Theory is a novel theoretical rubric fordescribing the topology of n-dimensional functions describedin pixels, as opposed to traditional Morse Theory, whichdescribes topology of function in analytic or continuous(Morse) spaces. Demonstration of the practical applicationof Morse Theory to hard problems of segmentation in clinicalMedical Modeling radiation (and eventually surgical) therapiesis the goal of this project. A major theme of this project ispreserve dimensional fidelity and accuracy from the imagesource in the segmented 3-D models. We intend to extend our

preliminary results in 3-D scalar images (Black and White CTimages and MRI images) to 3-D fused RGB color, CT, andMRI data into single 3-D model.

Significant technical hurdles remain to be overcome. We havemade tentative progress in overcoming our formidablechallenges. We are formally into our first year, or 1/6 of theproject. Medical imagery technology generates large images,and many of them. New theoretical findings, just being tested,indicate the promise of DMT to reduce the information densityof an image without distorting connectivity or topologicalrelationships in an image. Traditional Gaussian or otherblurring or multi resolution algorithms can complicatesegmentation because the blurring distorts or merges objectsthat are separated at higher resolutions. It is these relationshipsthat we need to preserve in DMT analysis, because clinicallysmall features can separate large objects. Blurring these objectscan make segmentation of these objects impossible or subjectto interpretation. Therefore, we need to process large imagesdown to a graph representation.

This report describes preliminary development of this graphingsoftware, and outlines our pathway to generating a softwaretool for graphical exploration and segmentation of large opaquebricks of data it to clinically useful dimensionally accuratemodels suitable for per patient therapeutic usage.

7/31/2019 rept2

4/58

Page 4 of 58


2 PROGRESS METRICS

The main work of this quarter was the development large scale networked based pixel services,

of rudimentary DMT graphs, and simple 3D display code. The interactive linkage between thegraph and the image is not yet implemented.

3D Isosurfaces as a segmentation technique suffer from problems selecting isosurfaces,and managing how they run together. Sample illustration of this problem is given below.We need to identify where two objects connect, and how to disconnect them by selectingthe proper isosurface seed (simple segmentation) or by building a barrier between theobjects (carving segmentation).

DMT Graph Generation and display by Dr Cox (generation) and Dr Wolberg.

Large Image Storage RAID built by Dr. Karron

Pixel sharing facility for large image memory mapping by Dr. Karron

The following section gives Microsoft Project 2000 snapshot GANTT, Flow, andResource Charts of our project underway. As these are big charts, fitting them to the pagemay make sections hard to read. Since this document is designed to be read in AdobeAcrobat Reader, you can zoom in and enlarge the detail on these charts in the softwarereader that you may not be able to see in the printed form on paper.

7/31/2019 rept2

5/58

Page 5 of 58


2.1 GANTTCHARTS

7/31/2019 rept2

6/58

Page 6 of 58


7/31/2019 rept2

7/58

Page 7 of 58


7/31/2019 rept2

8/58

Page 8 of 58


7/31/2019 rept2

9/58

Page 9 of 58


7/31/2019 rept2

10/58

Page 10 of 58


2.2 FLOW CHARTS

7/31/2019 rept2

11/58

Page 11 of 58


7/31/2019 rept2

12/58

Page 12 of 58


7/31/2019 rept2

13/58

Page 13 of 58


2.3 RESOURCE CHARTS

7/31/2019 rept2

14/58

Page 14 of 58


7/31/2019 rept2

15/58

Page 15 of 58


7/31/2019 rept2

16/58

Page 16 of 58


7/31/2019 rept2

17/58

Page 17 of 58


7/31/2019 rept2

18/58

Page 18 of 58


7/31/2019 rept2

19/58

Page 19 of 58


7/31/2019 rept2

20/58

Page 20 of 58


7/31/2019 rept2

21/58

Page 21 of 58


3 COLLABORATIONS AND PARTNERSHIPS

Currently CASI has not entered into any partnership agreements other than the specific

instances given below.

3.1 SGI

Sgi has provided two machines, one remains their property, and the other was ordered butnot paid for pending invoicing. We can only assume that despite repeated requests to payfor the equipment, SGI is in no hurry for payment. CASI must be careful not to spend thefunds allocated for this purchase, and not to let it turn into a loaner. CASI wants clearownership of the code we develop by developing the code only on equipment it clearly

owns. The developer organization has made an agreement with CASI, a copy of which isin the appendix. Since that agreement was signed, the developer organization has increasedthe discounts offered to participating developers from 50% to 70% for full systempurchases. A copy of the new developer discount schedule is appended. SGI has providedliberal technical support, and has been supportive of our effort. Dr Karron plans to spendsome time at SGI as a visiting scientist sometime this summer. The potential for kickbackor quid-pro-quo understandings exists with any large corporation. This must bescrupulously avoided. CASI expects to build its foundation system on large SGIsupercomputers, and SGI expects to sell systems through CASI. However, no specificactions on the part of CASI or SGI are predicated on any exchange of money, oragreements to purchase. CASI seem to be enjoying the benefits of being the golden boy

or pet developer of the SGI medical marketing organization.

3.2 CITY UNIVERSITY OF NEWYORK (CUNY)

CASI has not yet managed to come to a mutually beneficial agreement with CityUniversity of New York a/k/a CUNY either as a research collaborator-supplying faculty oras an incubator, supplying office and laboratory space... This does not mean that CASIcannot come to an agreement in the next project year. Part of the problem is that the PI (DrKarron) teaches at CUNY (specifically, he is an adjunct professor of computer science in

the City College of New York a/k/a CCNY, part of the City University of New York).Another mitigating situation is that all of the research and student staff at CASI is alreadyCCNY faculty or students, and the academic standing of the researchers attracts facultysupport on a personal level, instead of an institutional level, at significant cost savings toCASI and ATP.

7/31/2019 rept2

22/58

Page 22 of 58


3.3 NYUMEDICAL CENTER (DEPARTMENT OF RADIATION ONCOLOGY)

Our collaboration with NYUMC is based on support the clinical efforts of Dr BruceEllerin, MD PHD. Dr Ellerin works mainly in clinical prostate treatment, and isresponsible for managing the department caseload. He would be our most direct customeronce our software is ready for an unsophisticated computer user. Because of this, we havepushed back the starting date for our NYUMC collaboration until we can let Bruce loose onour computers without danger to either of us.

3.4 COLUMBIA UNIVERSITY (DEPARTMENTS OF MEDICAL INFORMATICS ANDRADIOLOGY)

A new exciting proposed collaboration between Columbia University (Dr Imielinska & Dr

Laine) and CASI (Dr Karron) will expand on our clinical thrust in prostate therapy. In ourproposal, we proposed to focus on Prostate and Breast as the clinical problems with thegreatest national benefit both economically and for the American standard of living. WithDrs Imielinska and Laines research on developing segmentation and visualization tools forSurgical Planning and Training for Retropubic Radical Prostatectomy, we see a naturalbasis for collaboration by mutual interests. Significantly, Dr Imielinska also worked withDr Karron in the Visible Human Project, where she is a PI, and Dr Laines radiologicalresearch makes use of Morse Theory. This team has significant computer and imageprocessing experience that our NYUMC collaborators lack. However, NYUMC is rightacross the street, and we plan to run cases right out of their clinic when the softwaredevelopment state warrants and we clear human subject protocols.

Prostate cancer is the most commonly diagnosed cancer in American men and the secondmost common cause of cancer deaths. Last year in the United States, it was estimated that184,500 men were diagnosed with prostate cancer and about 39,200 men will die of it.Over the last decade, the number of radical prostatectomies to treat men with newlydiagnosed prostate cancer has quadrupled. The rise in the number of radicalprostatectomies has occurred for several reasons. One reason is the dramatic increase inthe detection of localized disease attributed primarily to the use of serum prostate-specificantigen (PSA) testing in conjunction with digital rectal examination. More than 95% oftumors detected in aggressive early-detection programs are clinically organ-confined, thus,a higher proportion of newly detected tumors are potentially amenable to radical

prostatectomy. Radical retropubic prostatectomy has become the treatment of choice forotherwise healthy patients with clinically localized prostate cancer whose life expectancyexceeds 10 years. This technique is technically demanding with many perioperative andpostoperative complications. The anatomic location of the prostate contributes largely tothe technical difficulty encountered in this procedure. The prostate is located beneath thepubic symphysis making visualization and resection of the prostate and peri-prostaticstructures difficult.

7/31/2019 rept2

23/58

Page 23 of 58


Despite advances in our understanding of male pelvic anatomy, radical prostatectomyremains a difficult operation with a steep learning curve. With the rise in incidence oflocalized prostate cancer, the necessity for this operation has markedly increased. The goalof this project is to develop and evaluate tools for segmentation and 3D visualization of

male pelvic anatomy, from patient specific image data, for use in training urologists toperform radical prostatectomy.

The long-term goal of this research will be to improve and develop tools to assist andenable surgeons in minimally invasive procedures.

Tailoring a segmentation methodology for a specific application (A) (radicalprostatectomy), specific body region (B) (pelvis, abdomen), and imaging protocol (P) (e.g.Spiral CT scans of the abdomen and pelvis) will result in 3D visualization of patientanatomy that is best suited for a computer aided surgical tool.

7/31/2019 rept2

24/58

Page 24 of 58


4 ACCOMPLISHMENTS NARRATIVES

4.1 ACCOMPLISHMENTS

Dr Karron has been taking responsibility for building a state of the art image-processinginfrastructure supporting the growing image processing requirements of DMT. Key towinning clinical acceptance is intuitive interactivity, close coupling between the DMTgraph and the rendered image. This is required for rapid and repeatable segmentation. Weare in the process of scaling up the size of the images we can graph, and learningoptimization techniques to reduce the algorithm complexity to achieve the desiredgraphing.

Dr Karron has is developing a standard format based shared memory system that willenable users and programmers to attached to data pre-loaded into memory on a host

machine, and share that memory on the same machine or over a network. The data can beaccessed in various fashions, to reduce long memory accesses with the resultant cashinvalidation and memory cash thrashing. This tuning effort is being done in collaborationwith Silicon Graphics.

By using the new SGI system, older software developed by Dr Karron is coming back tolife. Dr Karron will present animation sequences that were too slow and large on theprevious casi computers now run fluidly and demonstrate the power and beauty of theSpiderWeb isosurface algorithm, on which DMT is being built on.

Dr Karron and Dr Cox have presented two professional meetings since the first Q report,namely the SIAM (Society of Industrial and Applied Math) Special Interest Group onMedical Imaging in Boston, as well as at CCNY Biology Department seminar series,hosted by Dr Karen Hubbard.

4.2 INFRASTRUCTURE,HARDWARE AND SOFTWARE SYSTEMS REPORT

In order to achieve the innovative high-end performance required of high-end image and3D modeling research, CASI elected to build its own computers and computer networkfrom almost scratch, from components. The crown jewel of the CASI development shop is

its two Octane 2 super workstations both with dual R14000 CPU. Additionally, eachcomputer has two network interfaces, a slow public interface, and a private gigabitnetwork, to segment high-speed interaction between computers from interaction betweeneach computer and the greater internet. This is done for security and performance issues.

CASI has built a super high performance image storage system out of redundant arrays ofinexpensive hard disks. CASI built two raids, a super high performance raid constructed

7/31/2019 rept2

25/58

Page 25 of 58


of 24 each fast access 34 gigabyte 15,000 RPM Seagate Cheetah disks, the fastest availablein the industry, and a mirror raid of 10 each immense 180 gigabyte slow 7200 rpmBarracuda disks.

In building such a high performance computer network, CASI stretches the state of the art

in both hardware and software. CASI has also created a serious noise problem from thehigh capacity cooling fans and the low capacity air conditioning and power suppliesavailable in Dr. Karrons home.

4.2.1Network Hardware Overview:

The onsite network consists of twelve workstations and two servers. Six workstations areIntel Xeon / Pentium 4 Processor-based machines, used primarily for office management,literature research, multimedia, and accounting. All six machines were purchased underthe NIST ATP Cooperative Program. Two workstations are Silicon Graphics Octaneworkstations, used solely for software development. One Silicon Graphics Octane waspurchased under the NIST ATP Cooperative Program, the other contributed by Dr Karron.Three workstations are Silicon Graphics Indigo machines, used as additional seats forsoftware development, with one running as a domain name server. All three Indigoworkstations were contributed by Dr Karron. One IBM ThinkPad Notebook was alsopurchased under the program. It is used primarily for remote office management anddevelopment. Two SCSI RAID Servers are employed for data storage. One waspurchased under the NIST ATP Cooperative Program, the other contributed by Dr Karron.The network is built on an Extreme Networks Summit 7i Switch, also purchased under theprogram.

4.2.2Network Architecture:

The current network architecture is divided into two Virtual Local Area Networks(VLAN), namely Public and Private. This is done for the purposes of isolating high-bandwidth traffic to the faster Private network, leaving the slower Public network for lessdemanding traffic. The Private network is also used for security, isolating data that shouldnot be publicly accessible.

The Public network consists of a 15 static internet protocol (IP) address block with ISDNservice, provided by Thorn Communications. Each workstation is assigned one static IPaddress over a 10/100Base-T Ethernet connection. Domain name serving is done on-site.

The private network consists of a Class-C address block (192.168.0.0-192.168.255.255),over a 1000Base-T Ethernet connection. Each workstation/device is assigned one staticPrivate address via a secondary network interface installed in the workstations. Additional

7/31/2019 rept2

26/58

Page 26 of 58


internet access is available over the Private network from cable modem connectionsprovided by Time Warner Cable and RCN.

4.2.3Requirements:

The network must be structured such that local data transfers and storage can be conductedreliably and efficiently. This is achieved using the 1000Base-T Ethernet connections overthe private network regulated by the Summit 7i Switch, and RAID servers.

In/Outbound traffic to/from the internet must be fast and always on, allowing forconstant on-site internet access and remote access to the local network for authorized users.This is achieved by employing a reliable and sufficiently robust Internet Service, on-siteDNS, and backup internet access via ISDN, cable modem, and dialup.

Sharing and interoperability of data between platforms (Windows 2000, UNIX, LINUX,etc.) must be available. Sufficient redundancy of platform bridging must also be present.This is achieved using various sharing applications such as Samba, NFS Share, Sharity, etc.

4.2.4 Shortcomings and Solutions:

The current network configuration fails to meet requirements such that data storage is not

reliable due to hardware failures caused by improper cooling and power outages. To fullyresolve this issue, the site must be relocated to a location fitted with proper air-conditioningand capable of meeting our power demands. Sufficient data-backup and backup-powersystems are also needed.

In/Outbound connections are not reliable since ISDN lacks the appropriate bandwidth tohandle our traffic volume. Cable modem access is only a partial solution. To fully resolvethis issue, a commercial grade T-1 connection must be employed, either through siteinstallation or through relocation to a site pre-equipped. Backup connections (ISDN, cable,dial-up, etc.) must remain available in case of failure.

The current public IP address block is not large enough to accommodate the currentnumber of workstations and devices. This will significantly hinder future growth of thenetwork if left unattended. To resolve this issue, a larger IP block must be purchased. Thisis easily available upon deployment of a T-1 connection.

7/31/2019 rept2

27/58

Page 27 of 58


Sharing of data and interoperability between platforms is not fully functional due tosoftware/hardware configuration problems and pending license purchases. Upon purchaseof required software licenses and proper configurations, sharing and interoperability will befully functional.

4.3 GUIDE TO CASI SOFTWARE UNDER DEVELOPMENT:

The main driver routine is ZoneSegmentor. It calls the Critical Detector module and thencalls the Zone Locator Module the output produced consists of the criticality tree file thatindexes a random access zone file (which contains cell lists for each zone). The cells arethe volume elements or voxels needed for constructing isosurfaces within each zone. Thezone database is used by the front-end software under development by Dr Wolberg.

4.4 GRAPH GENERATION (DR JAMES COX).

Dr Cox is working on the DMT decomposition preprocessing software that convertsimage data into DMT Tree and associated zones. Dr Cox is working with Dr Wolberg andDr Karron, who is working on the front-end visualization software that utilizes the DMTbackend, in optimizing the backend, and in integration of the entire system. This willcomplete project milestones 3 and 4.

Dr Cox wrote code to process scalar volume data that produces a list of DMT criticalitiesin the data.

Initially the code found maximum and minimum pixel criticalities in the data, whichconsisted of a stack of slides. The code was then expanded to include pixel saddlecriticalities and verified on an artificially constructed data set for which criticalities wereknown the program was then expanded to recognize interstitial criticalities induced by so-called ambiguous Voxel faces. These are the disambiguation points that induce a globaltopology change in the isosurfaces. The computation was again verified on severalartificially produced data sets (supplied by Dr Nazma Ferdous) for which the criticalitieswere known.

The code was then expanded to include critical isosets. These can occur when there are

groups of adjacent identically valued data points. These were also tested on generated(artificial) data. The code was then tested on actual image data from the Visible Human.

During February, Dr Cox worked on the code to compute the DMT criticality graph (tree)from the criticality information. This involved a substantial amount of coding. Thegenerated trees were verified against the generated data with known DMT decomposition.They were subsequently tested on image data.

7/31/2019 rept2

28/58

Page 28 of 58


Next, we wrote the code to generate the zone decomposition during the DMTdecomposition. Initially, for testing, the points in each zone were listed. Then I modifiedthe code so that each zone was written as a sorted list of cells (voxels) for processing bythe imaging software. I modified the DMT tree representation to make it a search datastructure.

We first tested the DMT decomposition with an implementation of SpiderWeb imagingsoftware using the OpenGL GLUT library. This code was developed by Dr Cox and DrFerdous. We subsequently interfaced with the new (and much improved) imagingsoftware of Dr Wolberg for further testing. The new imaging software incorporates anoptimized version of our SpiderWeb, together with sophisticated visualization toolsdeveloped by Dr Wolberg.

In March, I modified the DMT decomposition to create a single direct access file with thezones, with the DMT tree serving as the index data structure.

We included a computation of each zones volume. This volume is an approximationbased on the pixel (or voxel) size of each zone. Dr Wolberg has a more precisecomputation. Our volume computation suffices for testing purposes. Testing of the DMTpreprocessing and tuning of the code continues. The tests have included some RGB datathat Dr Karron has supplied, using Dr Karron's recently developed technique forconversion of RGB data into multiple related scalar volume data sets based on imagecontent.

Dr Karron has converted CT data from the VHP data set to TIFF format. Dr Karron and Ideveloped software (a filter) to read and convert the TIFF slices into the scalar data set thatthe DMT preprocessing requires. We then modified the code to produce a second DMTtree, in a simplified format for Dr Wolberg, who is working on a sophisticated tree displaysystem. I also began work with Dr Karron on a simplified DMT tree display system fortesting purposes.

Dr Cox is presently working on the DMT decomposition of the head from the VHP dataset. The size of the data requires some modification and tuning of the DMT code, and heis presently engaged in this task. He has already successfully partially tested criticality treeculling software that removes inessential nodes for a more manageable tool. This isimportant because blurring the data by traditional means distorts the topology connectivity.Frequently important separate medically significant objects are separated by small thinmembranes that disappear when an image is reduced by traditional multiresolutionblurring algorithms.

4.5 WOLBERG ACCOMPLISHMENTS

The objective of Dr Wolbergs software effort during the first part of this review periodthis period was to accurately compute the volume of a 3D polyhedral region using Stokestheorem. The purpose of this work is to build up an important validation suite that tests theaccuracy of the segmentations derived from Digital Morse Theory. Accurate volumecalculations serve as key members to this validation suite.

7/31/2019 rept2

29/58

Page 29 of 58


It is well known from calculus that Stokes theorem provides the formulas by whichsurface integrals can be transformed into line integrals. Since our surface geometry isrepresented as polyhedral models derived from voxel data, a variant of Stokes theorem isnecessary for our domain. We began by first deriving two equivalent approaches to thisproblem. This led to the derivation of two compact equations relating the volume to the

vertices of the polyhedral model. We opted to implement the streamlined scalar tripleproduct approach because it offers greater computational benefits, i.e., fewer calculations.

The software was tested on surfaces for which the precise volume is known analytically: asphere and a torus. The sphere tests the simple case of a convex object. The torus tests themore difficult case of a convex object. In both cases, the software first determines that thesupplied object constitutes a closed surface. This is done by checking that every trianglehas three neighboring triangles, i.e., one per side.

A graphical user interface was implemented to permit the user to read graphic files in avertex/face format and visualize it in real-time. The user is able to modify the resolution ofthe displayed surfaces and recomputed the volume instantly. In this case, resolutionchanges take place by subdividing or merging existing triangular faces. It is important to

note that we will ultimately handle a change of resolution by subdividing or merging voxeldata, and then extract the polyhedral data directly from the voxel.

Work was also conducted on a curve-editing tool that permits the outline of a serial sectionto be drawn using a Bezier curve. The area of that drawn curve is then computed using asummation of cross products computed among consecutive pairs of vertices. It can beshown that this approach is derivable from Green's Theorem.

It is planned that an animation be created that can help the user visualize the transitionfrom one layer of data to the next, and see its effect on volume. For this pedagogical tool,both Green's theorem and Stokes theorem are necessary to compute the volume over a 2Dregion and a 3D surface, respectively.

The objective of our software effort was the implementation of a graphical user interfaceto permit the user to read graphic files in a vertex/face format and visualize it in real-time.The user is able to modify the resolution of the displayed surfaces and recompute thevolume instantly. In this case, resolution changes take place by subdividing or mergingexisting triangular faces. It is important to note that we will ultimately handle a change ofresolution by subdividing or merging voxel data, and then extract the polyhedral datadirectly from the voxel.

This quarter I also began to write graphical user interface software that makes use ofquaternions. Quaternions are a useful mathematical construct that facilitates an intuitive3D rotation controller using 2D mouse input. It is important to note that counter-intuitiveresults often occur when trying to rotate 3D objects with a 2D pointing device becauserotation does not commute. That is, there is more than one way to rotate a sphere fromone orientation to another. Quaternions are widely acknowledged to offer the best solutionto this problem. The objective of this work is to permit the user to freely and intuitivelyrotate volumetric data without the familiar gimbal lock effects that plague conventionaltechniques based on Euler angles.

7/31/2019 rept2

30/58

Page 30 of 58


I also worked on the SpiderWeb algorithm for determining hit points in voxels and theresulting generation of triangles to render. The code also addresses the manipulation ofvolumes that are too large to fit in RAM. As a result, a preprocess is run whichdecomposes the volumetric data into interval trees that facilitate the rapid identification ofactive voxels that must be rendered in response to a user-specified isovalue. The

SpiderWeb algorithm has been integrated into the unified software test bed underdevelopment.

We endeavored the design and implementation of graphical user interface software thatmakes use of quaternions. The software interface consists of two large concentric circlesdrawn in the window.

The object to be rotated is construed to lie in a bounding sphere. The edge of the innercircle is interpreted to be the silhouette of the sphere. If the user moves the mouse insidethe inner circle, the sphere (and enclosed object) rotates according to the intuitivequaternion behavior. It is often common to select constrained rotation axes, such as the x-,y-, or z-axes. To facilitate this option, four small circles are drawn in the north, south, east,and west positions between the two large concentric circles. The following rules apply for

the determination of the rotation axis that is used, depending on where the user beginsdragging the mouse.

Starting point, P, for mouse dragging Rotation axis

------------------------------------- -------------

Inside inner circle line perpendicular. to vector

connecting previous

pt to current point

outside outer circle z-axis

north/south circles x-axis

east / west circles y-axis

Elsewhere between two concentric circles line perpendicular. to P and circle if the SHIFTkey is held down while the mouse is dragged, the rotation takes place using the object's(local) coordinate system. Otherwise, the rotation takes place about the camera/world(fixed) coordinate system.

Dr Wolberg started the design and implements a tool to display and navigate a criticalitytree. He wrote software to parse the tree data, as supplied by Dr Jim Cox. The parsingsoftware performs the following operations:

1) Allocates nodes[], an array of tree nodes. Each node contains the following info:

a) Volume: volume associated with node (available in tree input file),

b) Depth: level of the tree in which the node lies,

c) ChildNum: the number of children for that node,

d) Children: an integer array containing indices of the children.

7/31/2019 rept2

31/58

Page 31 of 58


e) childMax: the maximum number of children allowable in children[] (initially 8). If wetry to install more than 8 children for a node, we run a call to realloc() and update themax number to 16, or 24, as required dynamically).

2) While parsing the file and setting up nodes[], we simultaneously init depth[], an array of

tree depths (levels). For each entry in depth[], there is a data structure that contains thefollowing info:

a) nodeNum: how many nodes are present at that depth,

b) nodeMax: maximum allowable nodes that can be stored

c) nodes: an integer array that indicates which tree nodes are present at that depth.

Visualization software has been written to take the above data and draw it in one of severalmodes: flat 2D tree display, concentric circles (with the root at the center), and hemi-spherical mapping (with the root at the North Pole). The design and implementation of

tree visualization tool to display and navigate a criticality tree continued with theimplementation by Dr Wolberg of three different forms of tree visualization:

1) Flat 2D rendering with nodes lying along horizontal/vertical rows and columns.

a) 2) Concentric ring 2D rendering with the root located at the center and all childrenon the next outer concentric ring. This act as a bird's eye view of hemisphericrendering (see (3)).

b) 3) Hemispheric 3D rendering with tree draped upon hemisphere. The root lies onthe North Pole, the leaves lie along the equator.

The spacing between the nodes is determined by their respective volumes. I have alreadyimplemented precise volume computations based on the scalar triple product (a variant of

Stokes theorem). Those nodes that represent volumetric data with high volume occupy alarger segment of the hemisphere than low volume nodes. Various interface options areavailable to permit the user to rotate and navigate the tree without being overwhelmed bythe potentially large number of tree nodes. The program reads/parses a criticality tree file.When the user double-clicks on a drawn node in any of the above three methods, thedescendants of that node become minimized, i.e., they collapse into the node. Double-clicking again then exposes them. When a branch is minimized, the remaining tree isredrawn to best fit the new space. A flag can be used to override this feature. Quaternioncode is used to permit the user to intuitively manipulate the hemisphere.

The next stage of the code will couple the tree visualization with the volume. Pointing tonodes of the tree will render volumes that correspond to the node's isovalues. In addition

to rendering the subvolumes, they can be highlighted within the whole volume with theuse of transparency functions and cut-away views. The benefit of this approach is that aconvenient dual form of the volume will be available to the user for direct editing.

7/31/2019 rept2

32/58

Page 32 of 58


4.6 SCREEN SHOTS

flat.jpg: A flat 2D tree display. The root lies in the top-left corner.

All six children of the root are depicted as nodes in the next row.In general, the children of a node are always depicted in theSubsequent row. A unique color is applied to an entire branchStemming from the root. This display clearly illustrates theTopology of the tree. In a future version, the actual valuesof the volume and isovalues associated with the nodes will bedepicted with the use of scaled color value and radii, respectively.

7/31/2019 rept2

33/58

Page 33 of 58


ring.jpg: A flat 2D polar representation of the tree. The root lies in the center.Each successive generation of nodes is made to lie on a series ofconcentric rings. The outermost ring consists exclusively of leaves.

7/31/2019 rept2

34/58

Page 34 of 58


sphere1.jpg: A 3D representation of the tree. The ring.jpg data is essentially

draped onto a hemisphere. The root sits at the North Pole and the

Equator consists exclusively of leaves. This figure occludes all

Back-facing nodes.

7/31/2019 rept2

35/58

Page 35 of 58


sphere2.jpg: Same as sphere1.jpg except that the back-facing nodes are visible,

i.e., no blackface culling is applied. This version helps give the

User a sense of the complete, albeit cluttered, views of the tree.

7/31/2019 rept2

36/58

Page 36 of 58


volume1.jpg: Each node along a branch of the tree represents a volume associated with aparticular isovalue. This figure depicts one such volume.

7/31/2019 rept2

37/58

Page 37 of 58


volume2.jpg,

7/31/2019 rept2

38/58

Page 38 of 58


7/31/2019 rept2

39/58

Page 39 of 58


7/31/2019 rept2

40/58

Page 40 of 58


volume4.jpg: Several depictions of a volume rendered with incrementally

Larger isovalues. These images are associated with successive

Nodes along a branch of the criticality tree.

7/31/2019 rept2

41/58

Page 41 of 58


5 CASIBUSINESS ACTIVITIES (GURFEIN)

5.1 ITAC

In an effort to enable CASI to expand its horizons and to capitalize on the technicalprogress made via the ATP grant, two proposals were submitted to the IndustrialTechnology Assistance Corporation (ITAC), a non-profit economic development arm,funded by New York State and New York City.

The first proposal was for a broad-based effort, for $2.4 million, to pursue various potentialattractive applications of Digital Morse Theory (DMT). CASI proposed an iterative market

analysis/technical feasibility approach to be followed by applying the knowledgedeveloped from the activity funded by the ATP grant. The potential applications to beexamined ranged from several medical areas to reconnaissance, meteorology, metrology,mapping, commercial animation and decoding, among others.

5.2 INQTELPROPOSAL

A second proposal was submitted to In-Q-Tel, via ITAC, at their request, for an applicationto be chosen by In-Q-Tel. (In-Q-Tel is a venture capital company that is wholly owned by

the Central Intelligence Agency. Its role is to access new IT companies, solutions, andapproaches to address the CIAs priority problems.) This proposal, for $1 million, wouldlikely be in the field of reconnaissance or decoding, but that is not certain what applicationsthey have in mind yet.

5.3 SGI

CASI is currently a participant is SGIs Developer Program. Among other potentialbenefits, Developers receive substantial discounts (varying from 40-50% on many items)on purchases from SGI. CASI has already saved considerably via this program. (See the

attached agreement.)

CASI has been approached by SGI senior scientists at SGI for assisting CASI in our DMTeffort, enabling CASI to optimize its use of SGI equipment. There have been noagreements made specifically in this regard, verbal or written. Any prospective agreementswould be submitted for ATP approval. A group at SGI that does business development hasalso solicited CASI to provide equity based funding. As CASI is still in the basic research

7/31/2019 rept2

42/58

Page 42 of 58


phase, CASI feels it is premature to start selling the company until it has some customersand products.

5.4 NYSTAR

CASI is likely to submit a proposal in cooperation with Columbia University in the first Qof next fiscal year for joint NYSTAR funding of Columbia-CASI collaboration. Atechnical description of the collaboration is given above.

7/31/2019 rept2

43/58

Page 43 of 58


6 APPENDIX

6.1 SGIDEVELOPERS CONTRACT

7/31/2019 rept2

44/58

Page 44 of 58


7/31/2019 rept2

45/58

Page 45 of 58


7/31/2019 rept2

46/58

Page 46 of 58


7/31/2019 rept2

47/58

Page 47 of 58


7/31/2019 rept2

48/58

Page 48 of 58


7/31/2019 rept2

49/58

Page 49 of 58


6.2 INQTELPROPOSAL

Virtual Reality, Tele-operation, Simulation, and Advanced Imagery Research

300 East 33rd Street, Suite 4N, New York, New York, 10016

Telephone and Voice Mail: +1(212) 686 8748Fax: +1(212) 448 0261.

Electronic Mail: [email protected]/World Wide Web: http://www.casi.net

6.2.1 ProposalI. Overview: Computer Aided Surgery, Inc. (CASI), established in 1995, is a cutting-edge software developer that has emphasized medical research applications. CASI haswon a number of past grants from the Defense Advanced Research Projects Agency(DARPA) and the National Institutes ofHealth (NIH). In October of 2001, CASI wasawarded an Advanced Technology Program (ATP) grant from the National Institute of

Standards and Technology (NIST). This grant is aimed at generating software usingDigital Morse Theory (DMT) for enabling accurate and precise segmentation ofanatomical organs from difficult medical imagery. Essentially, the goal, as cited in thegrant, is for this software to delineate edges between adjacent soft tissue organs. Theprimary applications are for radiation oncology planning and for surgical planning. Thebenefit to be derived from this software will relieve the radiation oncologist from thetedious and time-consuming effort of carefully circumscribing the volume to be treated; itwill yield a more accurate and repeatable volume. When reduced to practice, this willresult in significant cost savings to the payers as well as attractive growth and income forCASI. ATP funds are restricted for direct technical research and development costs only.Additional funds are now needed for the overhead costs of this project and the

commercialization of this technology for these medical applications and a wide range ofother applications.

6.2.1.1 II. ATP Grant

II.a. the Goal of the ATP Grant: The goal of the ATP grant is to generate the algorithmsand thereby the DMT software necessary for segmenting soft tissue organs from eachother, enabling the accurate automated circumscription of anatomical organ volume. Thetypical modality is CT imagery, using Houndsfield numbers (density). The density

between adjacent organs is very similar, making this a non-trivial problem. (Note: Thereare competing products in the marketplace, but they are not useful in soft tissue adjacent tosoft tissue organs, but only in soft tissue adjacent to bone.) In addition to developing theDMT software, CASI will validate the software and will have clinical validation performedat NYU Medical Center.

II.b. Potential DMT Applications: In addition to the medical applications noted above,there are numerous potential timesaving applications where DMT can be applied, yielding

7/31/2019 rept2

50/58

Page 50 of 58


superior results. There are many fields where the raw data imagery is either diffuse, lacksresolution, and contains low contrast, among other shortcomings. DMT may be aneffective tool for a wide range of imagery applications. These include satellite imagery forreconnaissance, mapping, geodesy, surveying, and topography; meteorology; air trafficcontrol; medical education and training; medical animation; entertainment industry

animation and fluid mechanics, among many others.

II.c. the ATP Advantage: Piggybacking on the technical foundation to be derived fromthe ATP activity, CASI proposes to apply DMT software to the applications of interest toIn-Q-Tel. Some of this effort may be accomplished in parallel to the ATP activity.

III. Benefits of DMT Software

IIIa. Economic benefits are expected to be very significant. Taking the example ofradiation therapy planning for prostate cancer, it takes a radiation oncologist about an hour

to carefully circumscribe the volume of the prostate, using a cursor on a computer monitorimage. This is done slice-by-slice and is very tedious. Medicare currently pays $770 perradiation plan. DMT software would enable this cost to be around $250, saving over $500per patient. On a national scale, with 200,000 new prostate cancer cases in 2001, thiswould amount to a saving of over $100 million per year just for prostate cancer.

III.b. Clinical benefits are numerous. The physician will have more time to spend withpatients. The physician will be relieved of a tedious task. The software should provide amore accurate and repeatable volume, thereby avoiding the predisposition of the physicianof being either too generous (radiating too much and affecting adjacent organs, e.g.,

bladder, rectum, with harsh results) or conversely to plan on radiating too little and leavingtoo many cancer cells untreated. The patients overall quality of care will be markedlyimproved and the physician will have more time to attend to patients.

III.c Collaborators For radiation oncology, CASI will be collaborating with universityacademic medical scientists. For other disciplines, CASI will be engaging collaborators,who are bellwether hands-on state-of-the-art researchers. Specifically, for satellite imagingwe intend to engage specialists involved in reconnaissance and mapping, such as theNational Reconnaissance Office (NRO), and the National Imagery and Mapping Agency(NIMA). For meteorology, experts from the National Oceanic and AtmosphericAdministration (NOAA) will be invited to be collaborators.

III.d. generally it is expected that other applications will benefit from both the time savingaspects and the accuracy and repeatability of DMT software.

7/31/2019 rept2

51/58

Page 51 of 58


IV. Need for Funds: Restrictions on our current ATP grant necessitate additional co-funding, as the ATP funds may not be used for commercialization, market surveys or anyfunction not directly related to the approved technical proposal.

V. Use of Funds: The primary use of funds will be for:

a. Investment in technology infrastructure support equipment including computers,high speed Internet, high quality photographic imaging printers, data backup anduninterruptible power supplies;

b. Adapting the software developed for the ATP activity to the three mostpromising applications of interest to In-Q-Tel;

c. Examining the feasibility of distributing the software via Internet;d. Hiring staffe. Leasing data port-equipped specially air-conditioned computer lab facilities;f. Building patent position;

g. Marketing of products via license and /or Internet ASP.

VI. Rationale and Schedule of Activities and Expenditures:

a. ATP funds may be only applied to the specific research tasks delineated inthe grant. This applies to both activities and equipment. Additionalinfrastructure equipment is needed to expand our commercial horizons.

b. The aim is for CASIs intellectual property portfolio to be diversified forproduct balance and future growth. . . In addition to adapting the DMT

software for other (non-medical) applications, CASI will engageprofessional consultants in order to get a hands-on understanding of theultimate users needs. As an example, if meteorology is selected, we willengage a meteorology concern to enable CASI to fully understand the needsof a meteorologist in analyzing weather imagery/data.

c. A data communications design feasibility study will be performed todetermine suitability of establishing an Application Service Provider fordistributing products. This is one of the two primary approaches that willbe pursued to market/distribute the DMT software; the other being licensingto software companies that specialize in the various application

technologies.

d. Support staff will be required. CASI will hire a marketing professional toimplement marketing effort. CASI will employ an administrator and abookkeeper, which will manage fiscal reporting and audit requirements.

7/31/2019 rept2

52/58

Page 52 of 58


e. In contemplation of hiring personnel and obtaining high-poweredcomputing equipment, CASI will move into suitable commercial / industrialspace. Requirements for this space are 10 workstations for scientists,technologists, programmers, management, and support specialists.

f. To preserve and optimize the intellectual property developed by this projectwe will engage patent counsel services. Developing patentable intellectualproperty is key to attracting private capital investment and assures CASIsability to license its technology.

g. In order to educate, promulgate, and disseminate the knowledge developedfrom this project, many papers will be presented at national symposia, andinternational meetings. Hands on demonstrations at conference exhibitionswill help accomplish dissemination of this technology.

VII. Key Personnel: (See attached bio sketches)

Dr. D.B.Karron-CEO & Chief Technology Officer has been active incomputer science for medical applications for over fifteen years. A formerresearch professor of Surgery at NYUMC and current professor of computerscience specializing in BioInformatics at CCNY, he is the PrincipalInvestigator of the current ATP grant. Dr. Karron has broad knowledge ofthe needs of the medical disciplines. He co-invented DMT with Professor JL Cox. Dr. Karron has been awarded previous grants from DARPA andNIH.

E. Gurfein-Chief Operating Officer has over 25 years of business experiencein scientific and technical companies. He has an undergraduate degree inmathematics, with advanced training from Courant Institute. He earned anMBA and managed technical projects at Sperry-Rand, General Precision,Perkin-Elmer Corporation and Engelhard Industries. He has extensivetechnical marketing and commercialization background. He was director oftechnical marketing for the Hubble Space Telescope. Mr. Gurfein hastechnical familiarity with optical systems and imagery, laser technology andapplications, inertial guidance systems, ballistic missile operations andcatalytic combustion, among others.

7/31/2019 rept2

53/58

Page 53 of 58


6.3 ITACPROPOSAL

Virtual Reality, Tele-operation, Simulation, and Advanced Imagery Research

300 East 33rd Street, Suite 4N, New York, New York, 10016

Telephone and Voice Mail: +1(212) 686 8748Fax: +1(212) 448 0261.

Electronic Mail: [email protected]/World Wide Web: http://www.casi.net

6.3.1 Proposal

I. Overview: Computer Aided Surgery, Inc. (CASI), established in 1995, is a cutting-edge software developer that has emphasized medical research applications. CASI haswon a number of past grants from the Defense Advanced Research Projects Agency(DARPA) and the National Institutes ofHealth (NIH). In October of 2001, CASI wasawarded an Advanced Technology Program (ATP) grant from the National Institute of

Standards and Technology (NIST). This grant is aimed at generating software usingDigital Morse Theory (DMT) for enabling accurate and precise segmentation ofanatomical organs from difficult medical imagery. Essentially, the goal, as cited in thegrant, is for this software to delineate edges between adjacent soft tissue organs. Theprimary applications are for radiation oncology planning and for surgical planning. Thebenefit to be derived from this software will relieve the radiation oncologist from thetedious and time-consuming effort of carefully circumscribing the volume to be treated; itwill yield a more accurate and repeatable volume. When reduced to practice, this willresult in significant cost savings to the payers as well as attractive growth and income forCASI. ATP funds are restricted for direct technical research and development costs only.Additional funds are now needed for the overhead costs of this project and the

commercialization of this technology for these medical applications and a wide range ofother applications.

6.3.2II. ATP Grant

II.a. the Goal of the ATP Grant: The goal of the ATP grant is to generate the algorithmsand thereby the DMT software necessary for segmenting soft tissue organs from eachother, enabling the accurate automated circumscription of anatomical organ volume. Thetypical modality is CT imagery, using Houndsfield numbers (density). The densitybetween adjacent organs is very similar, making this a non-trivial problem. (Note: Thereare competing products in the marketplace, but they are not useful in soft tissue adjacent to

soft tissue organs, but only in soft tissue adjacent to bone.) In addition to developing theDMT software, CASI will validate the software and will have clinical validation performedat NYU Medical Center.

II.b. Potential DMT Applications: In addition to the medical applications noted above,there are numerous potential timesaving applications where DMT can be applied, yieldingsuperior results. This plan addresses some of the more compelling applications. There are

7/31/2019 rept2

54/58

Page 54 of 58


many fields where the raw data imagery is either diffuse, lacks resolution, and contains lowcontrast, among other shortcomings. DMT may be an effective tool for a wide range ofimagery applications. These include satellite imagery for reconnaissance, mapping,geodesy, surveying, and topography; meteorology; air traffic control; medical educationand training; medical animation; entertainment industry animation and fluid mechanics,

among many others.

II.c. the ATP Advantage: Piggybacking on the technical foundation to be derived fromthe ATP activity, CASI plans to apply DMT software to the most compelling of theapplications cited. While all of the applications will be examined, finally the three mostattractive ones will be given a highly comprehensive evaluation. In addition to technicalconsideration from among the various modalities, i.e., visible spectrum, infrared, color,Houndsfield number, etc., for each application, a market survey will be performed for all.Some of these investigations will be performed in parallel with the ATP work.

III. Benefits of DMT SoftwareIIIa. Economic benefits are expected to be very significant. Taking the example ofradiation therapy planning for prostate cancer, it takes a radiation oncologist about an hourto carefully circumscribe the volume of the prostate, using a cursor on a computer monitorimage. This is done slice-by-slice and is very tedious. Medicare currently pays $770 perradiation plan. DMT software would enable this cost to be around $250, saving over $500per patient. On a national scale, with 200,000 new prostate cancer cases in 2001, thiswould amount to a saving of over $100 million per year just for prostate cancer.

III.b. Clinical benefits are numerous. The physician will have more time to spend withpatients. The physician will be relieved of a tedious task. The software should provide amore accurate and repeatable volume, thereby avoiding the predisposition of the physicianof being either too generous (radiating too much and affecting adjacent organs, e.g.,bladder, rectum, with harsh results) or conversely to plan on radiating too little and leavingtoo many cancer cells untreated. The patients overall quality of care will be markedlyimproved and the physician will have more time to attend to patients.

III.c Collaborators As with radiation oncology, CASI will be collaborating withuniversity academic medical scientists. CASI will be engaging collaborators, who arebellwether hands-on state-of-the-art researchers. Specifically, for satellite imaging we willengage specialists involved in reconnaissance and mapping, such as National

Reconnaissance Office (NRO), National Imagery and Mapping Agency (NIMA). Formeteorology, experts from the National Oceanic and Atmospheric Administration (NOAA)will be invited to be collaborators.

III.d. generally it is expected that other applications will benefit from both the time savingaspects and the accuracy and repeatability of DMT software.

7/31/2019 rept2

55/58

Page 55 of 58


IV. Need for Public Funds: Restrictions on our current ATP grant necessitate additionalco-funding, as the ATP funds may not be used for commercialization, market surveys orany function not directly related to the approved technical proposal. The nature of DMTsoftware development, while extremely encouraging, is not sufficiently mature to attractprivate equity funds. Therefore, on the expectation that DMT will be of great broad-based

economic benefit, CASI is seeking co-funding in the range of $2 million dollars. Thesefunds will be employed as described below.

V. Use of Funds: The primary use of funds will be for:

h. Investment in technology infrastructure support equipment including computers,high speed Internet, high quality photographic imaging printers, data backup anduninterruptible power supplies;

i. Market surveys for each of six (6) potential DMT applications;

j. Adapting the software developed for the ATP activity to the three mostpromising applications;

k. Examining the feasibility of distributing the software via Internet;l. Hiring staffm. Leasing data port-equipped specially air-conditioned computer lab facilities;n. Building patent position;o. Presenting papers, attending seminars and publicity;p. Marketing of products via license and /or Internet ASP.

VI. Rationale and Schedule of Activities and Expenditures:

h. ATP funds may be only applied to the specific research tasks delineated inthe grant. This applies to both activities and equipment. Additionalinfrastructure equipment is needed to expand our commercial horizons.

i. CASI will engage the market research arm of the Industrial TechnicalAssistance Corporation (ITAC), as an independent marketing researchconsultant, for designing a market survey. From over a dozen initialcandidates, six potential applications will be selected based on technical,marketing and profitability factors.

j. The aim is for CASIs intellectual property portfolio to be diversified forproduct balance and future growth. A market research study will beperformed in parallel on each of the six provisional (6) applications areas.This research will be done along with the ITAC market research activity.Based on these research results, the three most promising applications willbe selected for product development. In addition to adapting the DMTsoftware for each application, CASI will engage professional consultants inorder to get a hands-on understanding of the ultimate users needs. As an

7/31/2019 rept2

56/58

Page 56 of 58


example, if meteorology is selected, we will engage a meteorology concernto enable CASI to fully understand the needs of a meteorologist inanalyzing weather imagery/data.

k. A data communications design feasibility study will be performed to

determine suitability of establishing an Application Service Provider fordistributing products. This is one of the two primary approaches that willbe pursued to market/distribute the DMT software; the other being licensingto software companies that specialize in the various applicationtechnologies.

l. Support staff will be required. CASI will hire a marketing professional tosupervise the various marketing studies and then to implement marketingeffort. CASI will employ an administrator and a bookkeeper, which willmanage fiscal reporting and audit requirements.

m. In contemplation of hiring personnel and obtaining high-poweredcomputing equipment, CASI will move into suitable commercial / industrialspace. Requirements for this space are 10 workstations for scientists,technologists, programmers, management, and support specialists.

n. To preserve and optimize the intellectual property developed by this projectwe will engage patent counsel services. Developing patentable intellectualproperty is key to attracting private capital investment and assures CASIsability to license its technology.

o. In order to educate, promulgate, and disseminate the knowledge developedfrom this project, many papers will be presented at national symposia, andinternational meetings. Hands on demonstrations at conference exhibitionswill help accomplish dissemination of this technology.

VII. Key Personnel: (See attached bio sketches)

Dr. D.B.Karron-CEO & Chief Technology Officer has been active incomputer science for medical applications for over fifteen years. A formerresearch professor of Surgery at NYUMC and current professor of computer

science specializing in BioInformatics at CCNY, he is the PrincipalInvestigator of the current ATP grant. Dr. Karron has broad knowledge ofthe needs of the medical disciplines. He co-invented DMT with Professor JL Cox. Dr. Karron has been awarded previous grants from DARPA andNIH.

E. Gurfein-Chief Operating Officer has over 25 years of business experiencein scientific and technical companies. He has an undergraduate degree in

7/31/2019 rept2

57/58

Page 57 of 58


mathematics, with advanced training from Courant Institute. He earned anMBA and managed technical projects at Sperry-Rand, General Precision,Perkin-Elmer Corporation and Engelhard Industries. He has extensivetechnical marketing and commercialization background. He was director oftechnical marketing for the Hubble Space Telescope. Mr. Gurfein has

technical familiarity with optical systems and imagery, laser technology andapplications, inertial guidance systems, ballistic missile operations andcatalytic combustion, among others.

7/31/2019 rept2

58/58

Page 58 of 58

6.4 COPY OF JMIVPAPER ACCEPTANCE LETTER

Documents

rept2