CenSSIS Year Four Nuggets:      Ideas   ·   Tools   ·   People

 

Ideas: Connections Between Discoveries and Their Use to Society
- A CenSSIS NU-BU team has developed a new technique for imaging deep inside of turbid, diffuse media such as the breast or brain using a combination of light and ultrasound.
Figure 7A-7. CenSSIS graduate student Lei Sui (back) and Post-doc Emmanuel Bossy (front) work on the API system. In this technique, called Acousto-Photonic Imaging (API), a continuous laser beam illuminates a diffusive media where it interacts with a focused ultrasound beam. The light that passes through the ultrasound focus is "tagged" by the ultrasound allowing the researchers to distinguish it from the background light. This "tagged" light contains information about optical and mechanical properties of the tissue in the focal region. Scanning the ultrasound transducer allows the researchers to create images deep inside of tissue. A significant technical challenge in API has been in the detection of the "tagged" light, which is very weak with respect to the background signal. We have recently solved this problem by developing a novel photorefractive crystal (PRC) based detection scheme which allows them to obtain strong signals from the "tagged light" through coherent averaging over space. The discovery of enhanced detection of API signals using a PRC based sensor was disclosed at the recent meeting of Photonics West in San Jose, and was greeted with great enthusiasm by the community (see Figures 7A-7 and 7A-8).
Figure 7A-8. A typical API result (a) API signal observed under the focal pressure for a phantom with a reduced scattering coefficient of 3cm-1 (b) measurements with and without the PRC based detection system illustrating the signal enhancement.

- A CenSSIS RPI-MGH-BU team has developed new algorithms which compensate for patient and organ motions during delicate radiotherapy procedures. Initial clinical tests will commence at MGH and may extend to MSKCC.

The new 3D multi-object deformable segmentation algorithm is based on matching probability distributions of photometric variables, and incorporates learned shape and appearance models for the objects of interest. The main innovation over similar approaches is that there is no need to compute a pixelwise correspondence between the model and the image. This allows for a faster, more principled algorithm. The team has obtained promising results on difficult imagery for both synthetic images and 3D CT images of the male pelvis for the purpose of image-guided radiotherapy of the prostate (see Figure 7A-9)
Figure 7A-9. Segmentation results using a joint object model of the prostate, bladder and anterior rectal wall. The red contour shows the segmentation result at convergence. The blue contour shows the hand-drawn ground-truth contours supplied by a radiation oncologist.

- CenSSIS researchers at UPRM and WHOI have conducted a campaign in the US Virgin Islands to assess the health of deep sea coral reefs. CenSSIS image processing and mosaicing techniques enabled the team to determine that, surprisingly, deep water reefs are not being significantly eroded. Deepwater coral reefs in the US Virgin Islands may occupy a much larger area and be in better health than previously thought, based on evidence gathered by CenSSIS researchers at UPRM and WHOI using the SeaBED autonomous underwater vehicle. At the Hind Bank Marine Conservation District (MCD), four digital photo transects were conducted at depths ranging from 32 to 90 m. At 40 m depth in the MCD, they found well-developed coral reefs with up to 70% living coral cover. The SeaBED provided unprecedented information on a little known coral reef habitat that is common along the upper insular slopes of many Caribbean Islands.

Tools: Provision of Facilities, Software and Databases to Enhance Research Productivity
- The new CenSSIS Hyperspectral Toolbox developed at UPRM will allow users worldwide access to a valuable software library for environmental and bio medical image analysis. Researchers at the CenSSIS Hyperspectral Image Processing group released the first version of a MatLAB toolbox incorporating different algorithms for supervised and unsupervised classification of hyperspectral images. The tools available in the toolbox represent over 5 years of research sponsored by NSF and others. These tools have been demonstrated in the environmental remote sensing area where early versions of the toolbox were used by the Army Topographic Engineering Center in Fort Belvoir, Virginia for land cover classification. The tools are also being used by researchers in the biomedical area. New algorithms are continually being added to the toolbox as soon as they are tested and validated.

- CenSSIS resaechers at RPI have developed a patent pending Dual-Bootstrap Algorithm which can enable the creation of 3D maps of complex regions such as retinal vascular and coral reef structures. Collaboration is ongoing at GE and WHOI to test the method on real data. They have constructed a prototype software toolkit based on the concepts underlying this and related algorithms, and are using this toolkit in a CenSSIS course on image registration. In collaboration with MGH and GE they have begun implementing and testing a new deformable lung CT image registration algorithm using this toolkit.


- An RPI-NU-Lockheed Martin team has demonstrated that Terahertz (THz) probes can be used to quantitatively locate defects in slabs of foam insulation similar to those that were thought to cause the Columbia Space shuttle disaster.

Figure 7A-10. RPI students, Hua Zhong (right) and Xie Xu in THz Imaging Lab testing one of NASA's foams. [This photo was featured in Nature, August 14 issue, 2003.]

Our team uses THz waves to identify material defects in foam manufactured and used for space shuttle insulation. Defects, such as those believed responsible for the Columbia shuttle crash, have been difficult to spot using other methods.

- Researchers at NU and MGH have collaborated with Silicon Graphics Inc. to enable near real time imaging for an advanced mammogram screening technique. As a result of CenSSIS involvement, the reconstruction fell from three hours to under two minutes. Specifically, the CenSSIS R3 team was presented with the challenge of accelerating a new tomosynthesis breast cancer mammography reconstruction algorithm. We have demonstrated a 180X speedup on a 64-processor SGI Altix system, reducing the runtime from 3 hours on a Pentium 4 workstation to 1 minute. This work has also been a catalyst in attracting the attention of SGI to become a partner of CenSSIS. The parallelized code is now being used during clinical trials of the tomosynthesis device at MGH.


People: Development of a Well Prepared, Diverse Science and Engineering Workforce
- CenSSIS has intensified its Research Experience for Teachers (RET) program to provide 14 urban teachers collaborative links to NU, BU, MGH, WHOI, and RPI reseachers.
Professional development coupled with research and design experiences provides middle and high school teachers an opportunity to develop interesting, cutting-edge science and engineering classroom connections.
How might you engage a group of middle school students in a study of the problems related to hazardous waste identification and clean-up? One RET teacher, Mark Casto, after spending his summer using Cross Well Radar for Detection and Imaging, challenged his class with an outdoor exploration that hit the mark. (Mark had teams of students work outside to find a buried paint can - all came close but one group hit the mark.) Teachers like Mark maximize the RET experience by empowering their students to investigate important societal problems and encouraging them to consider STEM careers. (See Figure 7A-11.)
Figure 7A-11. Students work with graphical data and transfer it to a grid that covers a parcel of land. Once students have identified where the spill is on paper, they will take the grid outdoors where a parcel of land, with the same dimensions, has been marked out with stakes. They are to then locate where the spill area and leaking drum are located.

- The Four CenSSIS academic partners have initiated a program to engage incoming freshman in the activities of the Center.

For the past two years approximately 80 "CenSSIS Scholars" have been appointed. This diverse group of students has become actively engaged in CenSSIS efforts such as K-12 outreach, CenSSIS seminars and the High Tech Tools and Toys Laboratory experiences. The CenSSIS Scholars receive book vouchers in recognition of their participation. At NU, the upperclass CenSSIS Scholars began working with Boston-area teachers in the fall of 2003 and we are expanding this effort to include freshman CenSSIS Scholars in the spring of 2004. The CenSSIS Scholar volunteers are working with Boston teachers who were past participants in the CenSSIS Research Experiences for Teachers Program.
At BU, several CenSSIS Scholars have worked on projects in the High Tech Tools and Toys Lab, preparing imaging stations for our freshman intro to engineering lab. These include Adam Lavely, Howard Simpson, Mary Diana, and Eleonora Vidolova. Katherine Harnett is working in a faculty member's lab as an undergraduate researcher.
At RPI, two CenSSIS Scholars are working with Prof. Birsen Yazici on developing modules for the RPI High Tech Tools and Toys Class and CenSSIS Scholar, Rebecca Fellin is working on a CenSSIS research project with Prof. Jonathan Newell to develop a way of using a thin layer of purified agar, possibly partially dehydrated, to serve as a model for the electrical properties of the skin covering the breast.
In general, CenSSIS Scholars have sought stronger involvement with CenSSIS through pursuing other opportunities that become available. BU CenSSIS Scholar, Lolita Tam and NU CenSSIS Scholar, Lindsey Tessier applied and were accepted to the CenSSIS 2003 REU Program. Both Lolita and Lindsey made significant contributions to the research projects that they worked on with CenSSIS faculty. NU CenSSIS Scholars, Amanda Angell, Matthew Bouchard and William Price have all begun working with CenSSIS faculty on research work during the 2003/2004 school year. UPRM CenSSIS Scholar, Stephanie Mendoza and NU CenSSIS Scholar, Anthony Serino are members of the Executive Committee of the CenSSIS Student Leadership Council and as such, organize CenSSIS student meetings and conduct the annual SWOT Analysis.


- Five CenSSIS engineering students participated in a K-12 outreach project by visiting the Upham Elementary School in Wellesley, MA to help first through fifth graders brainstorm for their "Invention Convention". This biennial event challenges budding engineers and scientists (and their classmates) to conceive and reduce to practice new devices to solve everyday problems. Some examples include: a water bottle with a hidden handle which transforms into a squirt gun; a portable van de Graff generator to create interesting hairstyles; and a dog shaver to help with canine grooming. The graduate and undergraduate Northeastern University electrical and civil engineering students were organized by Prof. Carey Rappaport to meet with students, jog their creative minds, discuss important design considerations, and think logically. Both college students and elementary students enjoyed the event, and learned from each other. There will be additional visits to Upham School for follow-up and invention presentations