| |||||||||||||||||||||
Reeves and Zabih are each collaborating with radiologists at Weill Cornell Medical College on algorithms to automate such tasks as maximizing the quality of a radiogram, or even help diagnose lung cancer, aneurysms, and breast cancer.
“The best predictor of malignancy is rate of growth,” says Reeves, who hopes that ultimately his team’s algorithm might supplant the need for biopsies as diagnostic tools. Zabih works on another problem: MRIs, despite their spectacular resolution, are highly prone to motion artifacts, caused both by quirks in the imaging technology and the natural movements of the human body. His team developed an algorithm that automatically corrects for motion artifacts in MR angiography, a non-invasive technique for visualizing blood vessels, and provides radiologists with significantly superior images.
Professor Uri Keich in Computer Science works on this motif-finding problem, with its twin challenges: To find the most pronounced motifs in input sequences and to analyze their statistical significance to determine if they are merely artifacts of the size of the data. “I work on both aspects of the problem but mostly on the computational statistics problem,” Keich says. “The special difficulty is that due to the huge number of competing motifs we need to estimate accurately very small measurements of significance; for the same reason, it needs to be done extremely quickly. “This work led us to an innovative significance evaluation of other statistical tests that is especially important for analyzing large datasets.”
Traditional radiation treatment delivers a heavy dose to an area defined as the tumor spot, which is considerably larger than the actual tumor. But, Henderson asks, how likely is it that the tumor will stay in a fixed position, especially when the patient returns for treatment once a day over a period of four to six weeks and the tumor is shrinking? Henderson’s strategy, along with a senior physicist at Princess Margaret Hospital in Toronto, relies on CT scans, algorithms, and heavy computation—robust optimization—to determine the tumor’s location and deliver appropriate doses while sparing surrounding tissue. “The traditional way is to deliver a lot of radiation inside the spot and none outside; it’s an all-or-nothing plan. Instead, we want to see the highest concentration of radiation at the very center and a gradual diminishing as you move outward. Furthermore, there are many ways to deliver radiation—by adjusting the angles or the intensities of each of the beams. We want to provide a smoother dose no matter how much things vary inside.”
In Chemical and Biomolecular Engineering, Professor Fernando Escobedo does stochastic modeling of the mutation and selection processes in the germinal centers of the lymph nodes that produce B cells finely tuned to fight foreign invaders. Recent experimental findings suggest that a few key mutations are associated with the onset and proliferation of the cancer. Disease depends on the sequence of mutations, and that sequence is essentially random. That’s what Escobedo studies: how the population of cells within a premalignant lymph node of a person changes over time. This work has the potential to point to new therapies, such as targeting some of the players in the germinal centers to slow down the progress of the disease. “Cancer uses stochastic processes and natural selection to its advantage; we need to learn how to beat cancer in the stochastic game,” Escobedo says. |
The invention of Computed Tomography (CT) scans and Magnetic Resonance (MR) images considerably improved biomedical imaging, not least because these scans can code diagnostic images into numbers. Now those inventions may be getting a big boost from Professor 
“The technology is producing so many more images in so much more detail that having a human look at them is simply impractical,” says Reeves. His early research focused on automating the detection, measurement, and diagnosis of pre-cancerous nodules. Now, most CT manufacturers have products that follow his strategy for measuring a lesion and its blood vessels. 
Protein production can be regulated by transcription factors that bind to specific DNA sites, thus regulating the transcription rate of proximal genes. Finding these sites, known as Transcription Factor Binding Sites, is fundamental to understanding gene expression regulation.
Professor 
Low-grade lymphoma, a cancer of the B cells in the blood, affected an estimated 20,000 new cases in this country in 2005. The disease usually evolves slowly over many years but exhibits some sporadic spurts. While current treatments can induce complete remission of the disease, they do not cure it: relapses always ensue.