The focus of the Medical Informatics (MedIX) Laboratory is purposeful and pragmatic: “We process and model data to extract information for use in the medical field,” says Daniela Stan Raicu (above), a professor in the College of Computing and Digital Media (CDM) and lab co-director.

But the work itself seems almost magical.

In one project, researchers—faculty and students, both graduate and undergraduate—have built algorithms that read CT scans of lungs to predict potential malignancy of nodules. “The program can evaluate hundreds of thousands of pixels, from hundreds of images per person, while drawing on a rich database of comparable images, all with the intent of reducing uncertainty for radiologists,” says Raicu.

In another project, they’re documenting the behavior of Caenorhabditis elegans (C. elegans), a microscopic worm used for investigating neural activity. “The C. elegans has 302 neurons, of a total 959 cells, and all the connections have been mapped,” says Raicu. “Using a tracker device and computational programs that we built in the lab, we’re gathering and analyzing data describing the movements of 86 single worms—some normal, some mutated—filmed at 30 frames per second, for up to three hours at a time, with the hope of learning more about human brain function.”

“Here, in the lab, it’s easy to see how the power of computing could impact everyone’s life,” she adds. Undergraduate Miguel Carrazza (above, right) shares her enthusiasm for the convergence of data science and medicine: “How many people could be better off because of what we’re doing here? Every research project that gets worked on, every paper that gets published, might help save lives.”

Computer-Aided Diagnosis

The lab started 14 years ago when Raicu and Jacob Furst, a professor in CDM and lab co-director, joined forces to explore how machine learning and data mining could improve the confidence of radiologists as they read CT scans of lung nodules and recommend a next step in treatment. After the National Cancer Institute made available a collection of CT images of the chest for about 1,000 patients, they taught the computer how to evaluate nodules across seven characteristics important in the diagnostic process, including the level of spiculation of the nodule and its subtlety.

“Uncertainty is a problem in the diagnostic process, especially when multiple radiologists don’t agree about what they’re seeing,” says Furst. “In general, the more data you have, the better your predictions. So, we wanted to create a large, stable data set that a radiologist could use to confirm or refute a conclusion: Is the nodule potentially malignant or not? Click on the image, and get the computer’s opinion—an opinion that’s based on thousands of samples that have already been classified.”

In addition to improving diagnostic accuracy, the program will boost physician productivity, as Carrazza points out: “A typical radiologist looks at 14,000 images a year. If we can reduce the burden of that work flow somehow, if we can make the radiologist more efficient and effective, then we’re helping medicine take a huge step towards better patient care.”

Lessons of the Lowly Worm

Working with Rosalind Franklin University of Medicine and Science (RFUMS), and supported by a collaborative DePaul-RFUMS grant, researchers in the lab are also documenting the role of serotonin—a neurotransmitter that influences mood, appetite, sleep, learning and other brain functions in humans—in the behavior of C. elegans.

“The worms are an excellent experimental subject,” says Furst. “They’re simple enough that they can be mutated in very precise and understandable ways, and yet their neurotransmission process is fundamentally the same as ours. Does this mean that the work we’re doing in the lab could eventually help physicians treating people who have trouble absorbing serotonin? It’s a huge jump, of course, but that’s the promise.”

In the lab the researchers compare the paths of normal and serotonin-inhibited worms as both types move around plates, looking for food. “Our initial findings suggest that the mutated worms spend more time searching for food, and the normal worms spend more time eating food,” says Furst. “This suggests that serotonin-inhibited worms can’t be ‘satisfied’ readily.”

Studying the worms means mathematically describing their paths, based on thousands of video images, and then analyzing the huge amount of data that results. “Any differences in the worms’ paths can’t be seen; they can only be described mathematically,” says graduate student, Ian Wang (above, center). “And that math is very complex. We’re using sophisticated statistics, data clustering, and discrete time warping; we’ve built a very complex computational model.”

A Rare Opportunity

“For me, personally, this research is very important, whether I pursue a professional career in industry or enter a PhD program,” Wang adds.

Carrazza agrees on the value of doing research as a student.

“In class, no matter what the subject, students are never asked a question that doesn’t have an answer. You look in the back of the book, you search on Google—it’s there. But in research you get to work on problems that don’t have answers, and that exercises your brain in a completely different way, a way that I find to be more truly educational, more enriching and more satisfying. It’s so rare for students, especially undergraduates, to get to work on original, worthwhile research.”

Also, each summer since 2005, the MedIX lab has hosted eight undergraduates from around the country as part of the NSF "Research Experience for Undergraduates" (REU) program, an interdisciplinary collaboration between DePaul and the University of Chicago. Only about 75 universities offer REU programs in computer science, and DePaul’s program is one of only a few focused specifically on medical informatics.