Our Research
Our group’s work is focused on the automatic segmentation and labeling of intervertebral discs from clinical MRI images. We are working to perfect a probabilistic model for this purpose that will subsequently result in computer-aided diagnosis from these same images. Once complete, this work will help to reduce the reporting time for MRI results, as physicians will be presented with an automated quantitative analysis of the MRI, rather than images alone.
The computational tool we have developed uses both pixel (appearance) and object (relative location) features to localize and label each of the lumbar discs. The use of these two levels of analysis results in more accurate labeling of the discs—in our experimental results we achieved an accuracy rate of approximately 96%. In the future we plan to add domain knowledge and further levels of analysis, such as texture, disc orientation, and shape to boost accuracy levels even higher.
Currently, our tool is designed simply for the purpose of localizing and labeling the discs. It was also able, however, to accurately label lumbar discs even in the presence of significant disc abnormalities. Our next endeavor, then, is to extend the tool’s capabilities to more complicated cases of abnormalities and pathologies in order to have a maximum impact on computer-aided diagnosis of intervertebral disc degeneration (IDD).
In addition to the expediency that disc labeling and computer-aided diagnosis provides to the diagnosis of IDD cases, teleradiology can provide a further level of rapidity to this process. This important technology, which Dr. Vipin Chaudhary was instrumental in developing, allows for remote viewing of diagnostic images. Due to the vast numbers of IDD patients—approximately 12 million Americans have some form of IDD—and the current shortage of radiologists, teleradiology can provide a way for radiologists to quickly receive these images, regardless of whether they are immediately present. Consequently, the convergence of these projects will greatly expedite the diagnosis and subsequent access to treatment for millions of patients suffering from cases of IDD.
Participating Researchers & Departments
- Raja’ S. Alomari
- Vipin Chaudhary, Department of Computer Science and Engineering
- Jason Corso, Department of Computer Science and Engineering
- Gurmeet S. Dhillon, Radiologist, Proscan Imaging
- Jaehan Koh, Department of Computer Science and Engineering