Rutgers’ graduate student Sushil Mittal has collaborated with a group of researchers from Siemens Corporate Research and Siemens Healthcare on a biomedical research project. Under the tutelage of Professor Peter Meer, Mr. Mittal spearheaded the project while interning at Siemens Corporate Research in Princeton, NJ. Mr. Mittal is the first author on the first published paper, “Fast Automatic Detection of Calcified Coronary Lesions in 3D Cardiac CT Images*,” to feature research from the project.
The project aimed to address some of the difficulties related to detecting calcified coronary lesions. Despite recent advances in multi-detector computed tomography (MDCT), detection of coronary lesions remains a troubling task. There are multiple reasons why detection is so difficult. For one, the lesions are incredibly small. Secondly, coronary arteries are by nature long and winding, which makes identifying lesions difficult. Finally, the imaging data tends to be of a low resolution, and vulnerable to noise, blooming, and motion artifacts. In response to these issues, Mr. Mittal and the rest of the team have proposed a solution: a novel learning-based, fully automatic algorithm for the detection of calcified lesions in contrast enhanced 3-D CT data.
With guidance from his manager at Siemens and Professor Meer, Mr. Mittal was solely responsible for the implementation and execution of the project experiments. The experiments compared and evaluated the performance of two supervised learning methods—Probabilistic Boosting Tree (PBT) and Random Forest (RF) classifier. Both PBT and RF use rotation invariant features that are extracted along the centerline of the coronaries. On data collected from 165 patients, Mr. Mittal was able to achieve an approximate 90% detection rate for less than one false positive scan. Below are two detection results found on coronary images:
Original PBT RF
Siemens Corporate Research funded the project and Mr. Mittal is currently waiting on the review decision of the project’s second paper. He works with Professor Meer in the Robust Image Understanding Laboratory at Rutgers.
S. Mittal, Yefeng Zheng, B. Georgescu, F. Vega-Higuera, Shaohua Kevin Zhou, P. Meer, D. Comaniciu. Published in Machine Learning in Medical Imaging, Volume 6357 of Lecture Notes in Computer Science, and available online coewww.rutgers.edu/riul/research/papers/pdf/wkmiccai10.pdf
By Sean Patrick Cooper