Students win prizes for improving image processing techniques for liver cancer detection and much more
Students in EECS 556: Image Processing, explore methods to improve image processing in applications such as biomedical imaging and video and image compression
Students in EECS 556: Image Processing, explore methods to improve image processing in applications such as biomedical imaging and video and image compression. The techniques are fundamental to companies such as KLA-Tencor, which offered prizes to two teams of students. The winning teams are highlighted below.
The course is taught by Jeff Fessler, William L. Root Collegiate Professor of Electrical Engineering and Computer Science, a leading expert in medical imaging with current and past projects in X-ray CT, MRI, PET, SPECT, radiation therapy, and image registration.
Convolutional operator learning for imaging
Siying Li, Haowei Xiang, Alexander Zaitzeff, Xiyu Zhang
According to the team, Convolutional Analysis Operator Learning (CAOL) is a method to learn kernels/features from large datasets, which have many applications in signal/image processing, computer vision and machine learning. Designing computationally efficient and fast-convergence algorithms is a subject of great interest.
This project proposed an innovative method to reduce the problem dimension of CAOL and speed up its computation. The team’s method shows faster convergence than the original CAOL acceleration method using majorizers. When applying the learned kernels to sparse-view CT reconstruction, their proposed method gives better image quality than the original method.
Automatic segmentation of tumorous liver CT scans
Sang Choi, Caroline Crockett, Alexander Ritchie, Rebecca Shen
“Liver cancer is among the leading causes of cancer deaths,” explained Rebecca Shen. “Many methods of treatment planning and measuring liver cancer require segmentation of tumorous regions by trained professionals. However, this segmentation is manual, tedious, and prone to differences between medical practitioners. Our project explored the implementation of semi-automated segmentation algorithms using the Potts model and the watershed algorithm.”
Data courtesy MIDAS: “Livers and liver tumors with expert hand segmentations”