BMIPL presented two posters at the 2016 IEEE International Symposium on Biomedical Imaging (ISBI).
One was “Center Pixel Weight Estimation for Non-Local Means Filtering Using Local James-Stein Estimator with Bounded Self-Weights” By Minh Phuong Nguyen (MS student, UNIST) and Se Young Chun. This work (called BLJS) addressed the issue of selecting non-local means filter parameters using James-Stein estimator and yielded better quality images than previous state-of-the-art methods (e.g. LJS).
The other was “Joint Estimation of Activity Distribution and Attenuation Map for Tof-Pet Using Alternating Direction Method of Multiplier” by Se Young Chun, Kyeong Yun Kim (SNU), Jae Sung Lee (SNU), and Jeff Fessler (University of Michigan). This collaboration investigated a novel algorithm to estimate activity and attenuation from TOF PET data jointly faster than previous methods.
Prof Se Young Chun also collaborated with Prof Jaesik Choi and members at SAIL (Statistical Artificial Intelligence Lab, UNIST) for the IEEE ISBI 2016 Challenge “Skin Lesion Analysis towards Melanoma Detection”. Thanh, Haebeom and Janghoon (SAIL, UNIST) ranked the 1st and 3rd places at Part 2B: Lesion Dermoscopic Feature Segmentation and Part 3B: Segmented Lesion Classification, respectively. Deep neural networks were used to tackle these challenges. For more information on challenge results, click here.
For more information on ISBI, click here.