News Highlight

Title: Finding Many Small Structures in Microscopic Images

Speaker: Elena Bernardis, Department of Radiology, University of Pennsylvania, (https://www.rad.upenn.edu/sbia/Elena.Bernardis/)

Date: April 23th, 2012

Time: 2:00pm--3:30pm

Location: Babbio 321

Abstract:

Extracting and counting numerous `dots', i.e. small round regions, in
large microscopic images is encountered in a wide range of medical and
scientific applications. In this talk, I will present a dot finding
method for popping out these regions of interest based on spectral graph
partitioning, where pixels are nodes of a weighted graph, and finding
dots becomes dissecting the graph based on weighted connections between
nodes. In contrast to most previous approaches that rely on sharp
boundaries, our features encode local convexity information instead. By
distancing ourselves from all traditional image segmentation methods
that emphasize precision of boundary locations and shapes, we obtain a
solution that is paradoxically closer to the desired segmentation.
Specifically, our method can deal with faint image conditions and
ambiguous boundaries, very common in medical image applications. I will
present some results to show various possible applications to medical
image segmentation, materials research and texture analysis, as well as
current work extending to shapes beyond `dots’.

Biography:

Elena Bernardis is a postdoctoral researcher in the Department of
Radiology at the University of Pennsylvania. She recently received her
Ph.D. from the Department of Computer and Information Science at the
University of Pennsylvania, working as member of the GRASP Laboratory
with Prof. Jianbo Shi and Prof. Stella Yu. Before joining Prof. Shi's
computer vision group, she completed a Master of Mathematics at the
University of Pennsylvania and a dual-major Bachelor's degree in
Materials Science and Engineering and Mathematics at the University of
Minnesota.