IC Colloquium : Biomedical Image Analysis: Models and Methods for Reverse-engineering the Worm

Event details
Date | 19.03.2015 |
Hour | 10:15 › 11:30 |
Location | |
Category | Conferences - Seminars |
By : Dagmar Kainmueller - MPI-CBG
IC Faculty candidate
Abstract :
Biomedical image analysis tasks like segmentation and tracking are commonly approached by (1) modeling the task by means of an optimization problem, and (2) solving the optimization problem by an appropriate inference method. The respective objective function typically combines prior knowledge about the sought object, like e.g. its shape, appearance and dynamics, with cues obtained from the observed image data.
In my talk I will explore how capturing prior knowledge is vital for successful automation of challenging biomedical image analysis tasks. I will showcase a range of such tasks I have tackled. I will focus in detail on the task of segmenting and annotating (i.e. labeling) cell nuclei in the nematode worm C. Elegans in 3d microscopic images.
Automated segmentation and annotation of nuclei in C. Elegans is essential for the biological goal of "reverse engineering" how the DNA of the worm encodes its development. The model I will present for this task integrates the popular active shape models into a sparse graph matching objective, hence combining the benefits of global and local prior statistical shape models. Together with a novel inference method this model allows for fully automatic simultaneous segmentation and annotation of nuclei with unprecedented accuracy.
Bio :
Dagmar Kainmueller is a computer scientist working on biomedical image analysis. She is currently an ELBE postdoctoral researcher in Gene Myers' lab at the Max Planck Institute of Molecular Cell Biology and Genetics, where she investigates prior models, machine learning and inference techniques for accurate automatic analysis of microscopic images.
Dagmar studied computer science at the University of Karlsruhe, and obtained her PhD in medical image analysis from the University of Luebeck and Zuse-Institute Berlin in 2013. For her thesis she received the BVM award from the German society for image processing in medicine. With her methods she won a series of MICCAI Grand Challenges, among which the liver segmentation challenge in 2007. Her respective method to date still ranks first on benchmark data.
More information
IC Faculty candidate
Abstract :
Biomedical image analysis tasks like segmentation and tracking are commonly approached by (1) modeling the task by means of an optimization problem, and (2) solving the optimization problem by an appropriate inference method. The respective objective function typically combines prior knowledge about the sought object, like e.g. its shape, appearance and dynamics, with cues obtained from the observed image data.
In my talk I will explore how capturing prior knowledge is vital for successful automation of challenging biomedical image analysis tasks. I will showcase a range of such tasks I have tackled. I will focus in detail on the task of segmenting and annotating (i.e. labeling) cell nuclei in the nematode worm C. Elegans in 3d microscopic images.
Automated segmentation and annotation of nuclei in C. Elegans is essential for the biological goal of "reverse engineering" how the DNA of the worm encodes its development. The model I will present for this task integrates the popular active shape models into a sparse graph matching objective, hence combining the benefits of global and local prior statistical shape models. Together with a novel inference method this model allows for fully automatic simultaneous segmentation and annotation of nuclei with unprecedented accuracy.
Bio :
Dagmar Kainmueller is a computer scientist working on biomedical image analysis. She is currently an ELBE postdoctoral researcher in Gene Myers' lab at the Max Planck Institute of Molecular Cell Biology and Genetics, where she investigates prior models, machine learning and inference techniques for accurate automatic analysis of microscopic images.
Dagmar studied computer science at the University of Karlsruhe, and obtained her PhD in medical image analysis from the University of Luebeck and Zuse-Institute Berlin in 2013. For her thesis she received the BVM award from the German society for image processing in medicine. With her methods she won a series of MICCAI Grand Challenges, among which the liver segmentation challenge in 2007. Her respective method to date still ranks first on benchmark data.
More information
Practical information
- General public
- Free
- This event is internal
Contact
- Host : Sabine Süsstrunk