Project

Airway Segmentation from Chest CT Images

Segmentation of human airways from chest CT images is an important and difficult task. A precise and accurate segmentation is a prerequisite for further analysis in diagnosis and treatment of airway related diseases, such as chronic obstructive pulmonary disease . However, the low quality of acquired images, the branching structures, as well as pathologies make the task challenging. The database utilized in this study is mainly from the open challenge EXACT09.

My research on airway segmentation include the following conventional methods:

Region growing-based algorithm
Frangi tubeness algorithm
Circle fitting detector algorithm
Gradient vector flow-based algorithm
Cavity enhancement filtering-based algorithm

As well as the following deep learning-based methods:

3-D Frangi-Net
3-D U-Net and its variants

Retinal Vessel Segmentation from Fundus Images

Retinal vessel segmentation is a crucial step in fundus image analysis. It provides information of the distribution, thickness and curvature of the retinal vessels, thus greatly assists early stage diagnosis of circulate system related diseases, such as diabetic retinopathy.

Manual annotation is tedious and time consuming, thus automatic segmentation algorithms are investigated on. My research cover the following topics:

Lacunae and Blood Vessel Segmentation from Mouse Bone XRM Images

This research topic is part of the project 4D+nanoSCOPE: Advancing osteoporosis medicine by observing bone microstructure and remodelling using a four-dimensional nanoscope.

The project is funded by European Research Council (ERC) with € 12,366,635 million for 72 months. It is jointly proposed by Prof. Dr. Georg Schett, Director of the Department of Medicine 3, Universitätsklinikum Erlangen, Prof. Dr. Andreas Maier from the Department of Computer Science 5 at FAU, and Prof. Dr. Silke Christiansen from Fraunhofer Institute for Ceramic Technologies and Systems IKT.