Project Description:

Nuclei segmentation is considered as the fundamental step in quantitative microscopic image analysis. As manual segmentation, performed by experts, is a time-consuming, subjective and complex procedure, the need for automatic computerized methods to perform nuclei segmentation has emerged. While Deep Learning-based approaches are the state-of-the-art solutions for image segmentation tasks, there are still two main unsolved issues.

The first issue is the instance segmentation performance for challenging object cases, which needs to be significantly improved. The second issue is the generalization capability of these models. In this project, senior and young researchers aim to address both issues. In particular, for the first issue, our technical goal is to propose a novel DL-based approach for improved nuclei segmentation. For the second issue, we plan to improve the generalization capacity of the models by three different Deep Learning-based knowledge transfer methods including deep model adaption, deep style transfer for colour normalization and multi-modal staining fusion for artificial ground truth generation

Our project on FFG:

projekte.ffg.at/projekt/3258628