The goal of the thesis is a visualization of deep neural network (DNN) results on images of pathological data. The thesis is part of the research group that uses DNNs to identify cancerous regions in the high-resolution scans of optical microscopy slides of tissues. The work is done in collaboration with Masaryk Memorial Cancer Institute.

The student will collect and analyze needs of the project in terms of what types and sizes of scans and what types of images produced from the DNN analyses are used and what is expected interaction with the visualizations. Based on this the student will perform research of the existing options for the visualization and will analyse which of them is best as a starting point for implementation of the system. Based on the results of the analysis, the target system will be implemented as an open-source system, ideally by extending one of the available systems for visualization of large-scale imagery. The minimum required functionality is ability to display slides as a base layer, the annotations provided by the pathologist as an overlay layer, visualization of probabilities resulting from the DNNs as another layer. The system should be further extensible at least in terms of number of overlay layers. The system should be tested with the results of the research group led by doc. Brázdil and doc. Holub and should be well documented for further development.