My master's thesis focuses on the automated analysis of microscopic images of liverworts belonging to the Scapaniaceae family.
The main objective is to automatically extract morphological features of individual leaf cells and investigate how these features can help distinguish between different liverwort species.
Since manually analyzing microscopic images is time-consuming, image processing and machine learning techniques will be used to simplify and improve this process. First, the microscopic images will be prepared, and then different methods for segmenting individual cells will be tested, including Cellpose and potentially other suitable approaches.
The segmentation quality will be evaluated using manually annotated reference images. After segmentation, morphological features such as cell area, perimeter, length, width, and roundness will be extracted.
These features will then be analyzed to identify potential differences between the studied species and used to develop machine learning models for species classification. An important part of the thesis is also the development of a decision tree to determine which cell features are most relevant for distinguishing between different species.
The de.NBI Cloud will be used to store high-resolution microscopic images and perform the necessary image analyses and computational tasks.