One of the key steps in CT image analysis is segmentation, which enables specific structures or regions of interest to be isolated from the entire scanned volume for subsequent visualization and quantitative analysis. The choice of segmentation algorithm depends on the nature and complexity of the task. While simple thresholding is sufficient in some cases, more complex datasets require advanced methods such as watershed segmentation, deep learning-based segmentation, or manual segmentation.
Artificial intelligence enables us to leverage extensive datasets collected over many years of CT measurements and develop new, fully automated image segmentation methods. In deep learning-based segmentation, we use CT images together with manually created segmentation masks as training data. Through an iterative learning process, a model is developed that can subsequently automatically segment previously unanalyzed CT data.
