Pre- and Post-processing Strategies for Generic Slice-wise Segmentation of Tomographic 3D Datasets Utilizing U-Net Deep Learning Models Trained for Specific Diagnostic Domains