Zen 3.4 blue

All images were acquired at pixel dimensions of 1024 × 1024 and are shown as the maximum intensity of the Z-projections using an LSM 900 laser scanning confocal microscope (Zeiss). For the measurement of immunofluorescence intensity, images were captured with the same laser power, and the mean intensity of the fluorescence signals was measured and displayed in arbitrary units (a.u.). When necessary, images were processed to reduce background noise prior to quantification. Unprocessed raw images were shown in supplementary figures (Supplementary Figure S9). The data were analyzed using ZEN 3.4 Blue (Zeiss) and ImageJ software (National Institutes of Health) under the same processing parameters.

To quantify NET, images were acquired with an ApoTome II microscope using ZEN 3.4 (blue edition) software from Carl Zeiss(Jena, Thüringen c). We used machine learning for the detection and quantification of NET. A classification model was created with the deep learning module Intellesis Trainable Segmentation Carl Zeiss(Jena, Thüringen). Four images were randomly selected for detection training, the training module was set to multispectral mode, and three classes were created for classification: background, nuclei, and NET. To normalize the NET area measured in each field, we chose only fields with 80 nuclei stained with DAPI. Nuclear detection was automatically performed with the ZEN cell nuclear counting module. We set up 50-feature deep learning and conditional random field postprocessing for image segmentation. Starting from the final detection model, we set up the image analysis module, and three randomly selected images from three independent trials were analyzed. Finally, the NET release area was obtained.

All images were acquired at pixel dimensions of 1024 × 1024 and are shown as the maximum intensity of the Z‐projections using an LSM 900 laser scanning confocal microscope (Zeiss). For the measurement of immunofluorescence intensity, images were captured with the same laser power and the mean intensity of the fluorescence signals was measured and normalized to the mean DAPI signal intensity. The data were analysed using ZEN 3.4 Blue (Zeiss) and ImageJ software (National Institutes of Health) under the same processing parameters.