Home
NucDetect is a Python package that is designed to identify the typical protein foci which which can be observed after DNA double strand breaks (such as γH2AX foci). It uses fluorescence images (in the rgb color space) to to identify and quantify the nucleus and foci on both the red and green channel. After analysis, the results can be either obtained as table or as image with marked foci. Images can be arranged into experiments, which can be divided into groups. Each experiment may then be statistically analysed (each group will be compared to a Poisson distribution and a boxplot will be created). If the automatic detection of foci is not accurat, the user can manually modify the results to add missing foci, merge oversegmented nuclei or remove falsely identified nuclei and/or foci.
NucDetect combines classical image processing with machine learning. For each analysis, the user can choose to either use the image processing based approach for detection or an pre-trained Fully Convolutional Network with U-Net architecture. The differentiation between both approaches is grounded in the experimental status of the ML approach. In the future, both approaches might be compared to improve detection accuracy.
The FCN will be trained with data obtained from manually modified automatic detection. This means a classical algorithmic approach is used for nucleus and focus detection, which will be then manually modified to correct errors such as oversegmented nuclei, missing foci et cetera. The so generated data will be then used to train the CNN to obtain higher detection accuracy than a classical image processing approach.
The main window will be shown on start of the program. It contains the main toolbar (upper row of buttons), the result table (shows the analysis results), the image list (a list of all images found in the folder User/NucDetect/images) and the image toolbar (row of buttons below the image list).
The Images Toolbar contains following buttons (from left to right):
- Analyse (analyse the selected image)
- Analyse All (analyse all images in the image list)
- Delete (removes the selected image from the image list)
- Delete ALl (removes all images from the image list)
Images placed in the folder User/NucDetect/images will be automatically loaded and shown in the image list. If no images are added to the folder User/NucDetect/images, the user needs to place images there and press the reload button (two arrows forming a circle). This can also be used if new images are placed inside the Images folder whilst the program is open.