Conidia_Counter

Simple approach to count conidia(or other particles) distribution

This app is a simple calculator of Aspergillus fumigatus 3D conidia distribution in the lung airways. It can be used for calculation of fluorescent labeled particles in fluorescent labeled mice lung airways or other structures. 3D images can be acquired by confocal microscope. For this app lung airways have to be priorly segmented to binary pictures.

Input files: - directory with 3D stack of conidia (or other particles) TIFF files
directory with 3D stack of masked lung airways (or other objects) TIFF files
Output files: - .xls file with the calculation output table

Algorithm description:
1. Calculate optimal threshold for conidia channel segmentation. As different lung samples can have different image intensity, segmentation process can not only segment target particles, but also noise pixels. As a result, the number of segmented conidia pixel depends on the threshold used. The idea was to create automatic way to segment conidia images to achieve similar pixels density between different lung samples. For this, simple filter been implemented to find threshold, in which the amount of neighbor pixels for each pixel is less than designated amount. The threshold has been calculated for each image in 3D stack and then the mode of threshold distribution was used as an optimal threshold for image stack. However, unpublished data shows that the error, occurring due to choosing different thresholds is significantly smaller than error, occurring due to comparison of different lungs' samples. Considering this, an option to choose custom threshold has been added to the user interface.
2. Perform binary AND operation for segmented conidia and segmented lung images and then calculate the number of True pixels. Simultaneously, calculate pixels outside of the lung mask. The bar chart and output table contains the percentage of conidia inside and outside of the lung mask relative to overall amount of conidia in the lung sample. Conidia inside the lung mask are considered to belong to bronchi and outside - to alveoli, as alveoli have poor intensity signal due to small wall thickness and are not segmented with mask segmentation method used for this work.
The algorithm can be modified to calculate conidia distribution in different bronchi generations given appropriate masks. However in such a case volume normalization should be considered to account for different lung geometries.