Fig 1.
Characterization of the conidial morphology of five different fungal species using FFT.
Results generated by the conidia/spore morphology function of FFT for images representing conidia of five different species (Top). Area (A), length (B), width (C) and circularity (D) of conidia for different fungal species, as assessed from 10 images per fungal species by FFT, where * represents significance levels of unpaired t-test.
Fig 2.
Results of the conidia/spore counting function of FFT for five different fungal species.
Example of the conidia detected by FFT (blue) overlaid on the original images for each of the studied fungal species. Bar chart representing the performance of FFT as percent error between the conidia detected by FFT versus the number of manually counted conidia for 10–15 images per species.
Fig 3.
Differences in conidiation for three strains of A. oligospora as evaluated by FFT.
Number of spores detected in a 5 μl droplet of spore solution extracted from cultures of strains TWF154, TWF132 and TWF102, respectively (A-C), together with a bar chart representing the results obtained from 10 droplets per strain (D). In the bar chart, * represents significance levels of unpaired t-test and error bars represent SD.
Table 1.
Accuracy measures of FFT analysis of 10 images of A. oligospora. Precision, Recall, F-measure and MCC computed for binary images assessed using FFT with respect to its ground truth counterpart for each image of our test set.
Fig 4.
Output of the mycelium characterization function of FFT for different A. oligospora strains.
Changes in the developing mycelium of A. oligospora (strain TWF154) over 72 hours, computed from the time-series of mathematical graphs generated by FFT (Top). A. Change over time of the area covered by mycelium, computed as the convex hull of all nodes in the mathematical graph for each time-point. B. Positions of the hyphal tips at each time-point. C. Overlay of the hyphal segments (edges) at different time-points showing development of the mathematical graph over time. Fungal features detected in images of the TWF154, TWF132 and TWF102 strains after 72 hours of growth (D, E, F, respectively). Area covered by the mycelium (blue polygon), mathematical graph representing the hyphal segments (green network), and hyphal tips (yellow dots) detected by FFT. The graphical representations show changes over time for the number of hyphal tips (G), total mycelial length (H), and the area covered by the mycelium (I), computed as the mean of six replicates per strain and time-point.
Fig 5.
Mycelium development of different fungal species and in different media, as captured by FFT.
Images showing the evolution of the mathematical graphs generated by FFT and representing 48 hours of growth on LNM (A-C) and PDA (D-F) media for each fungal species. Measures of mycelia over time, obtained as the mean of two replicates per species and time-point for LNM (G-I) and PDA (J-L) media.
Fig 6.
Performance of the trap counting function of FFT for three strains of A. oligospora.
Images show the traps detected by FFT, highlighted in blue in the original images (A-C). The bar chart (D) shows the difference in trap number between the different strains of A. oligospora, as computed using the average of three images per replicate for a total of five replicates (* represents significance levels of unpaired t-test and error bars represent SD).
Fig 7.
General workflow of FFT. The “Input options” tab (Top) allows users to select the set of images for analysis and the quantification function to be applied. In the “Calibration” tab (Middle), users can select and test the effect of different parameter combinations on one image from the set of images. FFT output information is defined in the “Output options” tab (Bottom). Once all this information has been provided, the “Run” button executes FFT analysis on all images in the set of images.
Fig 8.
Workflow of the mycelium characterization algorithm.
Original image (A). Binary image obtained by subtracting a gaussian-filtered version of the original image, applying morphological binarization, and applying a maximum area filter (B). Result of applying the Thinning function to the image in B (C). Mathematical graph obtained using the MorphologicalGraph function, in which nodes (blue) represent hyphal junctions/tips and edges (red) represent hyphae (D).