Detecting Irregular Curvilinear Structures in Gray Scale and Color Imagery using Multi-Directional Oriented Flux

We present here additional performance curves and comparative results on the evaluated datasets. The images shown in the links below are the maximum (or minimum) intensity projections (MIP) of the original image stacks and those of the algorithm results.

The results presented in the "Best Segmentations" columns are obtained by first finding the threshold value that maximizes the Jaccard index (equivalently the VOC measure defined in Eq.(4) of [26]). For this, we used η=0.4, as also mentioned in the manuscript. We then followed the procedure described in the last part of Section 4.2 to compute the segmentations.

Finally, for stacks with floating-point voxel precision, such as the tubularity volumes, the conversion to 8-bits (after the MIP step) is performed using the Fiji (http://fiji.sc) software [27]. The 3D renderings under the Videos links below are created using the publicly available Vaa3D visualization and analysis software [28].