MDL Usage

File Formats

filename.raw: 3D volume dataset with dimension indicated by the filename (x*y*z), with no header.
filename.seed: text file with high curvature seed points, which contains a list of 3D points.
filename.skel: text file with skeleton points, which contains a list of 3D points.
filename.swc: file format to describe dendritic morphology. The branches of a dendritic tree are represented as a set of cylinders.
filename.vtk: common vtk format, which contains MDL skeleton output.

Code Components

Example of program results of skeleton points, backbone extraction and dendritic tree. (A) Original 3D image on 2D z-projection, (B) 3D Skeleton points after path-line formation method, (C) Extracted dendritic backbone, (D) Resultant dendritic tree of MDL backbone and spine models.

• AnisoDiffuse: Image smoothing algorithm – Anisotropic Diffusion

It is an iterative method that can produce smooth results while keeping useful edge structures.
Input and output: 3D image volume

• ConnComntwFldfill: Connected component analysis algorithm with flood filling method (to prevent from memory outflow problem)

Compute Connected Component with Flood filling method DepthFirstSearch may cause stack overflow for large datasets.
Input and output: 3D image volume

Usage: ConnComntwFldfill <input file> <sizeX> <sizeY> <sizeZ> <output file> <threshold>

• ConnectComponents: Connected component analysis algorithm with scanline filling method

Label the connected components of the input volume with zero background and remove the connected components with small number of voxels.
Input: original volume
Output: removed small objects

• Floodfill: Image floodfill algorithm

Flood filling accept a sequence of volumes. Input data contains either 0 or values greater or equal to 3.
Input and output: 3D image volume

• GradientVecField: Computing gradient vector field from intensity images

Compute the gradient vector field of any 3D intensity map.
Input : 3D volume density map with any sizes
Output: ASCII file with vector 3 components for all object voxels

• Skel_Extrvalley3D: Computing high curvature seed points

Extract the ridges and valleys feature in 3D vector field
Input: 3D vector field
Output: 3D image-scalar field

• Skel_streamline: Producing 3D skeleton points from vector field and seed points

Form path-lines from saddle points and seed points
Input: 1. 3D vector field
2. Seed points
Output: skelelton file

• MinSpanTree: MDL algorithm for graph (tree) generation

Generate MDL graph structure from a list of 3D points.
Input format: .skel (3D points)
Output format: .swc or .vtk (3D graph format)

• VolumeProcess: Image preprocessing algorithms (including image operations, crop, scale, etc.)

Volume dataset processing accepts a sequence of volumes
Input parameters
1. sizeExpand
2. preproess

Python Wrapping

This python code gives an example of program run on dataset MBFsp5 with dimension x-308, y-512 and z-49.

#!/usr/bin/env python
import os
data_dir = "/projects/MDLCode/zackdata"
os.system("volumeProcess/volumeProcess  %s/MBFsp5.308x512x49.raw  308 512 49 1 %s/volume_Processed.raw 2" % (data_dir, data_dir))
os.system("ConnCompntwFldfill/ConnCompntwFldfill %s/volume_Processed.raw 308 512 49 %s/components_Connected.raw 100" % (data_dir, data_dir))
os.system("AnisoDiffuse/AnisoDiffuse  %s/components_Connected.raw 308 512 49 %s/Aniso_Diffused.raw 0" % (data_dir, data_dir))
os.system("GradientVecField/GradientVecField %s/Aniso_Diffused.raw 308 512 49  %s/out.vec" % (data_dir, data_dir))
os.system("skel_Extrvalley3D/skel_Extrvalley3D %s/out.vec 308 512 49 %s/out.seed" % (data_dir, data_dir))
os.system("skel_streamline/skel_streamline %s/out.vec 308 512 49 %s/out.seed %s/out.skel" % (data_dir, data_dir, data_dir))
os.system("MinSpanTree/MinSpanTree %s/ out.skel components_Connected.raw 308 512 49 %s/out.vtk out.txt %s/Aniso_Diffused.raw" % (data_dir, data_dir, data_dir))