def find_ROI_nodes(graph,atlas,roi_label,use_wm=[False,False]):
""" Computes the nodes set for each ROI.
Parameters
----------
graph: RCSP Graph (mandatory)
Graph in which to perform the tractography.
atlas: numpy float array (mandatory)
3-D array holding the labels for the ROIs in voxel coordinates.
label: [int,int] (mandatory)
Labels of the two endpoint ROIs.
use_wm: [wm,wm] where wm is boolean
Whether to use WM voxels (nodes) of the ROIs. (Default: [False,False])
Returns
-----------
roi1_nodes: [node1,node2,...]
List of node IDs for first ROI.
roi2_nodes: [node1,node2,...]
List of node IDs for second ROI.
"""
n = graph.no_of_nodes() #number of nodes in the graph
#select endpoint ROI voxels
roi1_nodes = [] #source region
roi2_nodes = [] #target region
for node in range(0,n):
tissue = graph.node(node).properties['tissue'][0]
#coordinates
x = int(graph.node(node).properties['position'][0])
y = int(graph.node(node).properties['position'][1])
z = int(graph.node(node).properties['position'][2])
if tissue == float(tissue2int('GM')):
if atlas[x,y,z] == roi_label[0]:
roi1_nodes.append(node)
if atlas[x,y,z] == roi_label[1]:
roi2_nodes.append(node)
elif tissue == float(tissue2int('WM')):
if atlas[x,y,z] == roi_label[0] and use_wm[0]:
roi1_nodes.append(node)
if atlas[x,y,z] == roi_label[1] and use_wm[1]:
roi2_nodes.append(node)
#make sure to compute shortest paths for smaller region
if len(roi1_nodes) > len(roi2_nodes):
temp = roi1_nodes
roi1_nodes = roi2_nodes
roi2_nodes = temp
return roi1_nodes, roi2_nodes
def convert_graph(graph,prior=None,mask=None,atlas=None):
""" Convert the brain graph from python to C format, optionally including WM probabilities and ROI labels.\n\
Note: edge weights of the given graph are assumed to not contain any prior information!
Parameters
----------
graph: Graph in python or binary format. (mandatory)
prior: numpy array
Holds the prior probability information in a float 3D-array.
mask: numpy array
Holds binary inclusion/waypoint prior information in a float 3D-array.
atlas: numpy array
Holds a region label for each voxel in a float 3D-array.
Returns
-----------
new_graph: Graph in binary format.
"""
#map from current node IDs to [0, number of nodes) interval
id_to_idx = {}
#collect nodes
nodes = []
#current IDs of the nodes
if type(graph) is dict: nodelist = graph['nodes'].keys() #python format
elif type(graph) is Graph: nodelist = range(0,graph.no_of_nodes()) #binary format
for i, node_id in enumerate(nodelist):
id_to_idx[node_id] = i
#python format
if type(graph) is dict:
node = graph['nodes'][node_id]
#node coordinates
x = node['x']
y = node['y']
z = node['z']
tissue = tissue2int(node['tissue'])
cur_roi = node['region']
#binary format
elif type(graph) is Graph:
node = graph.node(node_id)
#node coordinates
x = int(node.properties['position'][0])
y = int(node.properties['position'][1])
z = int(node.properties['position'][2])
tissue = node.properties['tissue'][0]
cur_roi = int(node.properties['region'][0])
#ROI label information
if atlas is not None: roi = atlas[x,y,z] #overwrite current label with new label
else: roi = cur_roi #keep current label
#prior information
if prior is not None: #new prior
prob = prior[x,y,z]
#set probability for GM and WM voxels low but above 0 to avoid unconnected nodes
if prob == 0. and (tissue == 2 or tissue == 3): prob = 10**-10
else: prob = 1
#apply binary mask to prior
if mask is not None: prob *= mask[x,y,z]
#set the properties for current node
properties = {
'position': Property([ float(x), float(y), float(z) ]),
'tissue': Property(tissue),
'region': Property(float(roi)),
'prior': Property(float(prob))
}
nodes.append(Node(i, properties))
#collect edges
edges = []
for source_id in nodelist:
sid = id_to_idx[source_id] #new ID of node
edge_list = [] #edges of current node
#neighbor nodes
if type(graph) is dict: target_list = graph['W'][source_id].keys() #python format
elif type(graph) is Graph: target_list = [e.node for e in graph.edges(source_id)] #binary format
for target_id in target_list:
tid = id_to_idx[target_id] #new ID of neighbor node
#edge weight
if type(graph) is dict: ew = graph['W'][source_id][target_id] #python format
elif type(graph) is Graph: ew = graph.edge(source_id,target_id).weight #binary format
if ((prior is not None) or (mask is not None)): #edges need to be updated if prior changed or mask was applied
prob1 = nodes[sid].properties['prior'][0]
prob2 = nodes[tid].properties['prior'][0]
if (prob1>0 and prob2>0): #not removed by mask
#include prior of endnodes in the edge weight
if prior is not None: ew -= math.log(math.sqrt(prob1) * math.sqrt(prob2))
edge_list.append(Edge(tid, ew)) #keep old edge weight
else: edge_list.append(Edge(tid, ew)) #keep old edge weight
edges.append(edge_list) #add node's edges to list of edges
#create new graph
new_graph = Graph(nodes, edges)
return new_graph