added visualizations and some more bug fixes

Former-commit-id: 735b26c

imgtools/autopipeline.py

@@ -33,6 +33,7 @@ def __init__(self,
                  modalities="CT",
                  spacing=(1., 1., 0.),
                  n_jobs=-1,
+                 visualize=False,
                  missing_strategy="drop",
                  show_progress=False,
                  warn_on_error=False):
@@ -50,7 +51,7 @@ def __init__(self,
         self.existing = [None] #self.existing_patients()
 
         #input operations
-        self.input = ImageAutoInput(input_directory, modalities, n_jobs)
+        self.input = ImageAutoInput(input_directory, modalities, n_jobs, visualize)
 
         self.output_df_path = os.path.join(self.output_directory, "dataset.csv")
         #Output component table
@@ -220,7 +221,10 @@ def run(self):
 
     parser.add_argument("--modalities", type=str, default="CT",
                         help="List of desired modalities. Type as string for ex: RTSTRUCT,CT,RTDOSE")
-
+
+    parser.add_argument("--visualize", type=bool, default=False,
+                        help="Whether to visualize the data graph")
+
     parser.add_argument("--spacing", nargs=3, type=float, default=(1., 1., 0.),
                         help="The resampled voxel spacing in  (x, y, z) directions.")
 
@@ -236,6 +240,7 @@ def run(self):
                             modalities=args.modalities,
                             spacing=args.spacing,
                             n_jobs=args.n_jobs,
+                            visualize=args.visualize,
                             show_progress=args.show_progress)
 
     print(f'starting Pipeline...')

imgtools/modules/datagraph.py

@@ -1,10 +1,11 @@
 import os, time
 from typing import List
 from functools import reduce
-
 import numpy as np
 import pandas as pd
 from tqdm import tqdm
+from pyvis.network import Network
+
 
 class DataGraph:
     '''
@@ -24,7 +25,8 @@ class DataGraph:
     '''
     def __init__(self,
                  path_crawl: str,
-                 edge_path: str = "./patient_id_full_edges.csv") -> None:
+                 edge_path: str = "./patient_id_full_edges.csv",
+                 visualize: bool = False) -> None:
         '''
         Parameters
         ----------
@@ -43,6 +45,8 @@ def __init__(self,
         else:
             print("Edge table not present. Forming the edge table based on the crawl data...")
             self.form_graph()
+        if visualize:
+            self.visualize_graph()
 
     def form_graph(self):
         '''
@@ -82,6 +86,40 @@ def form_graph(self):
         print(f"Saving edge table in {self.edge_path}")
         self.df_edges.to_csv(self.edge_path, index=False)
 
+    def visualize_graph(self):
+        """
+        Generates visualization using Pyviz, a wrapper around visJS. The visualization can be found at datanet.html
+        """
+        print("Generating visualizations...")
+        data_net = Network(height='100%', width='100%', bgcolor='#222222', font_color='white')
+
+        sources = self.df_edges["series_y"]
+        targets = self.df_edges["series_x"]
+        name_src = self.df_edges["modality_y"]
+        name_tar = self.df_edges["modality_x"]
+        patient_id = self.df_edges["patient_ID_x"]
+        reference_ct = self.df_edges["reference_ct_y"]
+        reference_rs = self.df_edges["reference_rs_y"]
+
+        data_zip = zip(sources,targets,name_src,name_tar,patient_id,reference_ct,reference_rs)
+
+        for i in data_zip:
+            data_net.add_node(i[0],i[2],title=i[2],group=i[4])
+            data_net.add_node(i[1],i[3],title=i[3],group=i[4])
+            data_net.add_edge(i[0],i[1])
+            node = data_net.get_node(i[0])
+            node["title"] = "<br>Patient_id: {}<br>Series: {}<br>reference_ct: {}<br>reference_rs: {}".format(i[4],i[0],i[5],i[6])
+            node = data_net.get_node(i[1])
+            node["title"] = "<br>Patient_id: {}<br>Series: {}<br>reference_ct: {}<br>reference_rs: {}".format(i[4],i[1],i[5],i[6])
+
+        neigbour_map = data_net.get_adj_list()
+        for node in data_net.nodes:
+            node["title"] += "<br>Number of connections: {}".format(len(neigbour_map[node['id']])) 
+            node["value"] = len(neigbour_map[node['id']])
+
+        vis_path = os.path.join(("/").join(self.edge_path.split("/")[:-1]),"datanet.html")
+        data_net.show(vis_path)
+
     def _form_edge_study(self, df, all_study, study_id):
         '''
         For a given study id forms edge table

imgtools/modules/dose.py

@@ -112,8 +112,8 @@ def get_metadata(self):
                 # DVH specifc properties
                 doses_bin = np.cumsum(raw_data[0:n:2])
                 vol = raw_data[1:n:2]
-                self.dvh[ROI_reference]["dose_bins"] = doses_bin
-                self.dvh[ROI_reference]["vol"] = vol
+                self.dvh[ROI_reference]["dose_bins"] = doses_bin.tolist()
+                self.dvh[ROI_reference]["vol"] = vol.tolist()
 
                 # ROI specific properties
                 tot_vol = np.sum(vol)

imgtools/modules/pet.py

@@ -76,8 +76,8 @@ def get_metadata(self):
         except:
             pass
         try:
-            self.metadata["scan_time"] = datetime.datetime.strptime(self.df.AcquisitionTime, '%H%M%S.%f')
-            self.metadata["injection_time"] = datetime.datetime.strptime(self.df.RadiopharmaceuticalInformationSequence[0].RadiopharmaceuticalStartTime, '%H%M%S.%f')
+            self.metadata["scan_time"] = datetime.datetime.strptime(self.df.AcquisitionTime, '%H%M%S.%f').__str__()
+            self.metadata["injection_time"] = datetime.datetime.strptime(self.df.RadiopharmaceuticalInformationSequence[0].RadiopharmaceuticalStartTime, '%H%M%S.%f').__str__()
             self.metadata["half_life"] = float(self.df.RadiopharmaceuticalInformationSequence[0].RadionuclideHalfLife)
             self.metadata["injected_dose"] = float(self.df.RadiopharmaceuticalInformationSequence[0].RadionuclideTotalDose)
         except:

imgtools/ops/ops.py

@@ -69,11 +69,14 @@ class ImageAutoInput(BaseInput):
     modalities: str
         List of modalities to process. Only samples with ALL modalities will be processed. Make sure there are no space between list elements as it is parsed as a string.
 
+    visualize: bool
+        Whether to return visualization of the data graph
     """
     def __init__(self,
                  dir_path: str,
                  modalities: str,
-                 n_jobs: int = -1):
+                 n_jobs: int = -1,
+                 visualize: bool = False):
         self.dir_path = dir_path
         self.modalities = modalities
         self.dataset_name = self.dir_path.split("/")[-1]
@@ -93,24 +96,13 @@ def __init__(self,
         ####### GRAPH ##########
         # Form the graph
         edge_path = os.path.join(self.parent,f"imgtools_{self.dataset_name}_edges.csv")
-        graph = DataGraph(path_crawl=path_crawl,edge_path=edge_path)
+        graph = DataGraph(path_crawl=path_crawl,edge_path=edge_path,visualize=visualize)
         print(f"Forming the graph based on the given modalities: {self.modalities}")
         self.df_combined = graph.parser(self.modalities)
         self.output_streams = [("_").join(cols.split("_")[1:]) for cols in self.df_combined.columns if cols.split("_")[0]=="folder"]
         self.column_names = [cols for cols in self.df_combined.columns if cols.split("_")[0]=="folder"]
         self.series_names = [cols for cols in self.df_combined.columns if cols.split("_")[0]=="series"]
 
-        #Initilizations for the pipeline
-        for colnames in self.output_streams:
-            output_stream = ("_").join([items for items in colnames.split("_") if items!="1"])
-            modality = colnames.split("_")[0]
-            if modality in ["PT","CT","RTDOSE"]:
-                self.df_combined["size_{}".format(output_stream)] = None
-                if modality!="CT":
-                    self.df_combined["metadata_{}".format(output_stream)] = None
-            elif modality=="RTSTRUCT":
-                self.df_combined["roi_names_{}".format(output_stream)] = None
-
         print(f"There are {len(self.df_combined)} cases containing all {modalities} modalities.")
 
         self.readers = [read_dicom_auto for i in range(len(self.output_streams))]