Hard-coding postop part, renamed HGG to GBM.

gui/SinglePatientComponent/CentralAreaExecutionWidget.py

@@ -151,8 +151,8 @@ def on_pipeline_execution(self, pipeline_code: str) -> None:
             if code == 0:  # Operation cancelled
                 return
 
-            if diag.tumor_type == 'High-Grade Glioma':
-                self.model_name = "MRI_HGGlioma"
+            if diag.tumor_type == 'Glioblastoma':
+                self.model_name = "MRI_GBM"
             elif diag.tumor_type == 'Low-Grade Glioma':
                 self.model_name = "MRI_LGGlioma"
             elif diag.tumor_type == 'Metastasis':

gui/StudyBatchComponent/StudiesSidePanel/SingleStudyWidget.py

@@ -591,8 +591,8 @@ def __on_run_pipeline(self) -> None:
             if code == 0:  # Operation was cancelled by the user
                 return
 
-            if diag.tumor_type == 'High-Grade Glioma':
-                self.model_name = "MRI_HGGlioma"
+            if diag.tumor_type == 'Glioblastoma':
+                self.model_name = "MRI_GBM"
             elif diag.tumor_type == 'Low-Grade Glioma':
                 self.model_name = "MRI_LGGlioma"
             elif diag.tumor_type == 'Metastasis':

gui/UtilsWidgets/CustomQDialog/TumorTypeSelectionQDialog.py

@@ -14,8 +14,8 @@ def __init__(self, parent=None):
         self.select_tumor_type_label.setStyleSheet("""QLabel{background-color: rgba(248, 248, 248, 1);}""")
         self.base_layout.addWidget(self.select_tumor_type_label, 0, 0)
         self.select_tumor_type_combobox = QComboBox()
-        self.select_tumor_type_combobox.addItems(["High-Grade Glioma", "Low-Grade Glioma", "Meningioma", "Metastasis"])
-        self.tumor_type = "High-Grade Glioma"
+        self.select_tumor_type_combobox.addItems(["Glioblastoma", "Low-Grade Glioma", "Meningioma", "Metastasis"])
+        self.tumor_type = "Glioblastoma"
 
         self.base_layout.addWidget(self.select_tumor_type_combobox, 0, 1)
         self.exit_accept_pushbutton = QDialogButtonBox(QDialogButtonBox.Ok)

utils/logic/PipelineCreationHandler.py

@@ -31,7 +31,7 @@ class ModelNameType(Enum):
     """
     _init_ = 'value string'
 
-    HGGlioma = 0, 'MRI_HGGLioma'
+    HGGlioma = 0, 'MRI_GBM'
     LGGlioma = 1, 'MRI_LGGlioma'
     Meningioma = 2, 'MRI_Meningioma'
     Metastasis = 3, 'MRI_Metastase'
@@ -47,6 +47,7 @@ def create_pipeline(model_name: str, patient_parameters, task: str) -> dict:
     How to allow for all possible combinations of what to use/what to run/on which timestamps?
 
     @TODO. Still heavily hard-coded atm, will need to rely only on the pipeline json files
+    @TODO. Postop segmentation not adapting to the number of inputs, defaulting to the 4 inputs.
     Returns
     -------
     dict
@@ -61,14 +62,14 @@ def create_pipeline(model_name: str, patient_parameters, task: str) -> dict:
     elif task == 'preop_segmentation':
         return __create_segmentation_pipeline(model_name, patient_parameters)
     elif task == 'postop_segmentation':
-        download_model(model_name='MRI_Tumor_Postop')
+        download_model(model_name='MRI_GBM_Postop_FV_4p')
         return __create_postop_segmentation_pipeline(model_name, patient_parameters)
     elif task == 'other_segmentation':
         return __create_other_segmentation_pipeline(model_name, patient_parameters)
     elif task == 'preop_reporting':
         return __create_preop_reporting_pipeline(model_name, patient_parameters)
     elif task == 'postop_reporting':
-        download_model(model_name='MRI_Tumor_Postop')
+        download_model(model_name='MRI_GBM_Postop_FV_4p')
         return __create_postop_reporting_pipeline(model_name, patient_parameters)
     else:
         return __create_custom_pipeline(task, model_name, patient_parameters)
@@ -158,9 +159,15 @@ def __create_other_segmentation_pipeline(model_name, patient_parameters):
 
 def __create_postop_segmentation_pipeline(model_name, patient_parameters):
     """
-
+    The default postop segmentation model is the one with four inputs, but based on the loaded images another fitting
+    model could be used.
+    @TODO. Ideally, in the future, the disambiguation of the best model to use should be done in the backend. In that
+    case, how to properly retrieve the corresponding pipeline.json?
     """
-    infile = open(os.path.join(SoftwareConfigResources.getInstance().models_path, "MRI_Tumor_Postop", 'pipeline.json'), 'rb')
+    postop_model_name = "MRI_GBM_Postop_FV_4p"
+    if SoftwareConfigResources.getInstance().user_preferences.use_manual_sequences:
+        postop_model_name = select_appropriate_postop_model(patient_parameters)
+    infile = open(os.path.join(SoftwareConfigResources.getInstance().models_path, postop_model_name, 'pipeline.json'), 'rb')
     raw_pip = json.load(infile)
 
     pip = {}
@@ -226,6 +233,7 @@ def __create_preop_reporting_pipeline(model_name, patient_parameters):
         pip[pip_num]["target"] = ["Brain"]
         pip[pip_num]["model"] = "MRI_Brain"
         pip[pip_num]["description"] = "Brain segmentation in T1CE (T0)"
+        download_model("MRI_Brain")
 
     pip_num_int = pip_num_int + 1
     pip_num = str(pip_num_int)
@@ -493,9 +501,8 @@ def __create_postop_reporting_pipeline(model_name, patient_parameters):
     pip[pip_num]["inputs"]["4"]["space"]["timestamp"] = 1
     pip[pip_num]["inputs"]["4"]["space"]["sequence"] = "T1-CE"
     pip[pip_num]["target"] = ["Tumor"]
-    pip[pip_num]["model"] = "MRI_Tumor_Postop"
+    pip[pip_num]["model"] = "MRI_GBM_Postop_FV_4p"
     pip[pip_num]["description"] = "Tumor segmentation in T1CE (T1)"
-    # download_model(model_name='MRI_Tumor_Postop')
 
     pip_num_int = pip_num_int + 1
     pip_num = str(pip_num_int)
@@ -589,3 +596,13 @@ def __create_custom_pipeline(task, tumor_type, patient_parameters):
                 pip[pip_num]["description"] = k + " segmentation in T1CE (T{})".format(str(timestamp_order))
                 download_model(model_name=model_name)
     return pip
+
+
+def select_appropriate_postop_model(patient_parameters) -> str:
+    """
+    Temporary method, which will be deported in the backend, for selecting the best postoperative glioblastoma
+    segmentation model based on available inputs.
+    """
+    model_name = "MRI_GBM_Postop_FV_4p"
+
+    return model_name

utils/models_download.py

@@ -100,7 +100,8 @@ def download_model(model_name: str):
             model_params = cloud_models_list.loc[cloud_models_list['Model'] == model_name]
             url = model_params['link'].values[0]
             md5 = model_params['sum'].values[0]
-            dep = list(model_params['dependencies'].values)
+            tmp_dep = model_params['dependencies'].values[0]
+            dep = list(tmp_dep.strip().split(';')) if tmp_dep == tmp_dep else []
             models_path = os.path.join(expanduser('~'), '.raidionics', 'resources', 'models')
             os.makedirs(models_path, exist_ok=True)
             models_archive_path = os.path.join(expanduser('~'), '.raidionics', 'resources', 'models',
@@ -112,6 +113,9 @@ def download_model(model_name: str):
                 download_state = True
 
             if download_state:
+                if os.path.exists(models_archive_path):
+                    # Just in case, deleting the old cached archive, if a new one is to be downloaded
+                    os.remove(models_archive_path)
                 headers = {}
 
                 response = requests.get(url, headers=headers, stream=True)