RF: Listify and sort BOLD series by EchoTime once [no fasttrack]

Author: effigies

fmriprep/workflows/base.py

@@ -216,6 +216,14 @@ def init_single_subject_wf(subject_id: str):
             )
         )
 
+    bold_runs = [
+        sorted(
+            listify(run),
+            key=lambda file: config.execution.layout.get_metadata(file).get('EchoTime', 0),
+        )
+        for run in subject_data['bold']
+    ]
+
     if subject_data['roi']:
         warnings.warn(
             f"Lesion mask {subject_data['roi']} found. "
@@ -393,7 +401,7 @@ def init_single_subject_wf(subject_id: str):
     fmap_estimators, estimator_map = map_fieldmap_estimation(
         layout=config.execution.layout,
         subject_id=subject_id,
-        bold_data=subject_data['bold'],
+        bold_data=bold_runs,
         ignore_fieldmaps="fieldmaps" in config.workflow.ignore,
         use_syn=config.workflow.use_syn_sdc,
         force_syn=config.workflow.force_syn,
@@ -509,11 +517,10 @@ def init_single_subject_wf(subject_id: str):
 : For each of the {num_bold} BOLD runs found per subject (across all
 tasks and sessions), the following preprocessing was performed.
 """.format(
-        num_bold=len(subject_data['bold'])
+        num_bold=len(bold_runs)
     )
 
-    for bold_series in subject_data['bold']:
-        bold_series = sorted(listify(bold_series))
+    for bold_series in bold_runs:
         bold_file = bold_series[0]
         fieldmap_id = estimator_map.get(bold_file)
 
@@ -610,7 +617,7 @@ def init_single_subject_wf(subject_id: str):
 def map_fieldmap_estimation(
     layout: bids.BIDSLayout,
     subject_id: str,
-    bold_data: list,
+    bold_data: list[list[str]],
     ignore_fieldmaps: bool,
     use_syn: bool | str,
     force_syn: bool,
@@ -656,7 +663,7 @@ def map_fieldmap_estimation(
     # Pare down estimators to those that are actually used
     # If fmap_estimators == [], all loops/comprehensions terminate immediately
     all_ids = {fmap.bids_id for fmap in fmap_estimators}
-    bold_files = (sorted(listify(bold_file))[0] for bold_file in bold_data)
+    bold_files = (bold_series[0] for bold_series in bold_data)
 
     all_estimators = {
         bold_file: [fmap_id for fmap_id in get_estimator(layout, bold_file) if fmap_id in all_ids]

fmriprep/workflows/bold/fit.py

@@ -118,7 +118,7 @@ def init_bold_fit_wf(
     Parameters
     ----------
     bold_series
-        List of paths to NIfTI files.
+        List of paths to NIfTI files, sorted by echo time.
     precomputed
         Dictionary containing precomputed derivatives to reuse, if possible.
     fieldmap_id
@@ -204,27 +204,26 @@ def init_bold_fit_wf(
 
     layout = config.execution.layout
 
-    # Collect bold and sbref files, sorted by EchoTime
-    bold_files = sorted(bold_series, key=lambda fname: layout.get_metadata(fname).get("EchoTime"))
+    # Collect sbref files, sorted by EchoTime
     sbref_files = get_sbrefs(
-        bold_files,
+        bold_series,
         entity_overrides=config.execution.get().get('bids_filters', {}).get('sbref', {}),
         layout=layout,
     )
 
     # Fitting operates on the shortest echo
     # This could become more complicated in the future
-    bold_file = bold_files[0]
+    bold_file = bold_series[0]
 
     # Get metadata from BOLD file(s)
-    entities = extract_entities(bold_files)
+    entities = extract_entities(bold_series)
     metadata = layout.get_metadata(bold_file)
     orientation = "".join(nb.aff2axcodes(nb.load(bold_file).affine))
 
     bold_tlen, mem_gb = estimate_bold_mem_usage(bold_file)
 
     # Boolean used to update workflow self-descriptions
-    multiecho = len(bold_files) > 1
+    multiecho = len(bold_series) > 1
 
     have_hmcref = "hmc_boldref" in precomputed
     have_coregref = "coreg_boldref" in precomputed
@@ -699,27 +698,20 @@ def init_bold_native_wf(
     layout = config.execution.layout
 
     # Shortest echo first
-    all_metadata, bold_files, echo_times = zip(
-        *sorted(
-            (
-                (md := layout.get_metadata(bold_file), bold_file, md.get("EchoTime"))
-                for bold_file in listify(bold_series)
-            ),
-            key=lambda x: x[2],
-        )
-    )
-    multiecho = len(bold_files) > 1
+    all_metadata = [layout.get_metadata(bold_file) for bold_file in bold_series]
+    echo_times = [md.get("EchoTime") for md in all_metadata]
+    multiecho = len(bold_series) > 1
 
-    bold_file = bold_files[0]
+    bold_file = bold_series[0]
     metadata = all_metadata[0]
 
     bold_tlen, mem_gb = estimate_bold_mem_usage(bold_file)
 
     if multiecho:
-        shapes = [nb.load(echo).shape for echo in bold_files]
+        shapes = [nb.load(echo).shape for echo in bold_series]
         if len(set(shapes)) != 1:
             diagnostic = "\n".join(
-                f"{os.path.basename(echo)}: {shape}" for echo, shape in zip(bold_files, shapes)
+                f"{os.path.basename(echo)}: {shape}" for echo, shape in zip(bold_series, shapes)
             )
             raise RuntimeError(f"Multi-echo images found with mismatching shapes\n{diagnostic}")
         if len(shapes) == 2:
@@ -774,12 +766,12 @@ def init_bold_native_wf(
     # almost identically
     echo_index = pe.Node(niu.IdentityInterface(fields=["echoidx"]), name="echo_index")
     if multiecho:
-        echo_index.iterables = [("echoidx", range(len(bold_files)))]
+        echo_index.iterables = [("echoidx", range(len(bold_series)))]
     else:
         echo_index.inputs.echoidx = 0
 
     # BOLD source: track original BOLD file(s)
-    bold_source = pe.Node(niu.Select(inlist=bold_files), name="bold_source")
+    bold_source = pe.Node(niu.Select(inlist=bold_series), name="bold_source")
     validate_bold = pe.Node(ValidateImage(), name="validate_bold")
     workflow.connect([
         (echo_index, bold_source, [("echoidx", "index")]),