Merge pull request #442 from michaelhallquist/1.9.1.8-HOTFIX

Fix for SUSAN double FWHM -> sigma problem

clpipe/config/package.py

@@ -1,5 +1,5 @@
 PACKAGE_NAME = "clpipe"
-VERSION = "1.9.1.7"
+VERSION = "1.9.1.8"
 
 DESCRIPTION = "clpipe: MRI processing pipeline for high performance clusters"
 REPO_URL = "https://github.com/cohenlabUNC/clpipe"

clpipe/postprocutils/image_workflows.py

@@ -528,11 +528,6 @@ def build_SUSAN_workflow(
     if crashdump_dir is not None:
         workflow.config["execution"]["crashdump_dir"] = crashdump_dir
 
-    # Calculate fwhm
-    fwhm_to_sigma = sqrt(8 * log(2))
-    sigma = fwhm_mm / fwhm_to_sigma
-    print(f"fwhm_to_sigma: {fwhm_to_sigma}")
-
     # Setup identity (pass through) input/output nodes
     input_node = build_input_node()
     output_node = build_output_node()
@@ -563,7 +558,9 @@ def build_SUSAN_workflow(
         ),
         name="setup_usans",
     )
-    susan_node = pe.Node(SUSAN(fwhm=sigma, use_median=1, dimension=3), name="SUSAN")
+
+    # N.B. the SUSAN node in nipype does the FWHM -> sigma conversion, so we should not!
+    susan_node = pe.Node(SUSAN(fwhm=fwhm_mm, use_median=1, dimension=3), name="SUSAN")
 
     # Set WF inputs and outputs
     if in_file:

docs/changelog.md

@@ -1,5 +1,9 @@
 # Change Log
 
+## 1.9.1.8 (Oct 4, 2024)
+
+- Fix `fwhm_mm` usage in `SUSAN`, which was converting to sigma (SD of Gaussian), but then nipype applied that conversion, too.
+
 ## 1.9.1 (Dec 7, 2023)
 
 ### Enhancements