Verify each finding against the current code and only fix it if needed.

Nitpick comments:
In `@experiments/Convert_VTK_To_USD/convert_chop_tpv25_valve_to_usd.ipynb`:
- Around line 388-390: Replace the hardcoded parameters intensity_range=(25,
200), cmap="jet", use_sigmoid_scale=True with a single shared colormap preset
(e.g., TPV25_COLORMAP_PRESET or get_tpv25_colormap_preset()) and pass that
preset into the call; update the call site that currently supplies
intensity_range, cmap and use_sigmoid_scale to instead read those values from
the preset so all TPV25 colorization paths share the same mapping policy.

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@experiments/Convert_VTK_To_USD/convert_chop_alterra_valve_to_usd.ipynb`:
- Line 202: The section header "## 3. Convert TPV25" is incorrect; update the
notebook markdown cell that contains that exact string (the header "## 3.
Convert TPV25") to accurately describe the conversion being performed, e.g.
change it to "## 3. Convert Alterra" or "## 3. Convert Alterra valve to USD" so
the section title matches the Alterra conversion code and context.
- Around line 134-142: The code collects matches into alterra_series from
vtk_files using pattern but doesn't handle the case where no frames match;
before any access (e.g., before using alterra_series[0] later), check if
alterra_series is empty and raise a clear, dataset-specific exception (e.g.,
ValueError or a custom AlterraDatasetError) with a message that includes the
pattern and the source directory or vtk_files summary so callers can diagnose
missing frames/naming drift; perform this check immediately after
alterra_series.sort(...) and before downstream uses.
- Around line 96-99: The colormap configuration currently allows both "stress"
and "strain" in colormap_primvar_substrs which can silently pick strain and
misapply the fixed colormap_range_min/colormap_range_max; change
colormap_primvar_substrs to only ["stress"] and ensure downstream usage (where
colormap_name, colormap_range_min, colormap_range_max are applied) is based on
that stress-only list (or add a simple assertion that the selected primvar
equals "stress") so the fixed (25,200) range is only used for stress
visualizations.

In `@experiments/Convert_VTK_To_USD/convert_chop_tpv25_valve_to_usd.ipynb`:
- Around line 258-263: The code pins vessel_path to a specific prim name
("TPV25Valve_object4"), which is brittle; instead detect the correct mesh prim
under the TPV25Valve meshes at runtime. Modify the logic around
conversion_settings.separate_objects_by_connectivity /
separate_objects_by_cell_type to enumerate children of
"/World/Meshes/TPV25Valve" (or the parent used), inspect prim names/attributes
or topology/point-count to find the intended mesh (e.g., the prim whose
topology/face-count or connectivity attribute matches the source VTK component),
and set vessel_path to that discovered prim path before the primvar lookup that
follows (the variable currently named vessel_path and the primvar query that
relies on it). Ensure fallback to the single mesh path when no separate objects
exist.
- Around line 132-140: The code collects matching VTK files into tpv25_series by
iterating over vtk_files with the regex pattern and then sorts it; add a guard
after tpv25_series.sort(...) that checks if tpv25_series is empty and raise a
FileNotFoundError with a clear message (including pattern and/or input directory
context) so later access like tpv25_series[0] fails fast and provides actionable
information. Ensure the check references the same variable name tpv25_series and
include enough context in the exception message to identify the missing TPV25
frames.
- Around line 94-97: The colormap source list colormap_primvar_substrs currently
allows both "stress" and "strain", which can let TPV25 use a different scalar
while still using the same colormap bounds
(colormap_range_min/colormap_range_max) and thus break the "consistent stress
colormap" requirement; update colormap_primvar_substrs to include only "stress"
so the notebook always uses stress as the scalar for colorization (leave
colormap_name and colormap_range_min/colormap_range_max as-is unless there’s a
separate range for strain to adjust).

Verify each finding against the current code and only fix it if needed.

Inline comments:
In `@CLAUDE.md`:
- Around line 55-57: The fenced code block containing "4D CT → Segmentation →
Registration → Contour Extraction → USD Export" needs a language tag to satisfy
markdownlint MD040; update the fence opening from ``` to ```text (or another
appropriate language) so the block becomes a labeled code fence (refer to the
existing fenced block content in CLAUDE.md).

In `@experiments/Convert_VTK_To_USD/convert_chop_heart_vtk_to_usd.ipynb`:
- Around line 45-46: The cleanup currently checks and removes
RVOT28-Dias-{usdname}.usd in the CWD using Path.absolute(Path(...)) and
os.path.exists/os.remove, which doesn't match the later write to output_dir;
update the cleanup to target the same output_dir by building the target path via
output_dir / f"RVOT28-Dias-{usdname}.usd" (or output_dir.joinpath(...)) and use
pathlib's exists() and unlink() methods instead of os.path.exists/os.remove to
delete any stale file before writing; adjust references around usdname and
output_dir to ensure the same path is used for both cleanup and file creation.

In
`@experiments/Heart-Statistical_Model_To_Patient/heart_model_to_patient-CHOPValve.ipynb`:
- Around line 124-128: The notebook saves the UnstructuredGrid stored in
results["registered_template_model"] to a .vtp file which is incorrect; change
the save call for registered_model to write a .vtu file while leaving
registered_model_surface (PolyData) saved as .vtp. Locate the symbols
registered_model (from results["registered_template_model"]) and
registered_model_surface (from results["registered_template_model_surface"]) and
update the registered_model.save invocation to use "registered_model.vtu"
(keeping output_dir and the existing path construction).

In `@experiments/Reconstruct4DCT/reconstruct_4d_ct_class.ipynb`:
- Around line 87-94: The comment and enabled methods are inverted: restore ICON
and ants_icon to the full-run configuration by updating registration_methods
(currently ["ants"]) to include "icon" and "ants_icon" and align
number_of_iterations_list to include the corresponding iteration settings for
each method (match entries to "ants", "icon", and "ants_icon"), and update the
preceding comment text (the string on line with registration_methods) to
accurately describe that full_run includes ANTs, ICON, and ants_icon;
alternatively, if you prefer to keep quick_run with fewer methods, remove "icon"
and "ants_icon" from the quick_run list (the variable quick_run referenced
earlier) so the quick-run and full-run semantics are consistent.

In `@src/physiomotion4d/contour_tools.py`:
- Around line 179-182: The code assumes mesh.n_faces_strict > 0 and uniformly
triangulated faces; add a guard in the contour processing path (where
extract_contours() may return an empty mesh) to early-return or set
faces/vertices to empty when mesh.n_faces_strict == 0, and before computing
num_points_per_face and reshaping ensure the mesh is explicitly triangulated
(call the triangulation helper or convert polygonal faces to triangles) and
validate that len(mesh.faces) is divisible by mesh.n_faces_strict; update the
block referencing mesh.faces and mesh.n_faces_strict to perform these checks and
triangulation so the reshape cannot raise ZeroDivisionError or produce
mis-shaped arrays.

In `@src/physiomotion4d/segment_heart_simpleware.py`:
- Around line 286-297: The code unconditionally opens and parses landmarks_file
which makes the whole segmentation fail if landmarks.csv is missing or
malformed; modify the block around landmarks_file and self.landmarks.clear() to
be resilient: first check os.path.exists(landmarks_file) or wrap the open/csv
parsing in a try/except that catches FileNotFoundError, csv.Error and
ValueError, log a warning and leave self.landmarks empty (do not re-raise), and
when parsing each row validate coords length and numeric conversion before
assigning to self.landmarks so malformed rows are skipped with a warning; refer
to landmarks_file, self.landmarks.clear(), csv.DictReader and the
Measurement/Name parsing to locate the code to change.

In `@src/physiomotion4d/workflow_fit_statistical_model_to_patient.py`:
- Around line 155-167: The code assumes patient_models_data["meshes"]["heart"]
always exists after calling WorkflowConvertCTToVTK.run_workflow; update the
branch that sets patient_models to first heart mesh to first check that "meshes"
in patient_models_data and "heart" in patient_models_data["meshes"] and that
patient_models_data["meshes"]["heart"] is non-empty, and if any check fails
raise a clear ValueError (or fallback) indicating the heart mesh was not
produced and advising to provide patient_models or verify segmentation settings;
reference the WorkflowConvertCTToVTK.run_workflow call and the
patient_models_data / patient_models variables when implementing the guard.

---

Nitpick comments:
In `@experiments/Convert_VTK_To_USD/convert_chop_heart_vtk_to_usd.ipynb`:
- Line 12: Remove the unused "import os" at the top of the notebook and refactor
any path cleanup code that still uses os functions to use pathlib.Path methods
(e.g., replace os.path.exists / os.remove with Path.exists() and Path.unlink())
so the os import is no longer needed; search for usages around the path cleanup
logic and update the code to call Path(...).exists() and Path(...).unlink()
instead.

In `@src/physiomotion4d/segment_heart_simpleware.py`:
- Around line 327-330: The get_landmarks method currently returns the internal
mutable dict self.landmarks directly; change it to return a defensive copy to
prevent external mutation (e.g., return a shallow copy of the dict so callers
can't modify the internal mapping). Update the get_landmarks function to return
a copy of self.landmarks (or a deep copy if landmarks values may later become
mutable) while keeping the same return type and behavior otherwise.

In `@src/physiomotion4d/workflow_fit_statistical_model_to_patient.py`:
- Around line 207-208: Update the docstrings for the setter properties that
document mask_dilation_mm and roi_dilation_mm to reflect the new defaults:
change the mask_dilation_mm setter docstring default from "Default: 5mm" to
"Default: 0.0mm" and change the roi_dilation_mm setter docstring default from
"Default: 20mm" to "Default: 25.0mm"; locate the setters that reference
mask_dilation_mm and roi_dilation_mm (the property/method names for those
setters) and update their docstring default lines only so the docs match the
actual defaults assigned in the initializer.