Enh patch sampling

examples/tutorials/2025-paper-fig5.py

@@ -0,0 +1,29 @@
+# %%
+import siibra
+assert siibra.__version__ >= "1.0.1"
+import matplotlib.pyplot as plt
+
+# %%
+# 1: Retrieve probability map of a motor area in Julich-Brain
+parc = siibra.parcellations.get('julich 3.1')
+region = parc.get_region("4p right")
+pmap = parc.get_map('mni152', 'statistical').get_volume(region)
+
+# %%
+# 2: Extract BigBrain 1 micron patches with high probability in this area
+patches = siibra.features.get(pmap, "BigBrain1MicronPatch", lower_threshold=0.7)
+print(f"Found {len(patches)} patches.")
+
+# %%
+# 3: Display highly rated samples, here further reduced to a predefined section
+section = 3556
+candidates = filter(lambda p: p.bigbrain_section == 3556, patches)
+f, axs = plt.subplots(1, 3, figsize=(8, 24))
+for patch, ax in zip(list(candidates)[:3], axs.ravel()):
+    patchdata = patch.fetch().get_fdata().squeeze()
+    ax.imshow(patchdata, cmap='gray', vmin=0, vmax=2**16)
+    ax.axis('off')
+    ax.set_title(f"#{section} - {patch.vertex}", fontsize=10)
+
+
+# %%

siibra/__init__.py

@@ -36,10 +36,9 @@
 from .retrieval.cache import Warmup, WarmupLevel
 
 from . import configuration
-from . import experimental
 from .configuration import factory
 from . import features, livequeries
-from siibra.locations import Point, PointCloud
+from siibra.locations import Point, PointCloud, Plane, BoundingBox
 
 import os as _os
 logger.info(f"Version: {__version__}")
@@ -151,6 +150,7 @@ def __dir__():
         "MapType",
         "Point",
         "PointCloud",
+        "BoundingBox",
         "QUIET",
         "VERBOSE",
         "fetch_ebrains_token",

siibra/features/image/__init__.py

@@ -22,6 +22,8 @@
     XPCTVolumeOfInterest,
     LSFMVolumeOfInterest,
     DTIVolumeOfInterest
-    # SegmentedVolumeOfInterest
 )
-from .sections import CellbodyStainedSection
+from .sections import (
+    CellbodyStainedSection,
+    BigBrain1MicronPatch
+)

siibra/features/image/sections.py

@@ -15,12 +15,69 @@
 """Multimodal data features in 2D section."""
 
 from . import image
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from ...locations import AxisAlignedPatch
+    from ...features.anchor import AnatomicalAnchor
 
 
 class CellbodyStainedSection(
     image.Image,
-    configuration_folder='features/images/sections/cellbody',
-    category="cellular"
+    configuration_folder="features/images/sections/cellbody",
+    category="cellular",
 ):
     def __init__(self, **kwargs):
         image.Image.__init__(self, **kwargs, modality="cell body staining")
+
+
+class BigBrain1MicronPatch(image.Image, category="cellular"):
+    def __init__(
+        self,
+        patch: "AxisAlignedPatch",
+        section: CellbodyStainedSection,
+        vertex: int,
+        relevance: float,
+        anchor: "AnatomicalAnchor",
+        **kwargs
+    ):
+        self._patch = patch
+        self._section = section
+        self.vertex = vertex
+        self.relevance = relevance
+        image.Image.__init__(
+            self,
+            name=f"Cortical patch in {section.name}",
+            modality=section.modality,
+            space_spec=section._space_spec,
+            providers=list(section._providers.values()),
+            region=None,
+            datasets=section.datasets,
+            bbox=patch.boundingbox,
+            id=None
+        )
+        self._anchor_cached = anchor
+
+    def __repr__(self):
+        return (
+            f"<{self.__class__.__name__}(space_spec={self._space_spec}, "
+            f"name='{self.name}', "
+            f"section='{self._section.get_boundingbox().minpoint.bigbrain_section()}', "
+            f"vertex='{self.vertex}', providers={self._providers})>"
+        )
+
+    @property
+    def bigbrain_section(self):
+        return self.get_boundingbox().minpoint.bigbrain_section()
+
+    def fetch(self, flip=False, **kwargs):
+        assert "voi" not in kwargs
+        res = kwargs.get("resolution_mm", -1)
+        if flip:
+            return self._patch.flip().extract_volume(
+                self._section, resolution_mm=res
+            ).fetch()
+        else:
+            return self._patch.extract_volume(
+                self._section, resolution_mm=res
+            ).fetch()

siibra/features/image/volume_of_interest.py

@@ -79,10 +79,3 @@ class LSFMVolumeOfInterest(
     def __init__(self, modality, **kwargs):
         image.Image.__init__(self, **kwargs, modality=modality)
 
-# class SegmentedVolumeOfInterest(
-#     image.Image,
-#     configuration_folder="features/images/vois/segmentation",
-#     category="segmentation"
-# ):
-#     def __init__(self, **kwargs):
-#         image.Image.__init__(self, **kwargs, modality="segmentation")