Expose pose kwargs in RegisterFixed

src/xvr/register/base.py

@@ -48,9 +48,16 @@ class RegisterBase(ABC):
     def __init_subclass__(cls, **kwargs):
         super().__init_subclass__(**kwargs)
         RegisterBase._registry[cls.__name__.replace("Register", "").lower()] = cls
+
         base_sig = signature(RegisterBase.__init__)
         cls.__init__.__signature__ = _merge_signatures(signature(cls.__init__), base_sig)
 
+        if "__call__" in cls.__dict__:
+            cls.__call__.__signature__ = _merge_signatures(
+                signature(cls.__call__),
+                signature(RegisterBase.__call__),
+            )
+
     @classmethod
     def create(cls, method: str, **kwargs):
         if method not in cls._registry:

src/xvr/register/register.py

@@ -6,6 +6,7 @@
 from ..model.inference import predict_pose
 from ..model.modules.network import load_model
 from .base import RegisterBase
+from .logging import RegistrationResult
 
 
 class RegisterFixed(RegisterBase):
@@ -15,6 +16,29 @@ class RegisterFixed(RegisterBase):
     when initializing from a prior scan or a clinical estimate.
     """
 
+    def __call__(
+        self,
+        filename: str,
+        rot: tuple[float, float, float],
+        xyz: tuple[float, float, float],
+        orientation: str = "AP",
+        isocenter: bool = True,
+        **kwargs,
+    ) -> RegistrationResult:
+        """Run registration with a manually specified initial pose.
+
+        Args:
+            filename: Path to the X-ray image.
+            rot: Rotation angles (in degrees) as (rx, ry, rz).
+            xyz: Translation (in mm) as (x, y, z).
+            orientation: Starting orientation of the volume, e.g. "AP" or "lateral".
+            isocenter: If True, centers the pose at the subject isocenter.
+            **kwargs: See RegisterBase.__call__ for remaining arguments.
+        """
+        return super().__call__(
+            filename, rot=rot, xyz=xyz, orientation=orientation, isocenter=isocenter, **kwargs
+        )
+
     def get_initial_pose_estimate(
         self,
         _img,
@@ -68,7 +92,7 @@ def get_initial_pose_estimate(
         """Predict the initial pose from the X-ray using a neural network.
 
         Args:
-            gt: Preprocessed ground truth X-ray image.
+            img: Preprocessed ground truth X-ray image.
             intrinsics: Camera intrinsics for the X-ray.
 
         Returns: