Fix unspecified spatial size for rank-one resampling and add NIfTI regression tests

monai/transforms/spatial/functional.py

@@ -162,8 +162,11 @@ def spatial_resample(
     in_spatial_size = torch.tensor(original_spatial_shape[:spatial_rank])
     if isinstance(spatial_size, int) and (spatial_size == -1):  # using the input spatial size
         spatial_size = in_spatial_size
-    elif spatial_size is None and spatial_rank > 1:  # auto spatial size
-        spatial_size, _ = compute_shape_offset(in_spatial_size, src_affine, dst_affine)  # type: ignore
+    elif spatial_size is None:
+        if spatial_rank > 1:  # auto spatial size
+            spatial_size, _ = compute_shape_offset(in_spatial_size, src_affine, dst_affine)  # type: ignore
+        else:
+            spatial_size = in_spatial_size
     spatial_size = torch.tensor(
         fall_back_tuple(ensure_tuple(spatial_size)[:spatial_rank], in_spatial_size, lambda x: x is not None and x >= 0)
     )

tests/data/test_nifti_rw.py

@@ -249,6 +249,26 @@ def test_write_5d(self):
                 )
                 np.testing.assert_allclose(out.affine, np.diag([1.4, 2, 2, 1]), atol=1e-4, rtol=1e-4)
 
+    @parameterized.expand([((64, 64),), ((8, 64, 64),)])
+    def test_write_without_original_affine_and_spatial_shape(self, shape):
+        """Verify channel-less NIfTI writing preserves shape, voxel values, and affine without target metadata.
+
+        Args:
+            shape: Shape of the channel-less 2D or 3D input array.
+        """
+        data = np.arange(np.prod(shape), dtype=np.float32).reshape(shape)
+        with tempfile.TemporaryDirectory() as out_dir:
+            image_name = os.path.join(out_dir, "test.nii")
+            writer_obj = NibabelWriter(output_dtype=np.float32)
+            writer_obj.set_data_array(data, channel_dim=None, spatial_ndim=None)
+            writer_obj.set_metadata({"affine": np.eye(4), "original_affine": None})
+            writer_obj.write(image_name)
+
+            saved = nib.load(image_name)
+            self.assertEqual(saved.shape, data.shape)
+            np.testing.assert_array_equal(saved.get_fdata(), data)
+            np.testing.assert_allclose(saved.affine, np.eye(4))
+
 
 if __name__ == "__main__":
     unittest.main()

tests/transforms/test_spatial_resample.py

@@ -222,6 +222,20 @@ def test_inverse(self, img, device, data_param, expected_output):
         expected_affine = to_affine_nd(len(out.affine) - 1, torch.eye(4))
         assert_allclose(out.affine, expected_affine)
 
+    @parameterized.expand([(False, None), (False, -1), (True, None), (True, -1)])
+    def test_1d_unchanged(self, lazy, spatial_size):
+        """Verify rank-one no-op resampling preserves values, pending shape, and affine.
+
+        Args:
+            lazy: Whether to record the resampling lazily instead of executing eagerly.
+            spatial_size: Output size setting, either None or -1, expected to preserve the input size.
+        """
+        img = MetaTensor(torch.arange(8, dtype=torch.float32).reshape(1, 8))
+        result = SpatialResample(lazy=lazy)(img, spatial_size=spatial_size)
+        assert_allclose(result, img)
+        assert_allclose(result.peek_pending_affine(), img.affine)
+        self.assertEqual(tuple(result.peek_pending_shape()), (8,))
+
     def test_unchange(self):
         for i, p in enumerate(TEST_NDARRAYS_ALL):
             set_track_meta(i % 2)