fix constraint for pixdim in track_transform_meta() function

Signed-off-by: Wei_Chuan, Chiang <jamesis1356@gmail.com>
Co-authored-by: einsyang723 <tina910723@gmail.com>
Co-authored-by: IamTingTing  <6121smile@gmail.com>

Author: 3 people

monai/transforms/inverse.py

@@ -197,8 +197,9 @@ def track_transform_meta(
                 else:
                     raise
             out_obj.meta[MetaKeys.AFFINE] = convert_to_tensor(affine, device=torch.device("cpu"), dtype=torch.float64)
-            spacing = affine_to_spacing(out_obj.meta[MetaKeys.AFFINE])
-            out_obj.meta[MetaKeys.PIXDIM][1 : 1 + len(spacing)] = spacing
+            if MetaKeys.PIXDIM in out_obj.meta:
+                spacing = affine_to_spacing(out_obj.meta[MetaKeys.AFFINE])
+                out_obj.meta[MetaKeys.PIXDIM][1 : 1 + len(spacing)] = spacing
 
         if not (get_track_meta() and transform_info and transform_info.get(TraceKeys.TRACING)):
             if isinstance(data, Mapping):