Methods: Subjects with head and neck cancer underwent a 64Cu-ATSM PET-CT scan a week before surgery. After surgery, sea urchin spines, which can be identified with CT, optically and histologically, were inserted into the specimen to act as fiducial markers. The specimen was fixed and scanned CT ex-vivo. After slicing, blockface images were obtained for visual reference. From these thick sectioned slices, a subsection of tissue that included tumour and markers was extracted and embedded in paraffin blocks of size 30x21mm. Subsequently microtome sectioning and haematoxylin and eosin staining was performed to acquire 5μm thick slides and digitised using a microscope. A series of rigid registration steps were performed between the different imaging modalities. The PET and CT (in-vivo) images were registered with an accuracy  of 5.4mm. Anatomical landmarks were used for in-vivo and ex-vivo CT registration and the fiducial markers for histology to ex-vivo CT registration. Inter-marker distances were used to locate the transverse CT slice corresponding most closely to the histology sample’s extent. Errors were assessed using a leave-one-out strategy .
Results: The accuracy for registration of in-vivo to ex-vivo CTs was 2.77±0.08mm, and for registration of histology to ex-vivo CT was 1.69±0.07mm. The total registration error between PET and histology for ten histology samples was 6.33±0.22mm.
Conclusion: We have developed a semi-automated registration method to align PET and histology images with a registration accuracy of 6.33mm which is comparable to the PET spatial resolution.
|Title of host publication||National Cancer Research Institute (NCRI) Cancer Conference|
|Place of Publication||Liverpool|
|Publication status||Published - 5 Nov 2013|
- Image registration
- Tracer validation
- Radiotherapy planning