Abstract
This thesis explores co-registering bone sarcoma histology images with volumetric radiology. The spatial alignment framework presented leverages routine clinical data to close the loop between pre-surgical radiology and post-surgical high-fidelity histology, facilitating improved accuracy of image interpretation and tumour resection.
Bone sarcoma has poor treatment outcomes, burdening patients and healthcare systems. Surgery, the primary curative approach, relies on radiology and commonly removes a healthy tissue margin to account for planning and treatment inaccuracies. The discrepancy between imaging-defined and histologic tumour boundaries contributes to this inaccuracies. Quantifying this difference may allow margins to be safely narrowed, preserving healthy critical structures such as nerves and blood vessels.
This thesis used a decision matrix to assess clinical viability of potential co-registration approaches, ranking ease of clinical use, histopathology processing, surgical simplicity, and scanning efficiency. The bisected specimen approach emerged as most suitable, involving ex-vivo resected tumour bisection and CT scanning. The bisection reference surface can be identified on CT and used for three-dimensional histology alignment.
A co-registration framework using the bisected specimen approach was developed and tested using canine limbs. Mean histology plane orientation error between computed and physical dissection landmarks was 0.19 ± 1.8mm standard deviation (SD) across 114 measurements from 3 specimens. Further framework development compared three geometric transformation methods for co-registration with challenging, curved bisection surfaces. Mean error for the best-performing method was 0.06 ± 1.87mm SD across 114 measurements from 4 specimens. Further clinical testing is needed to assess framework robustness for a range of bone sarcoma presentations, and to refine workflows for clinical adoption.
This co-registration framework would enable a comprehensive database of spatially orientated histology and radiology data. This could support bone sarcoma treatment by enabling more accurate image interpretation for safer narrow-margin decisions, patient-specific planning, preservation of critical structures during tumour resection, and prediction of surgical outcome.