The development of novel treatment techniques, including radiation, minimally invasive surgery or targeted agents, requires an accurate assessment of the treatment results for evidence-based medicine, where through randomized studies a cause and effect relationship is determined and the most effective treatment can be identified. Evidence based medicine requires precise knowledge of what was delivered in the treatment and what the corresponding outcome was. This correlation is only possible through precise modeling of the patient over the course of treatment, reducing the uncertainties in treatment delivery by accounting for the changes in the patient and the discrepancies between the intended delivery and the actual delivery of the treatment. The patient modeling can then be extended to monitor the response of the patient by correlating an accurate representation of the delivered treatment to the tumor and normal tissue response following the treatment.
Deformable modeling will improve the precision in evidence-based medicine by allowing evaluation of the intended treatment versus outcome to be replaced with the actual delivered treatment. Ongoing research in my lab is focusing on updating the initial treatment intent to include:
i) geometric uncertainties in the delivery of the local cancer treatment (e.g. anatomical changes due to response of anatomy due to therapy in the patient during the delivery of radiation therapy, collaborations with David Fuller, M.D., Ph.D. in Radiation Oncology; accounting for uncertainties in the integration of imaging information for surgical guidance in neurosurgery, collaborations with Ken-Pin Hwang, Ph.D. and Ho-Ling Anthony Liu, Ph.D. in Imaging Physics and Jeffrey Weinberg, M.D. in Neurosurgery)
ii) correlation of geometric response of the tumor and normal tissue, through deformable modeling of the planning and follow-up images, collaborations with Eugene Koay, M.D., Ph.D. and Sunil Krishnan, M.D. in Radiation Oncology.
Image-Guided Neurosurgery (functional mapping, resection guidance, correlative pathology)
Innovation and Significance: Functional imaging provides essential information to guide neurosurgery, however once the skull is opened and surgery begins the correlation of the pre-treatment imaging with the intra-operative imaging. The use of anatomical modeling to account for this shift and accurately map the functional imaging onto the intra-operative imaging may significantly improve the efficacy of the surgery and reduce the morbidity. In addition, the ability to update the models with imaging information (including limited information such as 2D images or partial volume images) obtained throughout the surgical process can provide additional information to the surgeon on residual tumor and the position of the critical normal tissues. This research will be performed in collaboration with Jeffrey Weinberg, M.D. in Neurosurgery.
Innovation and Significance: The integration of functional imaging and anatomical models will be applied to image guided lung resection to investigate the benefits in reduced morbidity. Anatomical modeling of lung function, including correlations between ventilation, perfusion, and dynamic images will be performed to identify areas of high-functioning lung. Localized intra-operative pathology techniques combined with imaging will also be investigated. In addition, models will be developed to map the tumor from the inflated to deflated (at the time of surgical resection) lung. This research will be performed in collaboration with Dave Rice, M.B., B.Ch. in Thoracic and Cardiovascular Surgery.
Advances in image-guided therapies promises precision targeting at the time of treatment, however, defining where to place this dose remains a large source of uncertainty. Questions remain such as which imaging modality or combination of imaging modalities best represent the tumor, what is the extent of microscopic disease that is not visualized on the images, and what new imaging modalities and probes can be used to improve this visualization. Answering these questions requires accurate integration of these images into one model of the patient as well as validation of the results of these images with a 'gold standard' of histo-pathology assessment. However, the geometric and spatial discrepancies between histo-pathology and in-vivo imaging compromise the validation. My research has focused on developing deformable registration techniques to provide correlation between in-vivo images and histo-pathology for validation of tumor definition. These novel techniques, employing biomechanical model-based techniques, may overcome some of the limitations of standard techniques. They have also been shown to translate from clinical to pre-clinical studies, we are establishing collaborations with Jim Bankson, Ph.D. and the Small Animal Imaging Facility at MD Anderson. This research will be enhanced by the unique resources at MD Anderson through collaborations with Ignacio Wistuba, M.D. in Translational Molecular Pathology and Sunil Krishnan, M.D. in the Center for Radiation Oncology Research and has the potential to impact all solid tumors treated with localized therapy.
Assessment of tumor response remains a limiting factor in evaluation of novel treatment techniques, which is critical for evidence-based medicine. Quantitative assessment of therapeutic response through anatomical, functional, and metabolic imaging has been identified by the NIH has a critical component in the advancement of local therapies. The testing, validation, and translation of image-based assessments have been hampered by the inability to accurately link the serially acquired images indicating response over time with the therapy that was delivered. My research has shown that biomechanical model-based deformable registration can map anatomy between multi-modality images obtained over a relatively short time frame, accounting for anatomical motion. However, the complex, longer-term changes in shape and volume due to non-mechanical forces such as differential response of normal tissues to irradiation are not addressed by force-based mechanics applied to anatomical models to guide deformation. In addition, these complex changes challenge the ability to combine local therapy previously delivered with that intended to be delivered, such as in the chronic management of liver cancer. These uncertainties related to complex changes often result in conservative management of patient treatments, potentially reducing the effective management of the cancer, and increased uncertainties in the correlation of treatment with assessment of outcome. The development of biomechanical models, currently under investigation in my lab, to describe the volumetric and spatial response of an organ to a local therapy will enable the improvement of correlating novel imaging, currently being developed by John Hazle, Ph.D., Osama Mawlawi, Ph.D. and Jason Stafford, Ph.D. in Imaging Physics, with local therapy in radiation oncology, interventional radiology, and image-guided surgery.
Response modeling (Target Sites: head and neck and liver)
Innovation and Significance: The novelty of the proposed research is the ability to generate a biomechanical model of anatomical tissue, specifically the parotid gland and liver for this initial project, which describes the volumetric response of the normal tissue to RT using a modified thermal-expansion coefficient, which incorporates delivered dose and volumetric response. Using MR sialography (parotid gland) and contrast enhanced MR and/or CT (liver) along with MR elastography in both anatomical sites, the biomechanical model of the volumetric response will be developed and quantitatively validated. This will be an important new advancement in the field, which will enable an understanding of the uncertainties associated with modeling normal response and its correlation with delivered dose. This research will require a prospective imaging study for patients undergoing radiation therapy treatment for oropharyngeal cancer and primary or metastatic liver cancer.