This work was funded by the Ontario Research Fund for the Ontario Consortium for Adaptive Interventions in Radiation Oncology (OCAIRO).
References
[1] Patrick Y Wen and Santosh Kesari. Malignant gliomas in adults. The New England journal of medicine, 359(5):492–507, July 2008.
[2] Craig J Galb´an, Thomas L Chenevert, Charles R Meyer, Christina Tsien, Theodore S Lawrence, Daniel A Hamstra, Larry Junck, Pia C Sundgren, Tim- othy D Johnson, David J Ross, Alnawaz Rehemtulla, and Brian D Ross. The parametric response map is an imaging biomarker for early cancer treatment outcome. Nature medicine, 15(5):572–6, May 2009.
[3] Robert J Harris, Timothy F Cloughesy, Whitney B Pope, Phioanh L Nghiemphu, Albert Lai, Taryar Zaw, Johannes Czernin, Michael E Phelps, Wei Chen, and Benjamin M Ellingson. 18F-FDOPA and 18F-FLT positron emission tomogra- phy parametric response maps predict response in recurrent malignant gliomas treated with bevacizumab. Neuro-oncology, 14(8):1079–89, August 2012.
[4] Craig J Galb´an, Suresh K Mukherji, Thomas L Chenevert, Charles R Meyer, Daniel A Hamstra, Peyton H Bland, Timothy D Johnson, Bradford A Moffat, Alnawaz Rehemtulla, Avraham Eisbruch, and Brian D Ross. A feasibility study of parametric response map analysis of diffusion-weighted magnetic resonance imaging scans of head and neck cancer patients for providing early detection of therapeutic efficacy. Translational oncology, 2(3):184–90, August 2009.
[5] Seung Joon Choi, Jonghoon Kim, Jongbum Seo, Hyung Sik Kim, Jong-min Lee, and Hyunjin Park. Parametric response mapping of dynamic CT as an imaging biomarker to distinguish viability of hepatocellular carcinoma treated with transcatheter arterial chemoembolization. Abdominal imaging, 39(3):518– 25, June 2014.
[6] Craig J Galb´an, Meilan K Han, Jennifer L Boes, Komal A Chughtai, Charles R Meyer, Timothy D Johnson, Stefanie Galb´an, Alnawaz Rehemtulla, Ella A Kaze- rooni, Fernando J Martinez, and Brian D Ross. Computed tomography-based biomarker provides unique signature for diagnosis of COPD phenotypes and dis- ease progression. Nature medicine, 18(11):1711–5, November 2012.
[7] Craig J. Galb´an, Bing Ma, Dariya Malyarenko, Martin D. Pickles, Kevin Heist, Norah L. Henry, Anne F. Schott, Colleen H. Neal, Nola M. Hylton, Alnawaz Rehemtulla, Timothy D. Johnson, Charles R. Meyer, Thomas L. Chenevert, Lindsay W. Turnbull, and Brian D. Ross. Multi-Site Clinical Evaluation of DW- MRI as a Treatment Response Metric for Breast Cancer Patients Undergoing Neoadjuvant Chemotherapy. PLOS ONE, 10(3):e0122151, March 2015.
[8] C. J. Galban, T. L. Chenevert, C. R. Meyer, C. Tsien, T. S. Lawrence, D. A. Hamstra, L. Junck, P. C. Sundgren, T. D. Johnson, S. Galban, J. S. Sebolt- Leopold, A. Rehemtulla, and B. D. Ross. Prospective Analysis of Parametric
Response Map-Derived MRI Biomarkers: Identification of Early and Distinct Glioma Response Patterns Not Predicted by Standard Radiographic Assessment.
Clinical Cancer Research, 17(14):4751–4760, April 2011.
[9] N. Zha, D. P. I. Capaldi, D. Pike, D. G. McCormack, I. A. Cunningham, and G. Parraga. Principal component analysis of the CT density histogram to gener- ate parametric response maps of COPD. In Barjor Gimi and Robert C. Molthen, editors,SPIE Medical Imaging, page 941716. International Society for Optics and Photonics, March 2015.
[10] Jolliffe I. T. Principal Component Analysis. Springer-Verlag, New York, 2002. [11] J L Boxerman, K M Schmainda, and R M Weisskoff. Relative cerebral blood
volume maps corrected for contrast agent extravasation significantly correlate with glioma tumor grade, whereas uncorrected maps do not. AJNR. American journal of neuroradiology, 27(4):859–67, April 2006.
[12] Kikinis R. Pieper S., Halle M. 3d slicer. InProceedings of the 1st IEEE Interna- tional symposium on biomedical imaging: From Nano to Macro, pages 632–635, 2004.
[13] Rajan Jain, Laila Poisson, Jayant Narang, David Gutman, Lisa Scarpace, Scott N Hwang, Chad Holder, Max Wintermark, Rivka R Colen, Justin Kirby, John Freymann, Daniel J Brat, Carl Jaffe, and Tom Mikkelsen. Genomic mapping and survival prediction in glioblastoma: molecular subclassification strengthened by hemodynamic imaging biomarkers. Radiology, 267(1):212–20, April 2013.
[14] Anthony Lausch, Jeff Chen, Aaron D Ward, Stewart Gaede, Ting-Yim Lee, and Eugene Wong. An augmented parametric response map with consideration of image registration error: towards guidance of locally adaptive radiotherapy.
Physics in medicine and biology, 59(22):7039–58, November 2014.
[15] Jennifer L Boes, Benjamin A Hoff, Nola Hylton, Martin D Pickles, Lindsay W Turnbull, Anne F Schott, Alnawaz Rehemtulla, Ryan Chamberlain, Benjamin Lemasson, Thomas L Chenevert, Craig J Galb´an, Charles R Meyer, and Brian D Ross. Image registration for quantitative parametric response mapping of cancer treatment response. Translational oncology, 7(1):101–10, February 2014.
[16] Fang Han and Han Liu. Scale-Invariant Sparse PCA on High Dimensional Meta- elliptical Data. Journal of the American Statistical Association, 109(505):275– 287, January 2014.
Chapter 5
Conclusion and Future Work
5.1
Summary of findings
5.1.1
Radiation dose-response relationships for liver tumours
Radiation-dose response relationships define the probability of tumour control as a function of prescribed radiation dose. An improved understanding of these relation- ships could help to adapt dose prescriptions to patients (i.e. personalize) in order to improve tumour control or reduce side-effects. Chapter 2 contributed to this body of knowledge by defining dose-response relationships for patients treated with SBRT for hepatocellular carcinoma (HCC) and colorectal liver metastases (MET). To our knowledge, this study was the first to explicitly model dose-response relationships for HCC patients and the second to do so for MET patients [1]. 50% and 90% probabil- ities of 6-month local control were estimated to be achievable by 2 Gy per fraction equivalent doses (α/β = 10Gy) of 53 Gy and 84 Gy for the HCC group and 70 Gy and 95 Gy for the MET group, respectively. Results for the MET group also helped to support a previously reported MET dose-response relationship [1]. Overall, we found that a higher radiation dose was required to control MET tumours when com- pared to HCC and that RT provided improved tumour control for HCC patients whencompared to MET patients at our institution.
5.1.2
Augmented parametric response mapping
Guidance of next generation locally adaptive radiotherapy techniques such as sub- volume boosting requires image-based treatment response prediction. The PRM is an image-based method for prediction of overall treatment outcome (e.g. overall survival) which shows promise as a tool for guiding personalized locally adaptive radiotherapy (RT). However, image registration error (IRE) introduces uncertainty into this voxel- wise analysis technique which may limit its use for guiding RT. Chapter 3 proposed an augmented PRM method (A-PRM) to address this challenge. The original PRM method was extended to include an IRE-related PRM analysis confidence interval and also incorporated multiple graded classification thresholds to facilitate visualization. PRM and A-PRM analyses of CT-perfusion functional images with known simulated IRE were compared to analysis without simulated IRE to investigate the two methods in the presence of controlled IRE. The A-PRM was shown to help visualize and quantify IRE-related analysis uncertainty. The use of multiple graded classification thresholds also provided additional contextual information which could be useful for visually identifying adaptive RT targets (e.g. sub-volume boost regions). The A-PRM should facilitate reliable PRM guided adaptive RT by allowing the user to identify if a patient’s unique IRE-related PRM analysis uncertainty has the potential to influence target delineation.
5.1.3
Multi-parametric response mapping
Voxel-wise analysis of functional imaging acquired at two time points using the PRM has been shown to be an effective tool for early prediction of cancer treatment out- comes (e.g. overall survival) and may also be well-suited towards guiding personalized locally adaptive RT. However, the PRM method has been designed for and almost ex- clusively applied to analysis of longitudinally acquired pairs of single-parameter image data. Chapter 4proposed a novel approach towards multi-parametric response map- ping (MPRM) to address this challenge. The overall objective was to improve global treatment response prediction (e.g. overall survival) and facilitate future investiga- tions into voxel-wise response prediction for guidance of locally adaptive RT. MPRM analysis was applied to a multi-parametric dataset acquired from a group of n = 19 patients treated for high-grade glioma with comparisons to original single-parameter PRM analysis. Separate PRM and MPRM analyses of the contrast-enhancing lesion (CEL) and 1 cm of peripheral tissue (PERIPH) were performed. The original single- parameter PRM was found to be significantly predictive of median overall survival only when applied to rCBV data within the PERIPH ROI (AUCP ERIP H = 0.78, p<
0.05). The MPRM was found to be significantly predictive of median overall survival for both CEL and PERIPH analyses and offered improved prediction (AUCCEL =
0.82, AUCP ERIP H = 0.84, p < 0.05) suggesting the benefit of multi-parametric re-
sponse prediction using the MPRM. The significant predictions of OS for PERIPH analyses also supported the idea that functional changes in peritumoral regions can be predictive of treatment response. In summary, the proposed algorithm accounted for spatial heterogeneity in multi-parametric response, supported intuitive visualizations, and was found to improve prediction of overall survival. To mirror the accessibility of the original PRM method, the MPRM algorithm was also made publically available via the MATLAB file exchange.