Lung cancer is the most deadly cancer in the United States. Radiation therapy uses ionizing radiation with high energy to destroy lung tumor cells by damaging their genetic material, preventing those cells from reproducing. The most challenging aspect of modern radiation therapy for lung cancer is the motion of lung tumors caused by patient breathing during treatment. Most gating based radiotherapy derives the tumor motion from external surrogates and generates a respiratory signal to trigger the beam. We propose a method that monitors internal diaphragm motion, which can provide a respiratory signal that is more highly correlated to lung tumor motion compared to the external surrogates. We also investigate direct tracking of the tumor in fluoroscopic video imagery. We tracked fixed tumor contours in fluoroscopic videos for 5 patients. The predominant tumor displacements are well tracked based on optical flow. Some tumors or nearby anatomy features exhibit severe nonrigid deformation, especially in the supradiaphragmatic region. By combining Active Shape Models and the respiratory signal, the deformed contours are tracked within a range defined in the training period. All the tracking results are validated by a human expert and the proposed methods are promising for applications in radiotherapy. Another important aspect of lung patient treatment is patient setup verification, which is needed to reduce inter- and intra-fractions geometry uncertainties and ensure precise dose delivery. Currently, there is no universally accepted method for lung patient verification. We propose to register 4DCT and 2D x-ray images taken before treatment to derive the couch shifts necessary for precise radiotherapy. The proposed technique leads to improved patient care.

The complex dose distributions delivered by modern, conformal radiation therapy techniques present a considerable challenge in dose verification. Traditional measurement tools are difficult and laborious to use, since complete verification requires that the doses be determined in three dimensions (3D). The difficulty is further complicated by a required target accuracy of ± 5% for the dose delivery. Gel dosimetry is an attractive option for realizing a tissue-equivalent, 3D dose verification tool with high resolution readout capabilities. However, much important work remains to be completed prior to its acceptance in the clinic. The careful development of easily accessible, fast optical readout tools such as cone beam optical computed tomography (CT) in combination with stable and reliable low-toxicity gel dosimeters is one key step in this process. In this thesis, the performance capabilities and limitations of the two main classes of cone beam optical CT-based absorbing and scattering gel dosimetry are characterized, and their measurement improved through careful matching of dosimeter and scanner performance. These systems are then applied to the evaluation of clinically relevant complex dose distributions. Three-dimensional quality assurance assessments of complex treatment plan dose distributions are shown to be feasible using an optically absorbing Fricke-xylenol-orange-gelatin-based gel dosimeter. Better than 95% voxel agreement is achieved between the plan and the delivery, using 3% dose difference and 3 mm spatial distance-to-agreement gamma function comparison criteria. Small field dose delivery evaluations are demonstrated to be viable using an optically scattering N-isopropylacrylamide (NIPAM)-based polymer gel, with the same comparison criteria. Full treatment process quality assurance is also possible using a NIPAM dosimeter in-phantom, but is limited in its accuracy due to the inherent difficulty of managing the effects of stray light pertubation in the optical attenuation-to-dose calibration. / Thesis (Ph.D, Physics, Engineering Physics and Astronomy) -- Queen's University, 2010-09-02 15:01:48.501

Background: We evaluated the validity of a set of 11 quality indicators for prostate cancer radiotherapy and radical prostatectomy by examining their association with outcomes. The selected indicators were: hospital volume, pre-treatment risk assessment, patient consultation with a radiation oncologist, appropriate follow-up care, leg immobilization during radiotherapy, bladder filling during radiotherapy, portal film target localization, use of nerve sparing surgery, operative blood loss, margin status and pelvic lymph node dissection. The selected outcomes were: cause-specific survival, disease-free survival, late morbidity (urinary incontinence, gastrointestinal and genitourinary morbidity), change in node stage from clinical N0 to pathologic N1, and margin status. Methods: Our study sample consisted of 1570 prostate cancer patients who were diagnosed in Ontario between January 1, 1990 and December 31, 1998 who received radical prostatectomy within 6 months of diagnosis (n=646), or curative radiotherapy within 9 months of diagnosis (n=924). Quality of care, outcomes, and potential confounders were measured using patient chart and administrative data. Regression techniques were used to evaluate the associations between quality indicators and relevant outcomes. Results: For patients treated surgically, hospital volume met our test of validity. Patients treated in the lowest volume hospital (0-1 RP/month) were at greater risk of prostate cancer death than patients treated in the highest volume hospitals (7+ RP/month) (HR=5.37 95% CI=1.23-23.46). For patients treated with radiotherapy, leg immobilization and bladder filling during radiotherapy met our test of validity. Patients treated without leg immobilization were more likely to experience urinary incontinence (RR=2.18, 95% CI=1.23-3.87) and genitourinary late morbidities (RR=1.72, 95% CI=1.16-2.56) than patients who received leg immobilization. Patients who were treated with an empty bladder were more likely to experience GU late morbidities (RR=1.98, 95% CI=1.08-3.63) than those treated with a full bladder. The remaining indicators did not meet our test of validity. Conclusion: Our results support the validity of one surgical quality indicator and two radiotherapy quality indicators. Explanations for our non-significant findings, including limited study power, data quality, our definition and measurement of indicators, and a true failure to predict outcome(s) are discussed, and recommendations for further research are presented. / Thesis (Master, Community Health & Epidemiology) -- Queen's University, 2011-09-07 20:26:34.461

We explore both robust biologically guided intensity-modulated radiation therapy (BG-IMRT) and pattern recognition to identify responders to cancer treatment for lung cancer. Heterogeneous dose prescriptions that are derived from biological images are subject to uncertainty, due to potential noise in the image. We develop a robust optimization model to design BG-IMRT plans that are de-sensitized to uncertainty. Computational results show improvements in tumor control probability and deviation from prescription dose compared to a non-robust model, while maintaining tissue dose below toxicity levels. We applied machine learning algorithms to 4D gated positron emission tomography/computed tomography (PET/CT) scans. We identified classifiers which could outperform a naive classifier. Our work shows the potential of using machine learning algorithms to predict patient response. This could hopefully lead to more adaptive treatment plans, where the clinician would adapt the treatment based on the prediction provided at certain time intervals in the treatment.

We explore both robust biologically guided intensity-modulated radiation therapy (BG-IMRT) and pattern recognition to identify responders to cancer treatment for lung cancer. Heterogeneous dose prescriptions that are derived from biological images are subject to uncertainty, due to potential noise in the image. We develop a robust optimization model to design BG-IMRT plans that are de-sensitized to uncertainty. Computational results show improvements in tumor control probability and deviation from prescription dose compared to a non-robust model, while maintaining tissue dose below toxicity levels. We applied machine learning algorithms to 4D gated positron emission tomography/computed tomography (PET/CT) scans. We identified classifiers which could outperform a naive classifier. Our work shows the potential of using machine learning algorithms to predict patient response. This could hopefully lead to more adaptive treatment plans, where the clinician would adapt the treatment based on the prediction provided at certain time intervals in the treatment.

Adaptive radiation therapy re-optimizes treatment plans based on updated tumor geometries from magnetic resonance imaging scans. However, the imaging process is costly in labor and equipment. In this study, we develop a mathematical model that describes tumor evolution based on a Markov assumption. We then extend the model to predict tumor evolution with any level of information from a new patient: weekly MRI scans are used to estimate transition probabilities when available, otherwise historical MRI scans are used. In the latter case, patients in the historical data are clustered into two groups, and the model relates the new patient's behavior to the existing two groups. The models are evaluated with 33 cervical cancer patients from Princess Margaret Cancer Centre. The result indicates that our models outperform the constant volume model, which replicates the current clinical practice.

Adaptive radiation therapy re-optimizes treatment plans based on updated tumor geometries from magnetic resonance imaging scans. However, the imaging process is costly in labor and equipment. In this study, we develop a mathematical model that describes tumor evolution based on a Markov assumption. We then extend the model to predict tumor evolution with any level of information from a new patient: weekly MRI scans are used to estimate transition probabilities when available, otherwise historical MRI scans are used. In the latter case, patients in the historical data are clustered into two groups, and the model relates the new patient's behavior to the existing two groups. The models are evaluated with 33 cervical cancer patients from Princess Margaret Cancer Centre. The result indicates that our models outperform the constant volume model, which replicates the current clinical practice.

No description available.

Several scientific and technological advancements in radiation oncology have resulted in dramatic improvements in dose conformity and delivery to the target volumes using external beam radiation therapy (EBRT). However, radiation therapy acts as a double-edged sword leading to drastic side-effects, one of them being secondary malignant neoplasms in cancer survivors. The latency time for the occurrence of second cancers is around $10$-$20$ years. Therefore, it is very important to evaluate the risks associated with various types of clinically relevant radiation treatment protocols, to minimize the second cancer risks to critical structures without impairing treatment to the primary tumor volume. A widely used biologically motivated model (known as the initiation-inactivation-proliferation model) with heterogeneous dose volume distributions of Hodgkin's lymphoma survivors is used to evaluate the excess relative risks (ERR). There has been a paradigm shift in radiation therapy from purely photon therapy to other particle therapies in cancer treatments. The extension of the model to include the dependence of linear energy transfer (LET) on the radio-biological parameters and mutation rate for charged particle therapy is discussed. Due to the increase in the use of combined modality regimens to treat several cancers, it is extremely important to evaluate the second cancer risks associated with these anti-cancer therapies. The extension of the model to include chemotherapy induced effects is also discussed. There have been several clinical studies on early and late relapses of cancerous tumors. A tumor control probability (TCP) model with recurrence dynamics in conjunction with the second cancer model is developed in order to enable design of efficient radiation regimens to increase the tumor control probability and relapse time, and at the same time decrease secondary cancer risks. Evolutionary dynamics has played an important role in modeling cancer progression of primary cancers. Spatial models of evolutionary dynamics are considered to be more appropriate to understand cancer progression for obvious reasons. In this context, a spatial evolutionary framework on lattices and unstructured meshes is developed to investigate the effect of cellular motility on the fixation probability. In the later part of this work, this model is extended to incorporate random fitness distributions into the lattices to explore the dynamics of invasion probability in the presence and absence of migration.

This work describes the development and application of a novel combination of single-cell Raman spectroscopy (RS), automated data processing, and principal component analysis (PCA) for investigating radiation induced biochemical responses in human tumour cells. The developed techniques are first validated for the analysis of large data sets (~200 spectra) obtained from single cells. The effectiveness and robustness of the automated data processing methods is demonstrated, and potential pitfalls that may arise during the implementation of such methods are identified. The techniques are first applied to investigate the inherent sources of spectral variability between single cells of a human prostate tumour cell line (DU145) cultured {\it in vitro}. PCA is used to identify spectral differences that correlate with cell cycle progression and the changing confluency of a cell culture during the first 3-4 days after sub-culturing. Spectral variability arising from cell cycle progression is (i) expressed as varying intensities of protein and nucleic acid features relative to lipid features, (ii) well correlated with known biochemical changes in cells as they progress through the cell cycle, and (iii) shown to be the most significant source of inherent spectral variability between cells. This characterization provides a foundation for interpreting spectral variability in subsequent studies. The techniques are then applied to study the effects of ionizing radiation on human tumour cells. DU145 cells are cultured in vitro and irradiated to doses between 15 and 50 Gy with single fractions of 6 MV photons from a medical linear accelerator. Raman spectra are acquired from irradiated and unirradiated cells, up to 5 days post-irradiation. PCA is used to distinguish radiation induced spectral changes from inherent sources of spectral variability, such as those arising from cell cycle. Radiation induced spectral changes are found to correlate with both the irradiated dose and the incubation time post-irradiation, and to arise from biochemical differences in lipids, nucleic acids, amino acids, and conformational protein structures between irradiated and unirradiated cells. This study is the first use of RS to observe radiation induced biochemical effects in single cells, and is the first use of vibrational spectroscopy to observe such effects independent from cell cycle or cell death related processes. The same methods are then applied to a panel of human tumour cell lines, derived from prostate (DU145, PC3, LNCaP and PacMet), breast (MDA-MB-231 and MCF7) and lung (H460), which vary by p53 gene status and intrinsic radiosensitivity. One radiation induced PCA component is detected for each cell line by statistically significant changes in the PCA score distributions for irradiated samples, as compared to unirradiated samples, in the first 24 to 72 hours post-irradiation. These RS response signatures arise from radiation induced changes in cellular concentrations of aromatic amino acids, conformational protein structures, and certain nucleic acid and lipid functional groups. Correlation analysis between the radiation induced PCA components separates the cell lines into three unique RS response categories: R1 (H460, MCF7 and PacMet), R2 (MDA-MB-231 and PC3), and R3 (DU145 and LNCaP). These RS categories partially segregate according to radiosensitivity; the R1 and R2 cell lines are radioresistant and the R3 cell lines are radiosensitive (PacMet radiosensitivity (R1) unknown). The R1 and R2 cell lines further segregate according to p53 gene status, corroborated by cell cycle analysis post-irradiation. Preliminary results obtained from a mouse prostate tumour cell line (TRAMP-C2), irradiated both in vitro and in vivo, indicate that RS signatures of radiation response may also be detectable from tumour cells obtained from an in vivo system during radiation therapy treatment. These results indicate the potential for future RS studies designed to investigate, monitor, or predict radiation response. / Graduate