One major objective in radiation oncology is the personalisation of cancer treatment. The implementation of this concept requires the identification of biomarkers, which precisely predict therapy outcome. Besides molecular characterisation of tumours, a new approach known as radiomics aims to characterise tumours using imaging data. In the context of the presented thesis, radiomics was established at OncoRay to improve the performance of imaging-based risk models. Two software-based frameworks were developed for image feature computation and risk model construction. A novel data-driven approach for the correction of intensity non-uniformity in magnetic resonance imaging data was evolved to improve image quality prior to feature computation. Further, different feature selection methods and machine learning algorithms for time-to-event survival data were evaluated to identify suitable algorithms for radiomics risk modelling. An improved model performance could be demonstrated using computed tomography data, which were acquired during the course of treatment. Subsequently tumour sub-volumes were analysed and it was shown that the tumour rim contains the most relevant prognostic information compared to the corresponding core. The incorporation of such spatial diversity information is a promising way to improve the performance of risk models.:1. Introduction 2. Theoretical background 2.1. Basic physical principles of image modalities 2.1.1. Computed tomography 2.1.2. Magnetic resonance imaging 2.2. Basic principles of survival analyses 2.2.1. Semi-parametric survival models 2.2.2. Full-parametric survival models 2.3. Radiomics risk modelling 2.3.1. Feature computation framework 2.3.2. Risk modelling framework 2.4. Performance assessments 2.5. Feature selection methods and machine learning algorithms 2.5.1. Feature selection methods 2.5.2. Machine learning algorithms 3. A physical correction model for automatic correction of intensity non-uniformity in magnetic resonance imaging 3.1. Intensity non-uniformity correction methods 3.2. Physical correction model 3.2.1. Correction strategy and model definition 3.2.2. Model parameter constraints 3.3. Experiments 3.3.1. Phantom and simulated brain data set 3.3.2. Clinical brain data set 3.3.3. Abdominal data set 3.4. Summary and discussion 4. Comparison of feature selection methods and machine learning algorithms for radiomics time-to-event survival models 4.1. Motivation 4.2. Patient cohort and experimental design 4.2.1. Characteristics of patient cohort 4.2.2. Experimental design 4.3. Results of feature selection methods and machine learning algorithms evaluation 4.4. Summary and discussion 5. Characterisation of tumour phenotype using computed tomography imaging during treatment 5.1. Motivation 5.2. Patient cohort and experimental design 5.2.1. Characteristics of patient cohort 5.2.2. Experimental design 5.3. Results of computed tomography imaging during treatment 5.4. Summary and discussion 6. Tumour phenotype characterisation using tumour sub-volumes 6.1. Motivation 6.2. Patient cohort and experimental design 6.2.1. Characteristics of patient cohorts 6.2.2. Experimental design 6.3. Results of tumour sub-volumes evaluation 6.4. Summary and discussion 7. Summary and further perspectives 8. Zusammenfassung

Personalized treatment strategies for oncological patient management can improve outcomes of patient populations with heterogeneous treatment response. The implementation of such a concept requires the identification of biomarkers that can precisely predict treatment outcome. In the context of this thesis, we develop and validate biomarkers from multimodal imaging data for the outcome prediction after treatment in patients with locally advanced rectal cancer (LARC) and in patients with newly diagnosed glioblastoma multiforme (GBM), using conventional feature-based radiomics and deep-learning (DL) based radiomics. For LARC patients, we identify promising radiomics signatures combining computed tomography (CT) and T2-weighted (T2-w) magnetic resonance imaging (MRI) with clinical parameters to predict tumour response to neoadjuvant chemoradiotherapy (nCRT). Further, the analyses of externally available radiomics models for LARC reveal a lack of reproducibility and the need for standardization of the radiomics process. For patients with GBM, we use postoperative [11C] methionine positron emission tomography (MET-PET) and gadolinium-enhanced T1-w MRI for the detection of the residual tumour status and to prognosticate time-to-recurrence (TTR) and overall survival (OS). We show that DL models built on MET-PET have an improved diagnostic and prognostic value as compared to MRI.

BACKGROUND: Pancreatic cancer is one of the most lethal cancers. Despite enhanced understanding of the disease, the 5-year survival rate remains 8% due to the late stage of diagnosis and a lack of effective treatment options. Early detection of precancerous lesions, such as intraductal papillary mucinous neoplasms (IPMNs), is a strategy to prevent pancreas cancer related death. Standard qualitative imaging assessment cannot reliably distinguish between benign and malignant branch duct intraductal papillary mucinous neoplasms (BD-IPMNs). A more consistent risk prediction method is needed to inform clinical decision making such that patients with benign cysts may be spared from unnecessary surgical resection. OBJECTIVE: To assess whether a BD-IPMN malignancy risk prediction model which demonstrated strong potential on preoperative computed tomography (CT) images would show similar results on magnetic resonance imaging (MRI). METHODS: 19 pathologically proven BD-IPMN patients with preoperative contrast-enhanced CT and MRI and were included in the study. Five radiomics features were extracted from the portal-venous phase CT and MR images of the largest cyst. Associations between radiomics features extracted from CT and MR were assessed using Pearson correlations. RESULTS: Of the five radiomics features, average-weighted eccentricity (AWE) was most strongly correlated between imaging modalities in all patients (n=19, r=0.46, 95% CI=0.001-0.75, p=0.05), low-risk patients (r=0.63, 95% CI=0.09-0.88, p=0.028), and patients with a solid component or mural nodule (r=0.66, 95% CI=-0.32-0.96, p=0.15). However, when two outliers within the dataset were removed from analysis, AWE no longer correlated between MR and CT. None of the other radiomics features displayed significant correlations between the modalities. CONCLUSIONS: The CT-based risk prediction model cannot be applied to MR data suggesting that a new model should be created from MRI data alone. / 2021-06-17T00:00:00Z

L’imagerie est un outil pour réaliser le diagnostic, le bilan d’extension et le suivi thérapeutique de la grande majorité des tumeurs. La tomodensitométrie (TDM) est la méthode la plus utilisée et les images obtenues fournissent une cartographie tumorale fondée sur la densité des tissus. L’analyse plus approfondie de ces images acquises en routine clinique a permis d’extraire des informations supplémentaires quant à la survie du patient ou à la réponse au(x) traitement(s). Toutes ces nouvelles données permettent de décrire le phénotype d’une lésion de façon non invasive et sont regroupées sous le terme de radiomique. La plupart des études de radiomique se sont focalisées sur les paramètres de texture et ont évalué les données acquises à l’aide de TDM avec injection de produit de contraste (phase portale). Pour ces travaux de thèse, nous avons réalisé une analyse des paramètres de radiomique extraits à la fois des images TDM contrastées et non contrastées des tumeurs colorectales. La construction d’un modèle pronostique à l’aide de ces paramètres a permis d’étudier la complémentarité des informations fournies par les deux modalités. Dans un second temps, l’analyse des modifications transcriptomiques des cellules souches et cellules cancéreuses dans le cancer colorectal a permis de valider l’hypothèse que la quantification de modifications transcriptomiques peut également avoir une valeur pronostique. Finalement, l’étude des corrélations entre les données d’expression génétique et la radiomique en TDM a montré que la quantification de l’hétérogénéité tumorale en TDM reflète en partie les modifications transcriptomiques. / Imaging is the principal tool for diagnosis, extension assessment and therapeutic follow-up of the vast majority of tumors. Computed tomography (CT) is the most used method and provides an assessment of tumor tissue density. In-depth analysis of those images acquired in clinical routine has suppl ied additional data regarding patient survival or treatment response. All those new data allow to describe the tumor phenotype and are generally grouped under the generic term radiomics. Most of previous studies focused on texture analysis using contrast enhanced CT (portal phase). In the first part of this thesis, we carried out a radiomics analysis of both contrast-enhanced and nonenhanced CT images of the colorectal tumors.Imaging is the principal tool for diagnosis, extension assessment and therapeutic followup of the vast majority of tumors. Computed tomography (CT) is the most used method and provides an assessment of tumor tissue density. In-depth analysis of those images acquired in clinical routine has supplied additional data regarding patient survival or treatment response. All those new data allow to describe the tumor phenotype and are generally grouped under the generic term radiomics. Most of previous studies focused on texture analysis using contrast enhanced CT (portal phase). In the first part of this thesis, we carried out a radiomics analysis of both contrast-enhanced and non-enhanced CT images of the colorectal tumors.

ADVANCED IMAGING ANALYSIS FOR PREDICTING TUMOR RESPONSE AND IMPROVING CONTOUR DELINEATION UNCERTAINTY By Rebecca Nichole Mahon, MS A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy at Virginia Commonwealth University. Virginia Commonwealth University, 2018 Major Director: Dr. Elisabeth Weiss, Professor, Department of Radiation Oncology Radiomics, an advanced form of imaging analysis, is a growing field of interest in medicine. Radiomics seeks to extract quantitative information from images through use of computer vision techniques to assist in improving treatment. Early prediction of treatment response is one way of improving overall patient care. This work seeks to explore the feasibility of building predictive models from radiomic texture features extracted from magnetic resonance (MR) and computed tomography (CT) images of lung cancer patients. First, repeatable primary tumor texture features from each imaging modality were identified to ensure a sufficient number of repeatable features existed for model development. Then a workflow was developed to build models to predict overall survival and local control using single modality and multi-modality radiomics features. The workflow was also applied to normal tissue contours as a control study. Multiple significant models were identified for the single modality MR- and CT-based models, while the multi-modality models were promising indicating exploration with a larger cohort is warranted. Another way advances in imaging analysis can be leveraged is in improving accuracy of contours. Unfortunately, the tumor can be close in appearance to normal tissue on medical images creating high uncertainty in the tumor boundary. As the entire defined target is treated, providing physicians with additional information when delineating the target volume can improve the accuracy of the contour and potentially reduce the amount of normal tissue incorporated into the contour. Convolution neural networks were developed and trained to identify the tumor interface with normal tissue and for one network to identify the tumor location. A mock tool was presented using the output of the network to provide the physician with the uncertainty in prediction of the interface type and the probability of the contour delineation uncertainty exceeding 5mm for the top three predictions.

Cervical cancer remains the third most commonly diagnosed gynecological malignancy in the United States and throughout the world despite being potentially preventable. Patients diagnosed with cervical cancer may develop local recurrence in the cervix and surrounding structures (vaginal apex, parametrial, or paracervical), regional recurrence in pelvic lymph nodes, distant metastasis, or a combination of all. The management of such treatment outcomes has not been subject to rigorous investigation. Therefore, there is a need for studies and clinical trials that focus on decision making to support the choice of the best treatment modality that leads to the minimal number of adverse treatment outcomes. Medical imaging plays a vital role in the initial diagnosis, staging, and guiding treatment decisions for cancer patients. Positron Emission Tomography-Computed Tomography (PET/CT) hybrid scanner has proven to be a primary functional imaging modality in the oncology clinic. A typical oncological application of PET/CT aims to examine the whole body for high tracer uptake as a sign of tumorous lesions or metastasis using 18F-Fluoro-2-deoxy-D-glucose (18F-FDG). This radiopharmaceutical has been proven to be useful for the quantitative determination of regional glucose metabolism localized in the brain, heart, bladder, and, fortunately, in tumors. Currently, 18F-FDG measured on PET is the prominent radiotracer in cancer staging and follow-up imaging. In the –omics1 era, mining data to derive inherent information about a system has influenced the medical field, especially oncological imaging. The process of radiomics involves high throughput analysis of medical images to extract a large number of quantified features that are presented as a decision supporting tool for clinicians in terms of various clinical tasks such as staging, prediction, and prognosis. In recent studies, the focus of radiomics has exceeded the whole-tumor analysis to include the quantification of habitats, sub-regions within the tumor volume defined based on specific criteria, with the intent to investigate the diversity extent of the intratumor heterogeneity as robust descriptors and predictors of clinicopathological factors. The presented work is a retrospective analysis of a cohort consisting of pretreatment Positron Emission Tomography and Computed Tomography (PET/CT) hybrid scans of cervical cancer patients consecutively treated with radiochemotherapy. We extracted radiomic features from the primary cervical tumor volumes, and voxel intensity-based features from tumor habitats to analyze the tumors’ heterogeneity based on 18Flourodeoxyglocuse (18F-FDG) uptake of PET, and Hounsfield Units (HU) of CT to obtain useful tumor information, which might be associated with treatment outcomes. To our knowledge, a limited number of studies have focused on investigating the potential role of radiomic features on cervical cancer PET/CT images. Briefly, the workflow of this study consisted of investigating parameters that might affect radiomic features predictive performance by evaluating the reproducibility of radiomic features extracted from 18F-FDG PET images for segmentation methods, gray levels discretization, and PET reconstruction algorithms. Afterward, we used these features to predict cervical treatment outcomes after radiochemotherapy. Due to the use of human data, this research study acquired the approval of the institutional review board (IRB) at the University of South Florida.

Soft Tissue Sarcomas (STS) are among the most dangerous diseases, with a 50% mortality rate in the USA in 2016. Heterogeneous responses to the treatments of the same sub-type of STS as well as intra-tumor heterogeneity make the study of biopsies imprecise. Radiologists make efforts to find non-invasive approaches to gather useful and important information regarding characteristics and behaviors of STS tumors, such as aggressiveness and recurrence. Quantitative image analysis is an approach to integrate information extracted using data science, such as data mining and machine learning with biological an clinical data to assist radiologists in making the best recommendation on clinical trials and the course of treatment. The new methods in “Radiomics" extract meaningful features from medical imaging data for diagnostic and prognostic goals. Furthermore, features extracted from Convolutional Neural Networks (CNNs) are demonstrating very powerful and robust performance in computer aided decision systems (CADs). Also, a well-known computer vision approach, Bag of Visual Words, has recently been applied on imaging data for machine learning purposes such as classification of different types of tumors based on their specific behavior and phenotype. These approaches are not fully and widely investigated in STS. This dissertation provides novel versions of image analysis based on Radiomics and Bag of Visual Words integrated with deep features to quantify the heterogeneity of entire STS as well as sub-regions, which have predictive and prognostic imaging features, from single and multi-sequence Magnetic Resonance Imaging (MRI). STS are types of cancer which are rarely touched in term of quantitative cancer analysis versus other type of cancers such as lung, brain and breast cancers. This dissertation does a comprehensive analysis on available data in 2D and multi-slice to predict the behavior of the STS with regard to clinical outcomes such as recurrence or metastasis and amount of tumor necrosis. The experimental results using Radiomics as well as a new ensemble of Bags of Visual Words framework are promising with 91.66% classification accuracy and 0.91 AUC for metastasis, using ensemble of Bags of Visual Words framework integrated with deep features, and 82.44% classification accuracy with 0.63 AUC for necrosis progression, using Radiomics framework, in tests on the available datasets.

abstract: The rate of progress in improving survival of patients with solid tumors is slow due to late stage diagnosis and poor tumor characterization processes that fail to effectively reflect the nature of tumor before treatment or the subsequent change in its dynamics because of treatment. Further advancement of targeted therapies relies on advancements in biomarker research. In the context of solid tumors, bio-specimen samples such as biopsies serve as the main source of biomarkers used in the treatment and monitoring of cancer, even though biopsy samples are susceptible to sampling error and more importantly, are local and offer a narrow temporal scope. Because of its established role in cancer care and its non-invasive nature imaging offers the potential to complement the findings of cancer biology. Over the past decade, a compelling body of literature has emerged suggesting a more pivotal role for imaging in the diagnosis, prognosis, and monitoring of diseases. These advances have facilitated the rise of an emerging practice known as Radiomics: the extraction and analysis of large numbers of quantitative features from medical images to improve disease characterization and prediction of outcome. It has been suggested that radiomics can contribute to biomarker discovery by detecting imaging traits that are complementary or interchangeable with other markers. This thesis seeks further advancement of imaging biomarker discovery. This research unfolds over two aims: I) developing a comprehensive methodological pipeline for converting diagnostic imaging data into mineable sources of information, and II) investigating the utility of imaging data in clinical diagnostic applications. Four validation studies were conducted using the radiomics pipeline developed in aim I. These studies had the following goals: (1 distinguishing between benign and malignant head and neck lesions (2) differentiating benign and malignant breast cancers, (3) predicting the status of Human Papillomavirus in head and neck cancers, and (4) predicting neuropsychological performances as they relate to Alzheimer’s disease progression. The long-term objective of this thesis is to improve patient outcome and survival by facilitating incorporation of routine care imaging data into decision making processes. / Dissertation/Thesis / Doctoral Dissertation Biomedical Informatics 2017

This thesis deals with a classification of glioma grade in high and low aggressive tumours and overall survival prediction based on magnetic resonance imaging. Data used in this work is from BRATS challenge 2019 and each set contains information from 4 weighting sequences of MRI. Thesis is implemented in PYTHON programming language and Jupyter Notebooks environment. Software PyRadiomics is used for calculation of image features. Goal of this work is to determine best tumour region and weighting sequence for calculation of image features and consequently select set of features that are the best ones for classification of tumour grade and survival prediction. Part of thesis is dedicated to survival prediction using set of statistical tests, specifically Cox regression

Au-delà des techniques conventionnelles de diagnostic et de suivi du cancer, l’analyse radiomique a pour objectif de permettre une médecine plus personnalisée dans le domaine de la radiothérapie, en proposant une caractérisation non invasive de l’hétérogénéité tumorale. Basée sur l’extraction de paramètres quantitatifs avancés (histogrammes des intensités, texture, forme) issus de l’imagerie multimodale, cette technique a notamment prouvé son intérêt pour définir des signatures prédictives de la réponse aux traitements. Dans le cadre de cette thèse, des signatures de la récidive des cancers du col utérin ont notamment été développées, à partir de l’analyse radiomique seule ou en combinaison avec des biomarqueurs conventionnels, apportant des perspectives majeures dans la stratification des patients pouvant aboutir à une adaptation spécifique de la dosimétrie.En parallèle de ces études cliniques, différentes barrières méthodologiques ont été soulevées, notamment liées à la grande variabilité des protocoles et technologies d’acquisition des images, qui entraîne un biais majeur dans les études radiomiques multicentriques. Ces biais ont été évalués grâce à des images de fantômes et des images multicentriques de patients pour l’imagerie TEP, et deux méthodes de correction de l’effet de stratification ont été proposées. En IRM, une méthode de standardisation des images par harmonisation des histogrammes a été évaluée dans les tumeurs cérébrales.Pour aller plus loin dans la caractérisation de l’hétérogénéité intra-tumorale et permettre la mise en œuvre d’une radiothérapie personnalisée, une méthode d’analyse locale de la texture a été développée. Adaptée particulièrement aux images IRM de tumeurs cérébrales, ses capacités à différencier des sous-régions de radionécrose ou de récidive tumorale ont été évaluées. Dans ce but, les cartes paramétriques d’hétérogénéité ont été proposées à des experts comme des séquences IRM additionnelles.À l’issue de ce travail, une validation dans des centres extérieurs des modèles développés, ainsi que la mise en place d’essais cliniques intégrant ces méthodes pour personnaliser les traitements seront des étapes majeures dans l’intégration de l’analyse radiomique en routine clinique. / Beyond the conventional techniques of diagnosis and follow-up of cancer, radiomic analysis allows to personalize radiotherapy treatments, by proposing a non-invasive characterization of tumor heterogeneity. Based on the extraction of advanced quantitative parameters (histograms of intensities, texture, shape) from multimodal imaging, this technique has notably proved its interest in determining predictive signatures of treatment response. During this thesis, signatures of cervical cancer recurrence have been developed, based on radiomic analysis alone or in combination with conventional biomarkers, providing major perspectives in the stratification of patients that can lead to dosimetric treatment plan adaptation.However, various methodological barriers were raised, notably related to the great variability of the protocols and technologies of image acquisition, which leads to major biases in multicentric radiomic studies. These biases were assessed using phantom acquisitions and multicenter patient images for PET imaging, and two methods enabling a correction of the stratification effect were proposed. In MRI, a method of standardization of images by harmonization of histograms has been evaluated in brain tumors.To go further in the characterization of intra-tumor heterogeneity and to allow the implementation of a personalized radiotherapy, a method for local texture analysis has been developed. Specifically adapted to brain MRI, its ability to differentiate sub-regions of radionecrosis or tumor recurrence was evaluated. For this purpose, parametric heterogeneity maps have been proposed to experts as additional MRI sequences.In the future, validation of the predictive models in external centers, as well as the establishment of clinical trials integrating these methods to personalize radiotherapy treatments, will be mandatory steps for the integration of radiomic in the clinical routine.