Respiratory motion causes artifacts and blurring of cardiac structures in reconstructed images of SPECT and PET cardiac studies. Hysteresis in respiratory motion causes the organs to move in distinct paths during inspiration and expiration. Current respiratory motion correction methods use a signal generated by tracking the motion of the abdomen during respiration to bin list- mode data as a function of the magnitude of this respiratory signal. They thereby fail to account for hysteretic motion. The goal of this research was to demonstrate the effects of hysteretic respiratory motion and the importance of its correction for different medical imaging techniques particularly SPECT and PET. This study describes a novel approach for detecting and correcting hysteresis in clinical SPECT and PET studies. From the combined use of MRI and a synchronized Visual Tracking System (VTS) in volunteers we developed hysteretic modeling using the Bouc-Wen model with inputs from measurements of both chest and abdomen respiratory motion. With the MRI determined heart motion as the truth in the volunteer studies we determined the Bouc Wen model could match the behavior over a range of hysteretic cycles. The proposed approach was validated through phantom simulations and applied to clinical SPECT studies.

京都大学 / 0048 / 新制・課程博士 / 博士(医学) / 甲第18452号 / 医博第3907号 / 新制||医||1004(附属図書館) / 31330 / 京都大学大学院医学研究科医学専攻 / (主査)教授 伊達 洋至, 教授 武田 俊一, 教授 富樫 かおり / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

Lung cancer is the leading cause of cancer-related deaths worldwide. Radiotherapy is one of the main treatment modalities of lung cancer. However, the achievable accuracy of radiotherapy treatment is limited for lung-based tumors due to respiratory motion. Four-dimensional radiotherapy explicitly accounts for anatomic motion by characterizing the motion, creating a treatment plan that accounts for this motion, and delivering this plan to the moving anatomy. This thesis focuses on the current problems and solutions throughout the course of four-dimensional radiotherapy. For characterization of respiratory-induced motion, patient tumor motion data were analyzed. It is shown that tumor motion can be significant during radiotherapy treatment, and its extent, direction, and linearity vary considerably between patients, between treatment fractions, and between respiratory cycles. After this, approaches to four-dimensional intensity-modulated radiation therapy treatment planning were developed and investigated. Among the techniques to manage respiratory motion, tumor tracking using a dynamic multileaf collimator delivery technique was chosen as a promising method. A formalism to solve a general four-dimensional intensity-modulated radiation therapy treatment-planning problem was developed. Specific solutions to this problem accounting for tumor motion initially in one dimension and extending this to three dimensions were developed and investigated using four-dimensional computed tomography planning scans of lung cancer patients. For four-dimensional radiotherapy treatment delivery, accuracy of two-dimensional projection imaging methods was investigated. Geometric uncertainty due to the limitation of two-dimensional imaging in monitoring three-dimensional tumor motion during treatment delivery was quantified. This geometric uncertainty can be used to estimate proper margins when a single two-dimensional projection imager is used for four-dimensional treatment delivery. Lastly, tumor-tracking delivery using a moving average algorithm was investigated as an alternative delivery technique that reduces mechanical motion constraints of a multileaf collimator. Moving average tracking provides an approximate solution that can be immediately implemented for delivery of four-dimensional intensity-modulated radiation therapy treatment. The clinical implementation of four-dimensional guidance, intensity-modulated radiation therapy treatment planning, and dynamic multileaf collimator tracking delivery may have a positive impact on the treatment of lung cancer.

<p>Imaging respiratory induced tumor motion in the radiation therapy treatment room could eliminate the necessity for large motion encompassing margins that result in excessive irradiation of healthy tissues. Currently available image guidance technologies are ill-suited for this task. Two-dimensional fluoroscopic images are acquired with sufficient speed to image respiratory motion. However, volume information is not present, and soft tissue structures are often not visible because a large volume is projected onto a single plane. Currently available volumetric imaging modalities are not acquired with sufficient speed to capture full motion trajectory information. Four-dimensional cone-beam computed tomography (4D CBCT) using a gantry mounted 2D flat panel imaging device has been proposed but has been limited by high doses, long scan times and severe under-sampling artifacts. The focus of the work completed in this thesis was to find ways to improve 4D imaging using a gantry mounted 2D kV imaging system. Specifically, the goals were to investigate methods for minimizing imaging dose and scan time while achieving consistent, controllable, high quality 4D images.</p><p>First, we introduced four-dimensional digital tomosynthesis (4D DTS) and characterized its potential for 3D motion analysis using a motion phantom. The motion phantom was programmed to exhibit motion profiles with various known amplitudes in all three dimensions and scanned using a 2D kV imaging system mounted on a linear accelerator. Two arcs of projection data centered about the anterior-posterior and lateral axes were used to reconstruct phase resolved DTS coronal and sagittal images. Respiratory signals were obtained by analyzing projection data, and these signals were used to derive phases for each of the projection images. Projection images were sorted according to phase, and DTS phase images were reconstructed for each phase bin. 4D DTS target location accuracies for peak inhalation and peak exhalation in all three dimensions were limited only by the 0.5 mm pixel resolution for all DTS scan angles. The average localization errors for all phases of an assymetric motion profile with a 2 cm peak-to-peak amplitude were 0.68, 0.67 and 1.85 mm for 60 <super> o <super/> 4D DTS, 360<super> o <super/> CBCT and 4DCT, respectively. Motion artifacts for 4D DTS were found to be substantially less than those seen in 4DCT, which is the current clinical standard in 4D imaging. </p><p>We then developed a comprehensive framework for relating patient respiratory parameters with acquisition and reconstruction parameters for slow gantry rotation 4D DTS and 4D CBCT imaging. This framework was validated and optimized with phantom and lung patient studies. The framework facilitates calculation of optimal frame rates and gantry rotation speeds based on patient specific respiratory parameters and required temporal resolution (task dependent). We also conducted lung patient studies to investigate required scan angles for 4D DTS and achievable dose and scan times for 4D DTS and 4D CBCT using the optimized framework. This explicit and comprehensive framework of relationships allowed us to demonstrate that under-sampling artifacts can be controlled, and 4D CBCT images can be acquired using lower doses than previously reported. We reconstructed 4D CBCT images of three patients with accumulated doses of 4.8 to 5.7 cGy. These doses are three times less than the doses used for the only previously reported 4D CBCT investigation that did not report images characterized by severe under-sampling artifacts. </p><p>We found that scan times for 200<super> o <super/> 4D CBCT imaging using acquisition sequences optimized for reduction of imaging dose and under-sampling artifacts were necessarily between 4 and 7 minutes (depending on patient respiration). The results from lung patient studies concluded that scan times could be reduced using 4D DTS. Patient 4D DTS studies demonstrated that tumor visibility for the lung patients we studied could be achieved using 30<super> o <super/> scan angles for coronal views. Scan times for those cases were between 41 and 50 seconds. Additional dose reductions were also demonstrated. Image doses were between 1.56 and 2.13 cGy. These doses are well below doses for standard CBCT scans. The techniques developed and reported in this thesis demonstrate how respiratory motion can be imaged in the radiotherapy treatment room using clinically feasible imaging doses and scan times.</p> / Dissertation

<p>To date, image localization of mobile tumors prior to radiation delivery has primarily been confined to 2D and 3D technologies, such as fluoroscopy and 3D cone-beam CT (3D-CBCT). Due to the limited information from these images, larger volumes of healthy tissue are often irradiated in order to ensure the radiation field encompasses the entirety of the target motion. Since the overarching goal of radiation therapy is to deliver maximum dose to cancerous cells and simultaneously minimize the radiation delivered to healthy surrounding tissues, it would be ideal to use 4D imaging to obtain time-resolved volume images of the tumor motion during respiration. </p><p>4D-CBCT imaging has been previously investigated, but has not yet seen large clinical translation due to the obstacles of long acquisition time and large image radiation dose. Furthermore, 4D-CBCT currently requires the use of external surrogates to correlate the patient's respiration with the image acquisition process. This correlation has been under question by a multitude of studies demonstrating the uncertainties that exist between the surrogate and the actual motion of the internal anatomy. Errors in the correlation process may result in image artifacts, which could potentially lead to reconstructions with inaccurate target volumes, thereby defeating the purpose of even using 4D-CBCT. </p><p>It is therefore the aim of this dissertation to initially highlight an additional limitation of using 3D-CBCT for imaging respiratory motion and thereby reiterate the need for 4D-CBCT imaging in the treatment room, develop a simple and efficient technique to achieve markerless, self-sorted 4D-CBCT and finally to comprehensively evaluate its robustness across a variety of potential clinical scenarios with a digital human phantom. </p><p>People often spend a longer period of time exhaling as compared with inhaling, and some do so in an extremely disproportionate manner. To demonstrate the disadvantage of using 3D-CBCT in such instances, a dynamic thorax phantom was imaged with a large variety of simulated and patient-derived respiratory traces of ratios of time spent in the inspiration phase versus time spent in the expiration phase (I/E ratio). Canny edge detection and contrast measures were employed to compare the internal target volumes (ITVs) generated per profile. The results revealed that an I/E ratio of less than one can lead to potential underestimation of the ITV with the severity increasing as the inspiration becomes more disproportionate to the expiration. This occurs because of the loss of contrast in the inspiration phase, due to the fewer number of projections acquired there. The measured contrast reduction was as high as 94% for small targets (0.5 cm) moving large amplitudes (2.0 cm) and still as much as 22.3% for large targets (3.0 cm) moving small amplitudes (0.5 cm). This is alarming because the degraded visibility of the target in the inspiration phase may inaccurately impact the alignment of the planning ITV with that of the FB-CBCT and thereby affect the accuracy of the localization and consequent radiation delivery. These potential errors can be avoided with the use of 4D-CBCT instead, to form the composite volume and serve as the verification ITV for alignment.</p><p>In order to delineate accurate target volumes from 4D-CBCT phase images, it is crucial that the projections be properly associated with the patient's respiration. Thus, in order to improve previously developed 4D-CBCT techniques, the basics of Fourier Transform (FT) theory were utilized to extract the respiratory signal directly from the acquired projection data. Markerless, self-sorted 4D-CBCT reconstruction was achieved by developing methods based on the phase and magnitude information of the Fourier Transform. Their performance was subsequently compared to the gold standard of visual identification of peak-inspiration projections. Slow-gantry acquired projections of two sets of physical phantom data with sinusoidal respiratory cycles of 3 and 6 seconds as well as three patients were used as initial evaluation of the feasibility of the Fourier technique. Quantitative criteria consisted of average difference in respiratory phase (ADRP) and percentage of projections assigned within 10% respiratory phase of the gold standard (PP10). For all five projection datasets, the results supported feasibility of both FT-Phase and FT-Magnitude methods with ADRP values less than 5.3% and PP10 values of 87.3% and above. </p><p>Because the technique proved to be promising in the initial feasibility study, a more comprehensive evaluation was necessary in order to assess the robustness of the technique across a larger set of possibilities that may be encountered in the clinic. A 4D digital XCAT phantom was used to generate an array of respiratory and anatomical variables that affect the performance of the technique. The respiratory variables studied included: inspiration to expiration ratio, respiratory cycle length, diaphragmatic motion amplitude, AP chest wall expansion amplitude, breathing irregularities such as baseline shift and inconsistent peak-inspiration amplitude, as well as six breathing profiles derived from cine-MRI images of three healthy volunteers and three lung cancer patients. The anatomical variables studied included: male and female patient size (physical dimension and adipose content), body-mass-index (BMI) category, tumor location, and percentage of the lung in the field-of-view (FOV) of the projection data. CBCT projections of each XCAT phantom were then generated. Additional external imaging factors such as image noise and detector wobble were added to select cases with different percentages of lung in the projection FOV to investigate any effects on the robustness. FT-Phase and FT-Magnitude were each applied and quantitatively compared to the gold standard. Both methods proved to be robust across the studied scenarios with ADRP<10% and PP10>90%, when incorporating minor modifications to region-of-interest (ROI) selection and/or low-frequency location to certain cases of diaphragm amplitude and lung percentage in the FOV of the projection (for which a method may have previously struggled). Nevertheless, in the instance where one method initially faltered, the other method prevailed and successfully identified peak-inspiration projections. This is promising because it suggests that the two methods provide complementary information to each other. To ensure appropriate clinical adaptation of markerless, self-sorted 4D-CBCT, perhaps an optimal integration of the two methods can be developed.</p> / Dissertation

Medical imaging has seen a rapid development in its clinical use in assessment of treatment outcome, disease monitoring and diagnosis over the last few decades. Yet, the vast amount of available image data limits the practical use of this potentially very valuable source of information for radiologists and physicians. Therefore, the design of computer-aided medical image analysis is of great importance to imaging in clinical practice. This thesis deals with the problem of deformable image registration in the context of lung imaging, and addresses three of the major challenges involved in this challenging application, namely: designing an image similarity for multi-modal scans or scans of locally changing contrast, modelling of complex lung motion, which includes sliding motion, and approximately globally optimal mathematical optimisation to deal with large motion of small anatomical features. The two most important contributions made in this thesis are: the formulation of a multi-dimensional structural image representation, which is independent of modality, robust to intensity distortions and very discriminative for different image features, and a discrete optimisation framework, based on an image-adaptive graph structure, which enables a very efficient optimisation of large dense displacement spaces and deals well with sliding motion. The derived methods are applied to two different clinical applications in pulmonary image analysis: motion correction for breathing-cycle computed tomography (CT) volumes, and deformable multi-modal fusion of CT and magnetic resonance imaging chest scans. The experimental validation demonstrates improved registration accuracy, a high quality of the estimated deformations, and much lower computational complexity, all compared to several state-of-the-art deformable registration techniques.

Objectif : Etudier l’impact volumétrique et dosimétrique de l’intégration des données de tomographie par émission de positons (TEP) en radiothérapie (RT) à travers deux modèles cliniques : les cancers oto-rhino-laryngologiques (ORL) et les cancers pulmonaires. Matériel et méthodes : Pour les cancers ORL, après un travail préalable sur fantôme pour mise au point d’une méthode de segmentation automatique par seuillage adaptatif, deux séries de neuf et 15 patients présentant un cancer ORL traité par RT, ont bénéficié d’une TEP respectivement au 18F-Fluorodeoxyglucose (18F-FDG) et au 18F-Fluoromisonidazole (18F-FMISO), traceur de l’hypoxie. Les modifications volumétriques et dosimétriques induites par ces examens ont été analysées. Pour le 18F-FMISO, différents temps d’acquisition et différentes méthodes de segmentation ont également été étudiés. Pour les cancers pulmonaires, l’impact sur la RT d’une TEP-4D au 18F-FDG avec correction des effets de volume partiel (EVP) et application de différentes méthodes de segmentation, a été évalué à travers l’analyse des sept premiers patients inclus dans le protocole PULMOTEP, promu par le CHU de Bordeaux. Résultats : Pour les cancers ORL, la TEP au 18F-FDG a conduit à une réduction des volumes de RT de 40% tout en individualisant des zones de « mismatch » entre TEP et scanner. Pour la TEP au 18F-FMISO, un meilleur contraste des images était retrouvé à 4h. Cependant, les volumes segmentés à 3 et 4h n’étaient pas significativement différents, permettant d’envisager en pratique courante des acquisitions moins tardives à 3h. L’utilisation d’une TEP au 18F-FMISO permettait d’envisager la réalisation d’une « escalade de dose » sur les zones hypoxiques avec une augmentation du taux de probabilité de contrôle tumoral de 18,1% sans augmentation excessive de la toxicité. Pour les cancers pulmonaires, il n’était pas retrouvé d’impact de la correction du mouvement respiratoire, un seul patient sur les sept étudiés présentant une tumeur mobile. Un impact constant de la correction des EVP était par contre retrouvé avec une augmentation de l’activité tumorale maximale de 27% et une diminution des volumes segmentés de 15%.Conclusion : Pour les cancers ORL, la validation de ces résultats nécessite la réalisation d’études cliniques. Pour les cancers pulmonaires, l’utilisation d’une TEP-4D avec correction du mouvement respiratoire doit être envisagée au cas par cas. L’implémentation en clinique de logiciels de correction des EVP semble, par contre, à encourager. / Objective: To study the impact of Positron Emission Tomography (PET) data on radiotherapy (RT) planning through two clinical models: the head-and-neck cancers (HNC) and the pulmonary cancers. Methods and Materials: For HNC, after a previous phantom study in order to determinate an automatic segmentation method with adaptive thresholding, two series of nine and 15 patients selected for RT, underwent PET with 18F-Fluorodeoxyglucose (FDG) and 18F-Fluoromisonidazole (FMISO). The impact on RT target volumes (TV) and dosimetries was evaluated. For FMISO-PET, several time acquisitions and several segmentation methods were assessed. For pulmonary cancers, the use of a four-dimensional (4D) FDG-PET with partial volume effect (PVE) correction and several segmentation methods was evaluated through the first seven patients enrolled in the PULMOTEP protocol performed by the CHU of Bordeaux. Results: For HNC, FDG-PET led to a RT TV reduction of 40%, with mismatches between PET and CT data. For FMISO-PET images, a better contrast was obtained 4h after FMISO injection. However, segmented volumes obtained at 3 and 4h were not statistically different allowing PET- acquisitions at 3h in routine clinical practice. The use of FMISO-PET allows considering « dose escalation » on hypoxic TV with an increase of tumour control probability by 18,1% without excessive increase of toxicities. For pulmonary cancers, there was no impact of the respiratory motion correction but only one patient on seven presented a mobile tumour. PVE correction had impact on RT TV with an increase of the maximal tumoural activity by 27% and a volume reduction of 15%. Conclusion: For HNC, the validation of these results needs clinical and prospective studies. For pulmonary cancers, the use of 4D-PET must be decided case by case. On the other side, the implementation of automatic software for PVE correction seems interesting.

L'objectif de cette thèse est d’étudier temporellement des phénomènes évolutifs à l'aide de la tomographie d'émission monophotonique (TEMP). La première partie de cette thèse traite le problème du mouvement respiratoire dans les images TEMP de souris. Nous présentons ici une méthode permettant de détecter ce mouvement respiratoire dans les images TEMP 4D, d'extraire un signal respiratoire intrinsèque, et de déterminer la phase du cycle respiratoire sans mouvement la plus large possible. Les données enregistrées durant ces phases sans mouvement sont alors utilisées pour reconstruire une seule image TEMP 3D sans artefact de mouvement par acquisition. Les images ainsi reconstruites présentent un bon compromis en terme de statistiques et de précision des mesures par rapport aux images TEMP 3D de base et TEMP 4D. Dans la deuxième partie, nous étudions la cinétique d'incorporation de l'iode dans l'estomac de souris à partir d'images TEMP 4D. Afin de comprendre le rôle biologique de cette accumulation dans l'estomac, nous avons modélisé le phénomène par une approche d'analyse compartimentale avec un modèle simplifiée à deux compartiments (paroi et cavité stomacale) et une entrée (sang). Les courbes temps-activité (TAC) de chaque compartiment sont déduites des observations et une première estimation des paramètres a été obtenue. / The aim of this thesis is to investigate temporally evolving phenomena with the use of single photon emission computed tomography (SPECT).The first part of this thesis addresses the problem of respiratory motion in SPECT images of mice. The presented method permits us to detect the respiratory motion in 4D SPECT images, to extract an intrinsic respiratory signal and to determine the widest possible phase of the respiratory cycle without movement. The data recorded during these motionless phases are then used to reconstruct a single 3D SPECT image without motion artefacts per acquisition. Reconstructed motionless SPECT images present a good compromise in terms of statistics and accuracy of the measurements with respect to basic 3D SPECT and 4D SPECT images. In the second part, we study the iodide uptake kinetics in the stomach 99mTc-pertechnetate biodistribution with the of mice through the study of use of 4D SPECT images. To understand the biological role of the iodide accumulation in the stomach, we modelled the phenomenon with a compartmental analysis approach using a simplified two-compartment (stomach wall and cavity) model with one input (blood). Time activity curves (TAC) of each compartment are deduced from observations and a first estimation of the parameters was obtained.

Modern big data often emerge as tensors. Standard statistical methods are inadequate to deal with datasets of large volume, high dimensionality, and complex structure. Therefore, it is important to develop algorithms such as low-rank tensor decomposition for data compression, dimensionality reduction, and approximation. With the advancement in technology, high-dimensional images are becoming ubiquitous in the medical field. In lung radiation therapy, the respiratory motion of the lung introduces variabilities during treatment as the tumor inside the lung is moving, which brings challenges to the precise delivery of radiation to the tumor. Several approaches to quantifying this uncertainty propose using a model to formulate the motion through a mathematical function over time. [Li et al., 2011] uses principal component analysis (PCA) to propose one such model using each image as a long vector. However, the images come in a multidimensional arrays, and vectorization breaks the spatial structure. Driven by the needs to develop low-rank tensor decomposition and provided the 4DCT and Displacement Vector Field (DVF), we introduce two tensor decompositions, Population Value Decomposition (PVD) and Population Tucker Decomposition (PTD), to estimate the respiratory lung motion with high levels of accuracy and data compression. The first algorithm is a generalization of PVD [Crainiceanu et al., 2011] to higher order tensor. The second algorithm generalizes the concept of PVD using Tucker decomposition. Both algorithms are tested on clinical and phantom DVFs. New metrics for measuring the model performance are developed in our research. Results of the two new algorithms are compared to the result of the PCA algorithm.

Breathing is a vital phenomenon that implies synergy of various anatomical structures that constitute the thorax. Joint physiology remains a relatively poorly-known component of the overall thorax physiology. Quantitative literature related to in vivo thorax kinematics during breathing is scarce. The present work focuses specifically on developing and applying a methodology to reach this goal. The developed method combined processing of CT data obtained at different lung volumes and infographic techniques. Detailed ranges of motion (ROMs) and axes of movement (mean helical axes, MHAs) were obtained at costovertebral joints in 12 asymptomatic subjects; rib ROMs gradually decrease with increasing rib number; lung volume and rib level have a significant influence on rib ROM; MHAs did not differ between rib levels. In addition, the method was applied on a sample of 10 patients with cystic fibrosis. The pathological condition significantly influenced CVJ ROMs while the orientation of the MHAs did not differ. Finally, the sternal displacement, sternal angle variations and sternocostal joints (SCJ at rib1 to 7) kinematics during breathing motion were analyzed. Rib ranges of motion relative to sternum decreased with increasing rib number similarly to CVJ. Orientation of the MHAs did not differ between SCJ levels. A significant linear correlation was demonstrated between sternum vertical displacement and rib ranges of motion at both CVJ and SCJ. The present work substantially contributes to 3D modelling of human thorax in breathing at a joint level both qualitatively and quantitatively. / Doctorat en Sciences biomédicales et pharmaceutiques (Médecine) / info:eu-repo/semantics/nonPublished