Towards Detecting Changes in Underwater Image Sequences

Detecting changes between images of the same scene taken at different times is of great interest for monitoring and understanding the environment. It is widely used for on-land application but suffers from different constraints. Unfortunately, Change detection algorithms require highly accurate geometric and photometric registration. This requirement has precluded their use in underwater imagery in the past. In this paper, the change detection techniques available nowadays for on-land application were analyzed and a method to automatically detect the changes in sequences of underwater images is proposed. Target application scenarios are habitat restoration sites, or area monitoring after sudden impacts from hurricanes or ship groundings. The method is based on the creation of a 3D terrain model from one image sequence over an area of interest. This model allows for synthesizing textured views that correspond to the same viewpoints of a second image sequence. The generated views are photometrically matched and corrected against the corresponding frames from the second sequence. Standard change detection techniques are then applied to find areas of difference. Additionally, the paper shows that it is possible to detect false positives, resulting from non-rigid objects, by applying the same change detection method to the first sequence exclusively. The developed method was able to correctly find the changes between two challenging sequences of images from a coral reef taken one year apart and acquired with two different cameras

Recovering Euclidean deformable models from stereo-motion

In this paper we present a novel structure from motion (SfM) approach able to infer 3D deformable models from uncalibrated stereo images. Using a stereo setup dramatically improves the 3D model estimation when the observed 3D shape is mostly deforming without undergoing strong rigid motion. Our approach first calibrates the stereo system automatically and then computes a single metric rigid structure for each frame. Afterwards, these 3D shapes are aligned to a reference view using a RANSAC method in order to compute the mean shape of the object and to select the subset of points on the object which have remained rigid throughout the sequence without deforming. The selected rigid points are then used to compute frame-wise shape registration and to extract the motion parameters robustly from frame to frame. Finally, all this information is used in a global optimization stage with bundle adjustment which allows to refine the frame-wise initial solution and also to recover the non-rigid 3D model. We show results on synthetic and real data that prove the performance of the proposed method even when there is no rigid motion in the original sequence

A Motion compensated filtering approach to remove sunlight flicker in shallow water images

A common problem in video surveys in very shallow waters is the presence of strong light fluctuations, due to sun light refraction. Refracted sunlight casts fast moving patterns, which can significantly degrade the quality of the acquired data. Motivated by the growing need to improve the quality of shallow water imagery, we propose a method to remove sunlight patterns in video sequences. The method exploits the fact that video sequences allow several observations of the same area of the sea floor, over time. It is based on computing the image difference between a given reference frame and the temporal median of a registered set of neighboring images. A key observation is that this difference will have two components with separable spectral content. One is related to the illumination field (lower spatial frequencies) and the other to the registration error (higher frequencies). The illumination field, recovered by lowpass filtering, is used to correct the reference image. In addition to removing the sunflickering patterns, an important advantage of the approach is the ability to preserve the sharpness in corrected image, even in the presence of registration inaccuracies. The effectiveness of the method is illustrated in image sets acquired under strong camera motion containing non-rigid benthic structures. The results testify the good performance and generality of the approach

Patient-specific model of a scoliotic torso for surgical planning

A method for the construction of a patient-specific model of a scoliotic torso for surgical planning via inter- patient registration is presented. Magnetic Resonance Images (MRI) of a generic model are registered to surface topography (TP) and X-ray data of a test patient. A partial model is first obtained via thin-plate spline registration between TP and X-ray data of the test patient. The MRIs from the generic model are then fit into the test patient using articulated model registration between the vertebrae of the generic model’s MRIs in prone position and the test patient’s X-rays in standing position. A non-rigid deformation of the soft tissues is performed using a modified thin-plate spline constrained to maintain bone rigidity and to fit in the space between the vertebrae and the surface of the torso. Results show average Dice values of 0.975 ± 0.012 between the MRIs following inter-patient registration and the surface topography of the test patient, which is comparable to the average value of 0.976 ± 0.009 previously obtained following intra-patient registration. The results also show a significant improvement compared to rigid inter-patient registration. Future work includes validating the method on a larger cohort of patients and incorporating soft tissue stiffness constraints. The method developed can be used to obtain a geometric model of a patient including bone structures, soft tissues and the surface of the torso which can be incorporated in a surgical simulator in order to better predict the outcome of scoliosis surgery, even if MRI data cannot be acquired for the patient.

Articulated Model Registration of MRI/X-Ray Spine Data

This paper presents a method based on articulated models for the registration of spine data extracted from multimodal medical images of patients with scoliosis. With the ultimate aim being the development of a complete geometrical model of the torso of a scoliotic patient, this work presents a method for the registration of vertebral column data using 3D magnetic resonance images (MRI) acquired in prone position and X-ray data acquired in standing position for five patients with scoliosis. The 3D shape of the vertebrae is estimated from both image modalities for each patient, and an articulated model is used in order to calculate intervertebral transformations required in order to align the vertebrae between both postures. Euclidean distances between anatomical landmarks are calculated in order to assess multimodal registration error. Results show a decrease in the Euclidean distance using the proposed method compared to rigid registration and more physically realistic vertebrae deformations compared to thin-plate-spline (TPS) registration thus improving alignment.

Towards Detecting Changes in Underwater Image Sequences

Detecting changes between images of the same scene taken at different times is of great interest for monitoring and understanding the environment. It is widely used for on-land application but suffers from different constraints. Unfortunately, Change detection algorithms require highly accurate geometric and photometric registration. This requirement has precluded their use in underwater imagery in the past. In this paper, the change detection techniques available nowadays for on-land application were analyzed and a method to automatically detect the changes in sequences of underwater images is proposed. Target application scenarios are habitat restoration sites, or area monitoring after sudden impacts from hurricanes or ship groundings. The method is based on the creation of a 3D terrain model from one image sequence over an area of interest. This model allows for synthesizing textured views that correspond to the same viewpoints of a second image sequence. The generated views are photometrically matched and corrected against the corresponding frames from the second sequence. Standard change detection techniques are then applied to find areas of difference. Additionally, the paper shows that it is possible to detect false positives, resulting from non-rigid objects, by applying the same change detection method to the first sequence exclusively. The developed method was able to correctly find the changes between two challenging sequences of images from a coral reef taken one year apart and acquired with two different cameras

A Motion compensated filtering approach to remove sunlight flicker in shallow water images

A common problem in video surveys in very shallow waters is the presence of strong light fluctuations, due to sun light refraction. Refracted sunlight casts fast moving patterns, which can significantly degrade the quality of the acquired data. Motivated by the growing need to improve the quality of shallow water imagery, we propose a method to remove sunlight patterns in video sequences. The method exploits the fact that video sequences allow several observations of the same area of the sea floor, over time. It is based on computing the image difference between a given reference frame and the temporal median of a registered set of neighboring images. A key observation is that this difference will have two components with separable spectral content. One is related to the illumination field (lower spatial frequencies) and the other to the registration error (higher frequencies). The illumination field, recovered by lowpass filtering, is used to correct the reference image. In addition to removing the sunflickering patterns, an important advantage of the approach is the ability to preserve the sharpness in corrected image, even in the presence of registration inaccuracies. The effectiveness of the method is illustrated in image sets acquired under strong camera motion containing non-rigid benthic structures. The results testify the good performance and generality of the approach

Recovering Euclidean deformable models from stereo-motion

In this paper we present a novel structure from motion (SfM) approach able to infer 3D deformable models from uncalibrated stereo images. Using a stereo setup dramatically improves the 3D model estimation when the observed 3D shape is mostly deforming without undergoing strong rigid motion. Our approach first calibrates the stereo system automatically and then computes a single metric rigid structure for each frame. Afterwards, these 3D shapes are aligned to a reference view using a RANSAC method in order to compute the mean shape of the object and to select the subset of points on the object which have remained rigid throughout the sequence without deforming. The selected rigid points are then used to compute frame-wise shape registration and to extract the motion parameters robustly from frame to frame. Finally, all this information is used in a global optimization stage with bundle adjustment which allows to refine the frame-wise initial solution and also to recover the non-rigid 3D model. We show results on synthetic and real data that prove the performance of the proposed method even when there is no rigid motion in the original sequence

A coupling technique for non-matching finite element meshes

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Information Filtering for Ultrasound-based Real-time Registration

This paper presents methods based on Information Filters for solving matching problems with emphasis on real-time, or effectively real-time applications. Both applications discussed in this work deal with ultrasound-based rigid registration in computer-assisted orthopedic surgery. In the first application, the usual workflow of rigid registration is reformulated such that registration algorithms would iterate while the surgeon is acquiring ultrasound images of the anatomy to be operated. Using this effectively real-time approach to registration, the surgeon would then receive feedback in order to better gauge the quality of the final registration outcome. The second application considered in this paper circumvents the need to attach physical markers to bones for anatomical referencing. Experiments using anatomical objects immersed in water are performed in order to evaluate and compare the different methods presented herein, using both 2D as well as real-time 3D ultrasound.

Automatic extraction of femur contours from calibrated fluoroscopic images

Automatic identification and extraction of bone contours from X-ray images is an essential first step task for further medical image analysis. In this paper we propose a 3D statistical model based framework for the proximal femur contour extraction from calibrated X-ray images. The automatic initialization is solved by an estimation of Bayesian network algorithm to fit a multiple component geometrical model to the X-ray data. The contour extraction is accomplished by a non-rigid 2D/3D registration between a 3D statistical model and the X-ray images, in which bone contours are extracted by a graphical model based Bayesian inference. Preliminary experiments on clinical data sets verified its validity

Nonrigid Motion Compensation of Free Breathing Acquired Myocardial Perfusion Data

In this work, we present a novel method to compensate the movement in images acquired during free breathing using first-pass gadolinium enhanced, myocardial perfusion magnetic resonance imaging (MRI). First, we use independent component analysis (ICA) to identify the optimal number of independent components (ICs) that separate the breathing motion from the intensity change induced by the contrast agent. Then, synthetic images are created by recombining the ICs, but other then in previously published work (Milles et al. 2008), we omit the component related to motion, and therefore, the resulting reference image series is free of motion. Motion compensation is then achieved by using a multi-pass non-rigid image registration scheme. We tested our method on 15 distinct image series (5 patients) consisting of 58 images each and we validated our method by comparing manually tracked intensity profiles of the myocardial sections to automatically generated ones before and after registration. The average correlation to the manually obtained curves before registration 0:89 0:11 was increased to 0:98 0:02

Mejoras en el procesamiento de imágenes de TC para el tratamiento con radioterapia adaptativa del cáncer de próstata

El cáncer de próstata es el tipo de cáncer con mayor prevalencia entre los hombres del mundo occidental y, pese a tener una alta tasa de supervivencia relativa, es la segunda mayor causa de muerte por cáncer en este sector de la población. El tratamiento de elección frente al cáncer de próstata es, en la mayoría de los casos, la radioterapia externa. Las técnicas más modernas de radioterapia externa, como la radioterapia modulada en intensidad, permiten incrementar la dosis en el tumor mientras se reduce la dosis en el tejido sano. Sin embargo, la localización del volumen objetivo varía con el día de tratamiento, y se requieren movimientos muy pequeños de los órganos para sacar partes del volumen objetivo fuera de la región terapéutica, o para introducir tejidos sanos críticos dentro. Para evitar esto se han desarrollado técnicas más avanzadas, como la radioterapia guiada por imagen, que se define por un manejo más preciso de los movimientos internos mediante una adaptación de la planificación del tratamiento basada en la información anatómica obtenida de imágenes de tomografía computarizada (TC) previas a la sesión terapéutica. Además, la radioterapia adaptativa añade la información dosimétrica de las fracciones previas a la información anatómica. Uno de los fundamentos de la radioterapia adaptativa es el registro deformable de imágenes, de gran utilidad a la hora de modelar los desplazamientos y deformaciones de los órganos internos. Sin embargo, su utilización conlleva nuevos retos científico-tecnológicos en el procesamiento de imágenes, principalmente asociados a la variabilidad de los órganos, tanto en localización como en apariencia. El objetivo de esta tesis doctoral es mejorar los procesos clínicos de delineación automática de contornos y de cálculo de dosis acumulada para la planificación y monitorización de tratamientos con radioterapia adaptativa, a partir de nuevos métodos de procesamiento de imágenes de TC (1) en presencia de contrastes variables, y (2) cambios de apariencia del recto. Además, se pretende (3) proveer de herramientas para la evaluación de la calidad de los contornos obtenidos en el caso del gross tumor volumen (GTV). Las principales contribuciones de esta tesis doctoral son las siguientes: _ 1. La adaptación, implementación y evaluación de un algoritmo de registro basado en el flujo óptico de la fase de la imagen como herramienta para el cálculo de transformaciones no-rígidas en presencia de cambios de intensidad, y su aplicabilidad a tratamientos de radioterapia adaptativa en cáncer de próstata con uso de agentes de contraste radiológico. Los resultados demuestran que el algoritmo seleccionado presenta mejores resultados cualitativos en presencia de contraste radiológico en la vejiga, y no distorsiona la imagen forzando deformaciones poco realistas. 2. La definición, desarrollo y validación de un nuevo método de enmascaramiento de los contenidos del recto (MER), y la evaluación de su influencia en el procedimiento de radioterapia adaptativa en cáncer de próstata. Las segmentaciones obtenidas mediante el MER para la creación de máscaras homogéneas en las imágenes de sesión permiten mejorar sensiblemente los resultados de los algoritmos de registro en la región rectal. Así, el uso de la metodología propuesta incrementa el índice de volumen solapado entre los contornos manuales y automáticos del recto hasta un valor del 89%, cercano a los resultados obtenidos usando máscaras manuales para el registro de las dos imágenes. De esta manera se pueden corregir tanto el cálculo de los nuevos contornos como el cálculo de la dosis acumulada. 3. La definición de una metodología de evaluación de la calidad de los contornos del GTV, que permite la representación de la distribución espacial del error, adaptándola a volúmenes no-convexos como el formado por la próstata y las vesículas seminales. Dicha metodología de evaluación, basada en un nuevo algoritmo de reconstrucción tridimensional y una nueva métrica de cuantificación, presenta resultados precisos con una gran resolución espacial en un tiempo despreciable frente al tiempo de registro. Esta nueva metodología puede ser una herramienta útil para la comparación de distintos algoritmos de registro deformable orientados a la radioterapia adaptativa en cáncer de próstata. En conclusión, el trabajo realizado en esta tesis doctoral corrobora las hipótesis de investigación postuladas, y pretende servir como cimiento de futuros avances en el procesamiento de imagen médica en los tratamientos de radioterapia adaptativa en cáncer de próstata. Asimismo, se siguen abriendo nuevas líneas de aplicación futura de métodos de procesamiento de imágenes médicas con el fin de mejorar los procesos de radioterapia adaptativa en presencia de cambios de apariencia de los órganos, e incrementar la seguridad del paciente. I.2 Inglés Prostate cancer is the most prevalent cancer amongst men in the Western world and, despite having a relatively high survival rate, is the second leading cause of cancer death in this sector of the population. The treatment of choice against prostate cancer is, in most cases, external beam radiation therapy. The most modern techniques of external radiotherapy, as intensity modulated radiotherapy, allow increasing the dose to the tumor whilst reducing the dose to healthy tissue. However, the location of the target volume varies with the day of treatment, and very small movements of the organs are required to pull out parts of the target volume outside the therapeutic region, or to introduce critical healthy tissues inside. Advanced techniques, such as the image-guided radiotherapy (IGRT), have been developed to avoid this. IGRT is defined by more precise handling of internal movements by adapting treatment planning based on the anatomical information obtained from computed tomography (CT) images prior to the therapy session. Moreover, the adaptive radiotherapy adds dosimetric information of previous fractions to the anatomical information. One of the fundamentals of adaptive radiotherapy is deformable image registration, very useful when modeling the displacements and deformations of the internal organs. However, its use brings new scientific and technological challenges in image processing, mainly associated to the variability of the organs, both in location and appearance. The aim of this thesis is to improve clinical processes of automatic contour delineation and cumulative dose calculation for planning and monitoring of adaptive radiotherapy treatments, based on new methods of CT image processing (1) in the presence of varying contrasts, and (2) rectum appearance changes. It also aims (3) to provide tools for assessing the quality of contours obtained in the case of gross tumor volume (GTV). The main contributions of this PhD thesis are as follows: 1. The adaptation, implementation and evaluation of a registration algorithm based on the optical flow of the image phase as a tool for the calculation of non-rigid transformations in the presence of intensity changes, and its applicability to adaptive radiotherapy treatment in prostate cancer with use of radiological contrast agents. The results demonstrate that the selected algorithm shows better qualitative results in the presence of radiological contrast agents in the urinary bladder, and does not distort the image forcing unrealistic deformations. 2. The definition, development and validation of a new method for masking the contents of the rectum (MER, Spanish acronym), and assessing their impact on the process of adaptive radiotherapy in prostate cancer. The segmentations obtained by the MER for the creation of homogenous masks in the session CT images can improve significantly the results of registration algorithms in the rectal region. Thus, the use of the proposed methodology increases the volume overlap index between manual and automatic contours of the rectum to a value of 89%, close to the results obtained using manual masks for both images. In this way, both the calculation of new contours and the calculation of the accumulated dose can be corrected. 3. The definition of a methodology for assessing the quality of the contours of the GTV, which allows the representation of the spatial distribution of the error, adapting it to non-convex volumes such as that formed by the prostate and seminal vesicles. Said evaluation methodology, based on a new three-dimensional reconstruction algorithm and a new quantification metric, presents accurate results with high spatial resolution in a time negligible compared to the registration time. This new approach may be a useful tool to compare different deformable registration algorithms oriented to adaptive radiotherapy in prostate cancer In conclusion, this PhD thesis corroborates the postulated research hypotheses, and is intended to serve as a foundation for future advances in medical image processing in adaptive radiotherapy treatment in prostate cancer. In addition, it opens new future applications for medical image processing methods aimed at improving the adaptive radiotherapy processes in the presence of organ’s appearance changes, and increase the patient safety.

Convergence analysis and validation of low cost distance metrics for computational cost reduction of the Iterative Closest Point algorithm

The Iterative Closest Point algorithm (ICP) is commonly used in engineering applications to solve the rigid registration problem of partially overlapped point sets which are pre-aligned with a coarse estimate of their relative positions. This iterative algorithm is applied in many areas such as the medicine for volumetric reconstruction of tomography data, in robotics to reconstruct surfaces or scenes using range sensor information, in industrial systems for quality control of manufactured objects or even in biology to study the structure and folding of proteins. One of the algorithm’s main problems is its high computational complexity (quadratic in the number of points with the non-optimized original variant) in a context where high density point sets, acquired by high resolution scanners, are processed. Many variants have been proposed in the literature whose goal is the performance improvement either by reducing the number of points or the required iterations or even enhancing the complexity of the most expensive phase: the closest neighbor search. In spite of decreasing its complexity, some of the variants tend to have a negative impact on the final registration precision or the convergence domain thus limiting the possible application scenarios. The goal of this work is the improvement of the algorithm’s computational cost so that a wider range of computationally demanding problems from among the ones described before can be addressed. For that purpose, an experimental and mathematical convergence analysis and validation of point-to-point distance metrics has been performed taking into account those distances with lower computational cost than the Euclidean one, which is used as the de facto standard for the algorithm’s implementations in the literature. In that analysis, the functioning of the algorithm in diverse topological spaces, characterized by different metrics, has been studied to check the convergence, efficacy and cost of the method in order to determine the one which offers the best results. Given that the distance calculation represents a significant part of the whole set of computations performed by the algorithm, it is expected that any reduction of that operation affects significantly and positively the overall performance of the method. As a result, a performance improvement has been achieved by the application of those reduced cost metrics whose quality in terms of convergence and error has been analyzed and validated experimentally as comparable with respect to the Euclidean distance using a heterogeneous set of objects, scenarios and initial situations.