Abdominal Imaging. Computational and Clinical Applications

Mathematical modeling has the potential to assist radiofrequency ablation (RFA) of tumors as it enables prediction of the extent of ablation. However, the accuracy of the simulation is challenged by the material properties since they are patient-specific, temperature and space dependent. In this paper, we present a framework for patient-specific radiofrequency ablation modeling of multiple lesions in the case of metastatic diseases. The proposed forward model is based upon a computational model of heat diffusion, cellular necrosis and blood flow through vessels and liver which relies on patient images. We estimate the most sensitive material parameters, those need to be personalized from the available clinical imaging and data. The selected parameters are then estimated using inverse modeling such that the point-to-mesh distance between the computed necrotic area and observed lesions is minimized. Based on the personalized parameters, the ablation of the remaining lesions are predicted. The framework is applied to a dataset of seven lesions from three patients including pre- and post-operative CT images. In each case, the parameters were estimated on one tumor and RFA is simulated on the other tumor(s) using these personalized parameters, assuming the parameters to be spatially invariant within the same patient. Results showed significantly good correlation between predicted and actual ablation extent (average point-to-mesh errors of 4.03 mm).

The identification and localisation of a focal liver lesion (FLL) in Contrast-Enhanced Ultrasound (CEUS) video sequences is crucial for liver cancer diagnosis, treatment planning and follow-up management. Currently, localisation and classification of FLLs between benign and malignant cases in CEUS are routinely performed manually by radiologists, in order to proceed with making a diagnosis, leading to subjective results, prone to misinterpretation and human error. This paper describes a methodology to assist clinicians who regularly perform these tasks, by discharging benign FLL cases and localise potential malignancies in a fully automatic manner by exploiting the perfusion dynamics of a CEUS video. The proposed framework uses local variations of intensity to distinguish between hyper- and hypo-enhancing regions and then analyse their spatial configuration to identify potentially malignant cases. Automatic localisation of the potential malignancy on the image plane is then addressed by clustering, using Expectation-Maximisation for Gaussian Mixture Models. A novel feature that combines description of local dynamic behaviour with spatial proximity is used in this process. Quantitative evaluation, on real clinical data from a retrospective multi-centre study, demonstrates the value of the proposed method.

Delineation of hepatic tumours is challenging in CT due to limited inherent tissue contrast, leading to significant intra-/inter-observer variability. Perfusion CT (pCT) allows quantitative assessment of enhancement patterns in normal and abnormal liver. This study aims to develop a semi-automated perfusion analysis toolkit that classifies hepatic tissue based on perfusion-derived parameters. pCT data from patients with hepatic metastases were used in this study. Tumour motion was minimized through image registration; perfusion parameters were derived and then employed in the training of a machine learning algorithm used to classify hepatic tissue. This method was found to deliver promising results for 10 data sets, with recorded sensitivity and specificity of the tissue classification in the ranges of 0.92–0.99 and 0.98–0.99 respectively. This semi-automated method could be used to analyze response over the treatment course, as it is not based on intensity values.

Dynamic contrast-enhanced MRI (DCE-MRI) data have often been analyzed using classic standard tracer kinetic models that assume a fast-exchange limit (FXL) of water. Recently, it has been demonstrated that deviations from the FXL model occurs when contrast agent arrives at the target tissue. However, no systematic analysis has been reported for the liver tumor with dual blood supply. In this study, we compared kinetic parameter estimates from DCE-MRI in advanced hepatocellular carcinoma that have the same physiological meaning between five different FXL standard dual-input tracer kinetic models and their corresponding water-exchange-modified (WX) versions. Kinetic parameters were estimated by fitting data to analytic solutions of five different FXL models and their WX versions based on a full two-site-exchange model for transcytolemmal water exchange or a full three-site-two-exchange model for transendothelial and transcytolemmal water exchange. Results suggest that parameter values differ substantially between the FXL standard and WX tracer kinetic models, indicating that DCE-MRI data are water-exchange-sensitive.

Previous texture analysis studies of liver CT images have shown the potential to achieve hepatic malignancy or predict overall survival (OS). However, to date, most studies have mainly focused on assessing texture features of the non-contrast CT or portal-phase image in the dynamic contrast-enhanced CT sequence. The aim of this study was to quantify texture features of physiologically-based kinetic parametric images, and to develop prognostic kinetic textural biomarkers for 1-year survival (1YS) and OS in patients with advanced hepatocellular carcinoma (HCC) following antiangiogenic therapy in comparison among five different tracer kinetic models. Mean, standard deviation, coefficient of variation, skewness, and kurtosis of the pixel distribution histogram within HCC were derived from baseline first-pass perfusion CT parameters. Results suggest that texture analysis of kinetic parametric images can provide better chances of finding effective prognostic biomarkers for the prediction of survival than a mean value analysis alone.

The modeling of tracer kinetics with use of low-temporal-resolution data is of central importance for patient dose reduction in dynamic contrast-enhanced CT (DCE-CT) study. Tracer kinetic models of the liver vary according to the physiologic assumptions imposed on the model, and they can substantially differ in the ways how the input for blood supply and tissue compartments are modeled. In this study, single-input flow-limited (FL), Tofts-Kety (TK), extended TK (ETK), Hayton-Brady (HB), two compartment exchange (2CX), and adiabatic approximation to the tissue homogeneity (AATH) models were applied to the analysis of liver 4-phase DCE-CT data with fully continuous-time parameter formulation, including the bolus arrival time. The bolus arrival time for the 2CX and AATH models was described by modifying the vascular transport operator theory. Initial results indicate that single-input tracer kinetic modeling is feasible for distinguishing between hepatocellular carcinoma and normal liver parenchyma.

The delineation of tumor boundaries is an essential task for the diagnosis and follow-up of liver cancer. However accurate segmentation remains challenging due to tissue inhomogeneity and high variability in tumor appearance. In this paper, we propose a semi-automatic liver tumor segmentation method that combines a deformable model with a machine learning mechanism. More precisely, segmentation is performed by an MRF-based omni-directional deformable surface model that uses image information together with a two-class (tumor, non-tumor) voxel classification map. The classification map is produced by a kernel SVM classifier trained on texture features, as well as intensity mean and variance. The segmentation method is validated on a metastatic tumor dataset consisting of 27 tumors across a set of abdominal CT images, using leave-one-out validation. Compared to pure voxel and gradient approaches, our method achieves better performance in terms of mean distance and Dice scores on the group of 27 liver tumors and can deal with highly pathological cases.

We present a new method for automated placement of a navigator tracker for MRI liver scans. The tracker is used for the navigator echo sequence. It localizes the region acquiring the MR signal to monitor respiratory motion. Accurate placement of the tracker at the boundary between the lung and liver while observing scout images is a complicated task for operators, adversely affecting their workflow. Our proposed method uses ensemble-based classifiers to detect pixels and a right landmark on the upper edge of the liver, following identification of the area containing the edge pixels in the superior/inferior direction. The navigator tracker location is computed from the peak location of the upward convex shape formed by the edge pixels after fitting to a quadratic function. Our method placed the navigator tracker with a mean error of 6.79 mm for the desired location in 126 volunteers. A computational time was approximately 3 s.

Introduction: We compared pancreatic blood flow (PBF) measured by perfusion CT with deconvolution method and Laser Doppler Flow (LDF) to determine whether uncollected contrast material from the pancreas affects measurement of perfusion CT, using porcine pancreatitis model (n = 7). Materials and Methods: The pancreas was divided into head and tail. We arranged that PBF of tail was circulated by single input/output vessel. Ischemic pancreatitis was induced to tail, shutting off input/output vessel for 30 min. PBF was measured both by perfusion CT with deconvolution method and LDF, during the overall course. We calculated uncollected ratio of contrast material, assuming area-under-curve differences between the input and output vessels correspond to uncollected. Results: tail PBF measured by perfusion CT with deconvolution was significantly correlated with LDF, despite of pancreatitis (P < 0.05). The uncollected ratio was 19 %. Discussion: Although the uncollected ratio was not few, result of perfusion CT was significantly related with LDF.

Organ segmentation is a prerequisite for a computer-aided diagnosis (CAD) system to detect pathologies and perform quantitative analysis. For anatomically high-variability abdominal organs such as the pancreas, previous segmentation works report low accuracies when comparing to organs like the heart or liver. In this paper, a fully-automated bottom-up method is presented for pancreas segmentation, using abdominal computed tomography (CT) scans. The method is based on a hierarchical two-tiered information propagation by classifying image patches. It labels superpixels as pancreas or not via pooling patch-level confidences on 2D CT slices over-segmented by the Simple Linear Iterative Clustering approach. A supervised random forest (RF) classifier is trained on the patch level and a two-level cascade of RFs is applied at the superpixel level, coupled with multi-channel feature extraction, respectively. On six-fold cross-validation using 80 patient CT volumes, we achieved 68.8 % Dice coefficient and 57.2 % Jaccard Index, comparable to or slightly better than published state-of-the-art methods.

Diffusion Weighted imaging (DWI) of the body provides important information about the physiological and microstructural properties of tissues and has great potential for imaging inflammatory activity and improve diagnosis and follow up of Crohn’s disease. The two main challenges for DWI are the lack of realistic signal decay models of heterogeneous diffusion and inherently low signal-to-noise ratio (SNR), which makes robust parameter estimation challenging. Increasing the SNR requires long scan times that are not clinically practical. In this work, to address both challenges, we propose a novel Spatially-constrained Probability distribution model of incoherent Motion (SPIM) of water molecules. This model is composed of a probability model of diffusion that we propose to account for the heterogeneity of incoherent motion within multiple compartment tissue microenvironments in each voxel and a spatial homogeneity prior proposed by Freiman et al. for robust parameter estimation. We evaluated the performance of proposed SPIM model in both simulated and in-vivo DWI data from 5 healthy and 24 Crohn’s disease subjects. SPIM model substantially reduced parameter estimation errors, with a reduction of \(35\,\%\) for perfusion and \(7\,\%\) for perfusion fraction and \(4\,\%\) for diffusion parameters. Coefficient of variation of estimated parameters decreased using SPIM compared to simple bi-exponential signal decay model, which indicates an increase in robustness. Parameters estimated using SPIM model better discriminated enhancing and non-enhancing stages of Crohn’s disease.

Crohn’s disease (CD) is a chronic inflammatory bowel disease which can be visualized by magnetic resonance imaging (MRI). For CD grading, several non-invasive MRI based severity scores are known, most prominent the MaRIA and AIS. As these scores rely on manual MRI readings for individual bowel segments by trained radiologists, automated MRI assessment has been more and more focused in recent research. We show on a dataset of 27 CD patients that semi-automatically measured bowel wall thickness (ABWT) and dynamic contrast enhancement (DCE) completely outperform manual scorings: the segmental correlation to the Crohn’s Disease Endoscopic Index of Severity (CDEIS) of ABWT and DCE is significantly higher (r = .78) than that of MaRIA (r = .45) or AIS (r = .51). Also on a per-patient basis, the models with ABWT and DCE show significantly higher correlation (r = .69) to global CDEIS than MaRIA (r = .46).

We propose a graph cut (GC) based approach for combining annotations from multiple experts and segmenting Crohns disease (CD) tissues in magnetic resonance (MR) images. Random forest (RF) based semi supervised learning (SSL) predicts missing expert labels while a novel self consistency (SC) score quantifies the reliability of each expert label and also serves as the penalty cost in a second order Markov random field (MRF) cost function. The final consensus label is obtained by GC optimization. Experimental results on synthetic images and real CD patient data show our final segmentation to be more accurate than those obtained by competing methods. It also highlights the effectiveness of SC score in quantifying expert reliability and accuracy of SSL in predicting missing labels.

Optical colonoscopy is the preferred method for colon cancer screening and prevention. The goal of colonoscopy is to find and remove colonic polyps, precursors to colon cancer. However, colonoscopy is not a perfect procedure. Recent clinical studies report a significant polyp miss due to insufficient quality of colonoscopy. To complicate the problem, the existing guidelines for a “good” colonoscopy, such as maintaining a minimum withdrawal time of 6 min, are not adequate to guarantee the quality of colonoscopy. In response to this problem, this paper presents a method that can objectively measure the quality of an examination by assessing the informativeness of the corresponding colonoscopy images. By assigning a normalized quality score to each colonoscopy frame, our method can detect the onset of a hasty examination and encourage a more diligent procedure. The computed scores can also be averaged and reported as the overall quality of colonoscopy for quality monitoring purposes. Our experiments reveal that the suggested method achieves higher sensitivity and specificity to non-informative frames than the existing image quality assessment methods for colonoscopy videos.

In CT colonography (CTC), orally administered positive-contrast fecal-tagging agents can cause artificial elevation of the observed radiodensity of adjacent soft tissue. Such pseudo-enhancement makes it challenging to differentiate polyps and folds reliably from tagged materials, and it is also present in dual-energy CTC (DE-CTC). We developed a method that corrects for pseudo-enhancement on DE-CTC images without distorting the dual-energy information contained in the data. A pilot study was performed to evaluate the effect of the method visually and quantitatively by use of clinical non-cathartic low-dose DE-CTC data from 10 patients including 13 polyps covered partially or completely by iodine-based fecal tagging. The results indicate that the proposed method can be used to reduce the pseudo-enhancement distortion of DE-CTC images without losing material-specific dual-energy information. The method has potential application in improving the accuracy of automated image-processing applications, such as computer-aided detection and virtual bowel cleansing in CTC.

In CT colonography, orally administered positive-contrast fecal-tagging agents are used for differentiating residual fluid and feces from true lesions. However, the presence of high-density tagging agent in the colon can introduce erroneous artifacts, such as local pseudo-enhancement and beam-hardening, on the reconstructed CT images, thereby complicating reliable detection of soft-tissue lesions. In dual-energy CT colonography, such image artifacts can be reduced by the calculation of virtual monochromatic CT images, which provide more accurate quantitative attenuation measurements than conventional single-energy CT colonography. In practice, however, virtual monochromatic images may still contain some pseudo-enhancement artifacts, and efforts to minimize radiation dose may enhance such artifacts. In this study, we evaluated the effect of image-based pseudo-enhancement post-correction on virtual monochromatic images in standard-dose and low-dose dual-energy CT colonography. The mean CT values of the virtual monochromatic standard-dose CT images of 51 polyps and those of the virtual monochromatic low-dose CT images of 20 polyps were measured without and with the pseudo-enhancement correction. Statistically significant differences were observed between uncorrected and pseudo-enhancement-corrected images of polyps covered by fecal tagging in standard-dose CT (p < 0.001) and in low-dose CT (p < 0.05). The results indicate that image-based pseudo-enhancement post-correction can be useful for optimizing the performance of image-processing applications in virtual monochromatic CT colonography.

To segment the boundary of both inner and outer colon wall is of much significance for colonic polyps detection in computed tomographic colonography (CTC). However, the low contrast of CT attenuation values between colon wall and the surrounding tissues limits many traditional algorithms to achieve this task. Moreover, when sticking presents between two colon walls, the task turns to be much more complicated and the threshold level set segmentation method may fail in this situation. In view of this, we present a minimum surface overlay model to extract the inner wall in this paper. Combined with the superposition model, we are able to depict the outer wall of colon in a natural way. We validated the proposed algorithm based on 60 CTC datasets. Compared with the golden standard (the manual drawing by experts), the new presented method achieved with more than 95 % overlapping coverage rate (OCR).

This paper proposes a complete framework for 3D colon segmentation, including detection of its outer walls. Outer wall detection is a challenging problem due to its poor contrast with other structures appearing in the abdominal scans, especially small bowels and other fatty structures. Missing outer walls could severely affect detection of polyps; indicators of colon cancer. A completely automated framework was developed based on level sets as an initial phase of segmentation to extract the lumen. This phase is followed by discarding non-colonic structures. Outer walls of the colon are then detected, and finally the 3d convex active contour model is used to combine the results of both lumen and outer walls. The technique was tested on 30 colon computed tomography (CT) scans and proved effective in both outer walls and polyp detection. The accuracy of the proposed framework is up to 98.94 %.

This paper proposes a new visualization technique for tubular shape visualization called one-sided transparency (OST). The technique effectively removes the exterior face of a surface, making it transparent, while keeping the interior opaque for viewing. OST is particularly useful in virtual colonoscopy, giving superior visibility coverage with reduced data memory requirements compared to state-of-the-art techniques. The technique achieved surface visibility coverage of up to \( 99.5 \pm 0.2\,\% \) when applied on 5 clinical sets using a fly-over (FO) approach. OST also lends improvements to other VC approaches that depend on centerline extraction of the colon volume.

Spatial regularization is indispensable in image registration to avoid both physically implausible displacement fields and potential local minima in optimization methods. Typical \(\ell _2\)-regularization is incapable of correctly recovering non-smooth displacement fields, such as at sliding organ boundaries during time-series of breathing motion. In this paper, Total Variation (TV) regularization is used to allow for accurate registration near such boundaries. We propose a novel formulation of TV-regularization for parametric displacement fields and introduce an efficient and general numerical solution scheme using the Alternating Directions Method of Multipliers (ADMM). Our method has been evaluated on two public datasets of 4D CT lung images as well as a dataset of 4D MR liver images, demonstrating accurate registrations both inside and outside moving organs. The target registration error of our method is 2.56 mm on average in the liver dataset, which indicates an improvement of over 24 % in comparison to other published methods.

We propose a bilinear model of respiratory organ motion. The advantages of classical statistical shape modelling are combined with a preconditioned trajectory basis for separately modelling the shape and motion components of the data. The separation of a linear basis into bilinear form leads to a more compact representation of the underlying physical process and the resulting model respects the temporal regularity within the training data, which is an important property for modelling quasi-periodic data. Bilinear modelling is combined with a Bayesian reconstruction algorithm for sparse data under observation noise. By applying the model to liver motion data, we show that our bilinear formulation of respiratory motion is significantly more parsimonious and can even outperform linear PCA-based models.

Tube detection filters (TDFs) are useful for segmentation and centerline extraction of tubular structures such as blood vessels and airways in medical images. Most TDFs assume that the cross-sectional profile of the tubular structure is circular. This assumption is not always correct, for instance in the case of abdominal aortic aneurysms (AAAs). Another problem with several TDFs is that they give a false response at strong edges. In this paper, a new TDF is proposed and compared to other TDFs on synthetic and clinical datasets. The results show that the proposed TDF is able to detect large non-circular tubular structures such as AAAs and avoid false positives.

3D reconstruction from monocular laparoscopic sequences is a significant challenge since illumination changes and specular reflections are present in the majority of the images. In this paper we present a total variation based approach to dense reconstruction from monocular laparoscopic sequences. The method deals with specular reflections and makes use of photometric invariants to gain robustness to illumination changes. The method achieves a median reconstruction accuracy of 0.89 mm in an evaluation of 277 reconstructions of cirrhotic liver phantoms. The nodular structure of the liver is of interest for macroscopic analysis of liver cirrhosis and clinical diagnosis.

To find an effective way to quantitatively analyze the thickness variation of human bladder wall under different states, in this paper, we proposed a novel pipeline for thickness measurement, analysis, and mapping of bladder wall based on T2-weighted MRI images. The pipeline includes major steps of data acquisition, automatic segmentation of bladder wall, 3D thickness calculation, thickness normalization, and standardized bladder shape mapping. Based on the proposed pipeline, 20 datasets including 10 patients and 10 volunteers were used to explore the distribution pattern of wall thickness and find the difference between cancerous tissue and normal bladder wall. The results demonstrated the potential of wall thickness as a good indicator of bladder abnormalities, indicating its possible use in lesion detection on the bladder wall.

In interventional radiology, fluoroscopy is used to determine the position of the catheter inserted into a vessel. However, since vessels cannot be identified in fluoroscopic images, it is difficult to forward a catheter to a target region only with fluoroscopy. Thus, angiography and preoperative computed tomography (CT) images are used for the clinical purpose. CT images are useful for understanding the three-dimensional (3D) structure, but guidance of catheter is still difficult since the relationship between CT images and the fluoroscopic image is unclear. In this study, we developed a method for 3D representation of deformed vessels in CT images using an angiographic image acquired preoperatively under natural respiration and preoperative CT images. We implemented the registration algorithm and applied it to patient data. As a result, we confirmed that the vessels in CT images were correctly deformed, and a position error was two pixels in the median value.

Previously, we proposed a method for reconstructing 4D-MRI of thoracoabdominal organs that can visualize and quantify the three-dimensional dynamics of organs due to respiration. However, the data acquisition time of the method is long, say, 30 min. In this study, we assume an interleave acquisition of images with a smaller number of the encoding in the k-space to shorten the data acquisition time. We also propose to use a reconstruction technique named k-t SLR that utilizes sparse and low rank structures of the data matrix to avoid image degradation due to the small number of data acquisition. We performed a simulation experiment where we regarded 4D-MR images by our previous method as ideal images, generated down sampled data in k-space, and applied k-t SLR reconstruction to those data. We evaluated the resultant images from three viewpoints and confirmed that the combination of fast data collection with a small number of encoding and the subsequent k-t SLR reconstruction can produce high quality MR images.

This study presents a homogeneous region localization technique combined with automated evaluation of radiation dose-dependent contrast-to-noise ratio in dual energy abdominal CT. Patient body size was calculated using region growing segmentation and used to estimate size-specific dose estimate, and contrast-to-noise ratio are automatically evaluated by using proposed technique. Contrast-to-noise ratio turned out to be similar between low and high tube potential for both pre- and post-contrast phases, while radiation dose is remarkably lower at 80 kVp. Low tube potential can be recommended to reduce radiation dose while maintaining contrast-to-noise.

In this work we presented the technology for high-resolution efficient numerical modeling of bioimpedance measurements. This technology includes 3D image segmentation, adaptive unstructured tetrahedral mesh generation, finite-element discretization, and the analysis of simulation data. High resolution anatomically correct model based on Visible Human Project data was created. Sensitivity field distributions for a Kubicek-like scheme, as well as two eight-electrode segmental torso measurement schemes were computed and compared. All presented methods and techniques are well-known and are implemented in several open-source packages.