Non-calcified plaque-based coronary stenosis grading in contrast enhanced CT.

BACKGROUND: The high mortality rate associated with coronary heart disease has led to state-of-the-art non-invasive methods for cardiac diagnosis including computed tomography and magnetic resonance imaging. However, stenosis computation and clinical assessment of non-calcified plaques has been very challenging due to their ambiguous intensity response in CT i.e. a significant overlap with surrounding muscle tissues and blood. Accordingly, this research presents an approach for computation of coronary stenosis by investigating cross-sectional lumen behaviour along the length of 3D coronary segments. METHODS: Non-calcified plaques are characterized by comparatively lower-intensity values with respect to the surrounding. Accordingly, segment-wise orthogonal volume was reconstructed in 3D space using the segmented coronary tree. Subsequently, the cross sectional volumetric data was investigated using proposed CNN-based plaque quantification model and subsequent stenosis grading in clinical context was performed. In the last step, plaque-affected orthogonal volume was further investigated by comparing vessel-wall thickness and lumen area obstruction w.r.t. expert-based annotations to validate the stenosis grading performance of model. RESULTS: The experimental data consists of clinical CT images obtained from the Rotterdam CT repository leading to 600 coronary segments and subsequent 15786 cross-sectional images. According to the results, the proposed method quantified coronary vessel stenosis i.e. severity of the non-calcified plaque with an overall accuracy of 83%. Moreover, for individual grading, the proposed model show promising results with accuracy equal to 86%, 90% and 79% respectively for severe, moderate and mild stenosis. The stenosis grading performance of the proposed model was further validated by performing lumen-area versus wall-thickness analysis as per annotations of manual experts. The statistical results for lumen area analysis precisely correlates with the quantification performance of the model with a mean deviation of 5% only. CONCLUSION: The overall results demonstrates capability of the proposed model to grade the vessel stenosis with reasonable accuracy and precision equivalent to human experts.

Prediction of atrial fibrillation and stroke using machine learning models in UK Biobank.

Objective: Atrial fibrillation (AF) is the most common cardiac arrythmia, and it is associated with increased risk for ischemic stroke, which is underestimated, as AF can be asymptomatic. The aim of this study was to develop optimal ML models for prediction of AF in the population, and secondly for ischemic stroke in AF patients. Methods: To develop ML models for prediction of 1) AF in the general population and 2) ischemic stroke in patients with AF we constructed XGBoost, LightGBM, Random Forest, Deep Neural Network, Support Vector Machine and Lasso penalised logistic regression models using UK-Biobank's extensive real-world clinical data, questionnaires, as well as biochemical and genetic data, and their predictive performances were compared. Ranking and contribution of the different features was assessed by SHapley Additive exPlanations (SHAP) analysis. The clinical tool CHA2DS2-VASc for prediction of ischemic stroke among AF patients, was used for comparison to the best performing ML model. Findings: The best performing model for AF prediction was LightGBM, with an area-under-the-roc-curve (AUROC) of 0.729 (95% confidence intervals (CI): 0.719, 0.738). The best performing model for ischemic stroke prediction in AF patients was XGBoost with AUROC of 0.631 (95% CI: 0.604, 0.657). The improved AUROC in the XGBoost model compared to CHA2DS2-VASc was statistically significant based on DeLong's test (p-value = 2.20E-06). In addition, the SHAP analysis showed that several peripheral blood biomarkers (e.g. creatinine, glycated haemoglobin, monocytes) were associated with ischemic stroke, which are not considered by CHA2DS2-VASc. Implications: The best performing ML models presented have the potential for clinical use, but further validation in independent studies is required. Our results endorse the incorporation of some routinely measured blood biomarkers for ischemic stroke prediction in AF patients.

Applications of machine and deep learning to thyroid cytology and histopathology: a review.

This review synthesises past research into how machine and deep learning can improve the cyto- and histopathology processing pipelines for thyroid cancer diagnosis. The current gold-standard preoperative technique of fine-needle aspiration cytology has high interobserver variability, often returns indeterminate samples and cannot reliably identify some pathologies; histopathology analysis addresses these issues to an extent, but it requires surgical resection of the suspicious lesions so cannot influence preoperative decisions. Motivated by these issues, as well as by the chronic shortage of trained pathologists, much research has been conducted into how artificial intelligence could improve current pipelines and reduce the pressure on clinicians. Many past studies have indicated the significant potential of automated image analysis in classifying thyroid lesions, particularly for those of papillary thyroid carcinoma, but these have generally been retrospective, so questions remain about both the practical efficacy of these automated tools and the realities of integrating them into clinical workflows. Furthermore, the nature of thyroid lesion classification is significantly more nuanced in practice than many current studies have addressed, and this, along with the heterogeneous nature of processing pipelines in different laboratories, means that no solution has proven itself robust enough for clinical adoption. There are, therefore, multiple avenues for future research: examine the practical implementation of these algorithms as pathologist decision-support systems; improve interpretability, which is necessary for developing trust with clinicians and regulators; and investigate multiclassification on diverse multicentre datasets, aiming for methods that demonstrate high performance in a process- and equipment-agnostic manner.

DeepFMRI: End-to-end deep learning for functional connectivity and classification of ADHD using fMRI.

BACKGROUND: Resting state fMRI has emerged as a popular neuroimaging method for automated recognition and classification of brain disorders. Attention Deficit Hyperactivity Disorder (ADHD) is one of the most common brain disorders affecting young children, yet its underlying mechanism is not completely understood and its diagnosis is mainly dependent on behaviour analysis. NEW METHOD: In this paper, we propose an end-to-end deep learning architecture to diagnose ADHD. Our aim is to (1) automatically classify a subject as ADHD or healthy control, and (2) demonstrate the importance of functional connectivity to increase classification accuracy and provide interpretable results. The proposed method, called DeepFMRI, is comprised of three sequential networks, namely (1) a feature extractor, (2) a functional connectivity network, and (3) a classification network. The model takes fMRI pre-processed time-series signals as input and outputs a diagnosis, and is trained end-to-end using back-propagation. RESULTS: Experimental results on the publicly available ADHD-200 dataset demonstrate that this innovative method outperforms previous state-of-the-art. Different imaging sites contributed the data to the ADHD-200 dataset. For the New York University imaging site, our proposed method was able to achieve classification accuracy of 73.1% (specificity 91.6%, sensitivity 65.5%). COMPARISON WITH EXISTING METHODS: In this work, we propose a novel end-to-end deep learning method incorporating functional connectivity for the classification of ADHD. To the best of our knowledge, this has not been explored by existing studies. CONCLUSIONS: The results suggest that the proposed end-to-end deep learning architecture achieves better performance as compared to the other state-of-the-art methods. The findings suggest that the frontal lobe contains the most discriminative power towards the classification of ADHD.

Macrosight: A Novel Framework to Analyze the Shape and Movement of Interacting Macrophages Using Matlab®.

This paper presents a novel software framework, called macrosight, which incorporates routines to detect, track, and analyze the shape and movement of objects, with special emphasis on macrophages. The key feature presented in macrosight consists of an algorithm to assess the changes of direction derived from cell-cell contact, where an interaction is assumed to occur. The main biological motivation is the determination of certain cell interactions influencing cell migration. Thus, the main objective of this work is to provide insights into the notion that interactions between cell structures cause a change in orientation. Macrosight analyzes the change of direction of cells before and after they come in contact with another cell. Interactions are determined when the cells overlap and form clumps of two or more cells. The framework integrates a segmentation technique capable of detecting overlapping cells and a tracking framework into a tool for the analysis of the trajectories of cells before and after they overlap. Preliminary results show promise into the analysis and the hypothesis proposed, and lays the groundwork for further developments. The extensive experimentation and data analysis show, with statistical significance, that under certain conditions, the movement changes before and after an interaction are different from movement in controlled cases.

Distinct patterns of variation in the distribution of knee pain.

The patient's expression of pain using digital-body maps expands analytic opportunities for exploring the spatial variation of bodily pain. A common knee pain condition in adolescents and adults is patellofemoral pain (PFP) and recently PFP was shown to be characterized by a heterogeneous distribution of pain. Whether there are important patterns in these distributions remains unclear. This pioneering study assesses the spatial variation of pain using principal component analysis and a clustering approach. Detailed digital-body maps of knee pain were drawn by 299 PFP patients of mixed sex, age, and pain severity. Three pain distribution patterns emerged resembling an Anchor, Hook, and an Ovate shape on and around the patella. The variations in pain distribution were independent of sex, age, and pain intensity. Bilateral pain associated with a longer duration of pain and the majority characterized by the Hook and Ovate pain distributions. Bilateral and/or symmetrical pain between the left and right knees may represent symptoms associated with longstanding PFP. The distinct patterns of pain location and area suggest specific underlying structures cannot be ruled out as important drivers, although central neuronal mechanisms possibly exemplified by the symmetrical representation of pain may play a role in individuals with longstanding symptoms.

DAGAN: Deep De-Aliasing Generative Adversarial Networks for Fast Compressed Sensing MRI Reconstruction.

Compressed sensing magnetic resonance imaging (CS-MRI) enables fast acquisition, which is highly desirable for numerous clinical applications. This can not only reduce the scanning cost and ease patient burden, but also potentially reduce motion artefacts and the effect of contrast washout, thus yielding better image quality. Different from parallel imaging-based fast MRI, which utilizes multiple coils to simultaneously receive MR signals, CS-MRI breaks the Nyquist-Shannon sampling barrier to reconstruct MRI images with much less required raw data. This paper provides a deep learning-based strategy for reconstruction of CS-MRI, and bridges a substantial gap between conventional non-learning methods working only on data from a single image, and prior knowledge from large training data sets. In particular, a novel conditional Generative Adversarial Networks-based model (DAGAN)-based model is proposed to reconstruct CS-MRI. In our DAGAN architecture, we have designed a refinement learning method to stabilize our U-Net based generator, which provides an end-to-end network to reduce aliasing artefacts. To better preserve texture and edges in the reconstruction, we have coupled the adversarial loss with an innovative content loss. In addition, we incorporate frequency-domain information to enforce similarity in both the image and frequency domains. We have performed comprehensive comparison studies with both conventional CS-MRI reconstruction methods and newly investigated deep learning approaches. Compared with these methods, our DAGAN method provides superior reconstruction with preserved perceptual image details. Furthermore, each image is reconstructed in about 5 ms, which is suitable for real-time processing.

Fully automatic cervical vertebrae segmentation framework for X-ray images.

The cervical spine is a highly flexible anatomy and therefore vulnerable to injuries. Unfortunately, a large number of injuries in lateral cervical X-ray images remain undiagnosed due to human errors. Computer-aided injury detection has the potential to reduce the risk of misdiagnosis. Towards building an automatic injury detection system, in this paper, we propose a deep learning-based fully automatic framework for segmentation of cervical vertebrae in X-ray images. The framework first localizes the spinal region in the image using a deep fully convolutional neural network. Then vertebra centers are localized using a novel deep probabilistic spatial regression network. Finally, a novel shape-aware deep segmentation network is used to segment the vertebrae in the image. The framework can take an X-ray image and produce a vertebrae segmentation result without any manual intervention. Each block of the fully automatic framework has been trained on a set of 124 X-ray images and tested on another 172 images, all collected from real-life hospital emergency rooms. A Dice similarity coefficient of 0.84 and a shape error of 1.69 mm have been achieved.

Fully automatic segmentation and objective assessment of atrial scars for long-standing persistent atrial fibrillation patients using late gadolinium-enhanced MRI.

PURPOSE: Atrial fibrillation (AF) is the most common heart rhythm disorder and causes considerable morbidity and mortality, resulting in a large public health burden that is increasing as the population ages. It is associated with atrial fibrosis, the amount and distribution of which can be used to stratify patients and to guide subsequent electrophysiology ablation treatment. Atrial fibrosis may be assessed noninvasively using late gadolinium-enhanced (LGE) magnetic resonance imaging (MRI) where scar tissue is visualized as a region of signal enhancement. However, manual segmentation of the heart chambers and of the atrial scar tissue is time consuming and subject to interoperator variability, particularly as image quality in AF is often poor. In this study, we propose a novel fully automatic pipeline to achieve accurate and objective segmentation of the heart (from MRI Roadmap data) and of scar tissue within the heart (from LGE MRI data) acquired in patients with AF. METHODS: Our fully automatic pipeline uniquely combines: (a) a multiatlas-based whole heart segmentation (MA-WHS) to determine the cardiac anatomy from an MRI Roadmap acquisition which is then mapped to LGE MRI, and (b) a super-pixel and supervised learning based approach to delineate the distribution and extent of atrial scarring in LGE MRI. We compared the accuracy of the automatic analysis to manual ground truth segmentations in 37 patients with persistent long-standing AF. RESULTS: Both our MA-WHS and atrial scarring segmentations showed accurate delineations of cardiac anatomy (mean Dice = 89%) and atrial scarring (mean Dice = 79%), respectively, compared to the established ground truth from manual segmentation. In addition, compared to the ground truth, we obtained 88% segmentation accuracy, with 90% sensitivity and 79% specificity. Receiver operating characteristic analysis achieved an average area under the curve of 0.91. CONCLUSION: Compared with previously studied methods with manual interventions, our innovative pipeline demonstrated comparable results, but was computed fully automatically. The proposed segmentation methods allow LGE MRI to be used as an objective assessment tool for localization, visualization, and quantitation of atrial scarring and to guide ablation treatment.

Fusion of fMRI and non-imaging data for ADHD classification.

Resting state fMRI has emerged as a popular neuroimaging method for automated recognition and classification of different brain disorders. Attention Deficit Hyperactivity Disorder (ADHD) is one of the most common brain disorders affecting young children, yet its underlying mechanism is not completely understood and its diagnosis is mainly dependent on behavior analysis. This paper addresses the problem of classification of ADHD based on resting state fMRI and proposes a machine learning framework with integration of non-imaging data with imaging data to investigate functional connectivity alterations between ADHD and control subjects (not diagnosed with ADHD). Our aim is to apply computational techniques to (1) automatically classify a subject as ADHD or control, (2) identify differences in functional connectivity of these two groups and (3) evaluate the importance of fusing non-imaging with imaging data for classification. In the first stage of our framework, we determine the functional connectivity of brain regions by grouping brain activity using clustering algorithms. Next, we employ Elastic Net based feature selection to select the most discriminant features from the dense functional brain network and integrate non-imaging data. Finally, a Support Vector Machine classifier is trained to classify ADHD subjects vs. control. The proposed framework was evaluated on a public ADHD-200 dataset, and our results suggest that fusion of non-imaging data improves the performance of the framework. Classification results outperform the state-of-the-art on some subsets of the data.

Framework for detection and localization of coronary non-calcified plaques in cardiac CTA using mean radial profiles.

BACKGROUND AND OBJECTIVE: The high mortality rate associated with coronary heart disease (CHD) has driven intensive research in cardiac imaging and image analysis. The advent of computed tomography angiography (CTA) has turned non-invasive diagnosis of cardiovascular anomalies into reality as calcified coronary plaques can be easily identified due to their high intensity values. However, the detection of non-calcified plaques in CTA is still a challenging problem because of lower intensity values, which are often similar to the nearby blood and muscle tissues. In this work, we propose the use of mean radial profiles for the detection of non-calcified plaques in CTA imagery. METHODS: Accordingly, we computed radial profiles by averaging the image intensity in concentric rings around the vessel centreline in a first stage. In the subsequent stage, an SVM classifier is applied to identify the abnormal coronary segments. For occluded segments, we further propose a derivative-based method to localize the position and length of the plaque inside the segment. RESULTS: A total of 32 CTA volumes were analysed and a detection accuracy of 88.4% with respect to the manual expert was achieved. The plaque localization accuracy was computed using the Dice similarity coefficient and a mean of 83.2% was achieved. CONCLUSION: The consistent performance for multi-vendor, multi-institution data demonstrates the reproducibility of our method across different CTA datasets with a good agreement with manual expert annotations.

A hybrid energy model for region based curve evolution - Application to CTA coronary segmentation.

BACKGROUND AND OBJECTIVE: State-of-the-art medical imaging techniques have enabled non-invasive imaging of the internal organs. However, high volumes of imaging data make manual interpretation and delineation of abnormalities cumbersome for clinicians. These challenges have driven intensive research into efficient medical image segmentation. In this work, we propose a hybrid region-based energy formulation for effective segmentation in computed tomography angiography (CTA) imagery. METHODS: The proposed hybrid energy couples an intensity-based local term with an efficient discontinuity-based global model of the image for optimal segmentation. The segmentation is achieved using a level set formulation due to the computational robustness. After validating the statistical significance of the hybrid energy, we applied the proposed model to solve an important clinical problem of 3D coronary segmentation. An improved seed detection method is used to initialize the level set evolution. Moreover, we employed an auto-correction feature that captures the emerging peripheries during the curve evolution for completeness of the coronary tree. RESULTS: We evaluated the segmentation accuracy of the proposed energy model against the existing techniques in two stages. Qualitative and quantitative results demonstrate the effectiveness of the proposed framework with a consistent mean sensitivity and specificity measures of 80% across the CTA data. Moreover, a high degree of agreement with respect to the inter-observer differences justifies the generalization of the proposed method. CONCLUSIONS: The proposed method is effective to segment the coronary tree from the CTA volume based on hybrid image based energy, which can improve the clinicians ability to detect arterial abnormalities.

Discrete Wavelet Transform-Based Whole-Spectral and Subspectral Analysis for Improved Brain Tumor Clustering Using Single Voxel MR Spectroscopy.

Many approaches have been considered for automatic grading of brain tumors by means of pattern recognition with magnetic resonance spectroscopy (MRS). Providing an improved technique which can assist clinicians in accurately identifying brain tumor grades is our main objective. The proposed technique, which is based on the discrete wavelet transform (DWT) of whole-spectral or subspectral information of key metabolites, combined with unsupervised learning, inspects the separability of the extracted wavelet features from the MRS signal to aid the clustering. In total, we included 134 short echo time single voxel MRS spectra (SV MRS) in our study that cover normal controls, low grade and high grade tumors. The combination of DWT-based whole-spectral or subspectral analysis and unsupervised clustering achieved an overall clustering accuracy of 94.8% and a balanced error rate of 7.8%. To the best of our knowledge, it is the first study using DWT combined with unsupervised learning to cluster brain SV MRS. Instead of dimensionality reduction on SV MRS or feature selection using model fitting, our study provides an alternative method of extracting features to obtain promising clustering results.

Multilabel region classification and semantic linking for colon segmentation in CT colonography.

Accurate and automatic colon segmentation from CT images is a crucial step of many clinical applications in CT colonography, including computer-aided detection (CAD) of colon polyps, 3-D virtual flythrough of the colon, and prone/supine registration. However, the existence of adjacent air-filled organs such as the lung, stomach, and small intestine, and the collapse of the colon due to poor insufflation, render accurate segmentation of the colon a difficult problem. Extra-colonic components can be categorized into two types based on their 3-D connection to the colon: detached and attached extracolonic components (DEC and AEC, respectively). In this paper, we propose graph inference methods to remove extracolonic components to achieve a high quality segmentation. We first decompose each 3-D air-filled object into a set of 3-D regions. A classifier trained with region-level features can be used to identify the colon regions from noncolon regions. After removing obvious DEC, we remove the remaining DEC by modeling the global anatomic structure with an a priori topological constraint and solving a graph inference problem using semantic information provided by a multiclass classifier. Finally, we remove AEC by modeling regions within each 3-D object with a hierarchical conditional random field, solved by graph cut. Experimental results demonstrate that our method outperforms a purely discriminative learning method in detecting true colon regions, while decreasing extra-colonic components in challenging clinical data that includes collapsed cases.

Endoluminal surface registration for CT colonography using haustral fold matching.

Computed Tomographic (CT) colonography is a technique used for the detection of bowel cancer or potentially precancerous polyps. The procedure is performed routinely with the patient both prone and supine to differentiate fixed colonic pathology from mobile faecal residue. Matching corresponding locations is difficult and time consuming for radiologists due to colonic deformations that occur during patient repositioning. We propose a novel method to establish correspondence between the two acquisitions automatically. The problem is first simplified by detecting haustral folds using a graph cut method applied to a curvature-based metric applied to a surface mesh generated from segmentation of the colonic lumen. A virtual camera is used to create a set of images that provide a metric for matching pairs of folds between the prone and supine acquisitions. Image patches are generated at the fold positions using depth map renderings of the endoluminal surface and optimised by performing a virtual camera registration over a restricted set of degrees of freedom. The intensity difference between image pairs, along with additional neighbourhood information to enforce geometric constraints over a 2D parameterisation of the 3D space, are used as unary and pair-wise costs respectively, and included in a Markov Random Field (MRF) model to estimate the maximum a posteriori fold labelling assignment. The method achieved fold matching accuracy of 96.0% and 96.1% in patient cases with and without local colonic collapse. Moreover, it improved upon an existing surface-based registration algorithm by providing an initialisation. The set of landmark correspondences is used to non-rigidly transform a 2D source image derived from a conformal mapping process on the 3D endoluminal surface mesh. This achieves full surface correspondence between prone and supine views and can be further refined with an intensity based registration showing a statistically significant improvement (p<0.001), and decreasing mean error from 11.9 mm to 6.0 mm measured at 1743 reference points from 17 CTC datasets.

CT colonography: external clinical validation of an algorithm for computer-assisted prone and supine registration.

PURPOSE: To perform external validation of a computer-assisted registration algorithm for prone and supine computed tomographic (CT) colonography and to compare the results with those of an existing centerline method. MATERIALS AND METHODS: All contributing centers had institutional review board approval; participants provided informed consent. A validation sample of CT colonographic examinations of 51 patients with 68 polyps (6-55 mm) was selected from a publicly available, HIPAA compliant, anonymized archive. No patients were excluded because of poor preparation or inadequate distension. Corresponding prone and supine polyp coordinates were recorded, and endoluminal surfaces were registered automatically by using a computer algorithm. Two observers independently scored three-dimensional endoluminal polyp registration success. Results were compared with those obtained by using the normalized distance along the colonic centerline (NDACC) method. Pairwise Wilcoxon signed rank tests were used to compare gross registration error and McNemar tests were used to compare polyp conspicuity. RESULTS: Registration was possible in all 51 patients, and 136 paired polyp coordinates were generated (68 polyps) to test the algorithm. Overall mean three-dimensional polyp registration error (mean ± standard deviation, 19.9 mm ± 20.4) was significantly less than that for the NDACC method (mean, 27.4 mm ± 15.1; P = .001). Accuracy was unaffected by colonic segment (P = .76) or luminal collapse (P = .066). During endoluminal review by two observers (272 matching tasks, 68 polyps, prone to supine and supine to prone coordinates), 223 (82%) polyp matches were visible (120° field of view) compared with just 129 (47%) when the NDACC method was used (P < .001). By using multiplanar visualization, 48 (70%) polyps were visible after scrolling ± 15 mm in any multiplanar axis compared with 16 (24%) for NDACC (P < .001). CONCLUSION: Computer-assisted registration is more accurate than the NDACC method for mapping the endoluminal surface and matching the location of polyps in corresponding prone and supine CT colonographic acquisitions.

A robust and efficient approach to detect 3D rectal tubes from CT colonography.

PURPOSE: The rectal tube (RT) is a common source of false positives (FPs) in computer-aided detection (CAD) systems for CT colonography. A robust and efficient detection of RT can improve CAD performance by eliminating such "obvious" FPs and increase radiologists' confidence in CAD. METHODS: In this paper, we present a novel and robust bottom-up approach to detect the RT. Probabilistic models, trained using kernel density estimation on simple low-level features, are employed to rank and select the most likely RT tube candidate on each axial slice. Then, a shape model, robustly estimated using random sample consensus (RANSAC), infers the global RT path from the selected local detections. Subimages around the RT path are projected into a subspace formed from training subimages of the RT. A quadratic discriminant analysis (QDA) provides a classification of a subimage as RT or non-RT based on the projection. Finally, a bottom-top clustering method is proposed to merge the classification predictions together to locate the tip position of the RT. RESULTS: Our method is validated using a diverse database, including data from five hospitals. On a testing data with 21 patients (42 volumes), 99.5% of annotated RT paths have been successfully detected. Evaluated with CAD, 98.4% of FPs caused by the RT have been detected and removed without any loss of sensitivity. CONCLUSIONS: The proposed method demonstrates a high detection rate of the RT path, and when tested in a CAD system, reduces FPs caused by the RT without the loss of sensitivity.

Automatic prone to supine haustral fold matching in CT colonography using a Markov random field model.

CT colonography is routinely performed with the patient prone and supine to differentiate fixed colonic pathology from mobile faecal residue. We propose a novel method to automatically establish correspondence. Haustral folds are detected using a graph cut method applied to a surface curvature-based metric, where image patches are generated using endoluminal CT colonography surface rendering. The intensity difference between image pairs, along with additional neighbourhood information to enforce geometric constraints, are used with a Markov Random Field (MRF) model to estimate the fold labelling assignment. The method achieved fold matching accuracy of 83.1% and 88.5% with and without local colonic collapse. Moreover, it improves an existing surface-based registration algorithm, decreasing mean registration error from 9.7mm to 7.7mm in cases exhibiting collapse.

Registration of the endoluminal surfaces of the colon derived from prone and supine CT colonography.

PURPOSE: Computed tomographic (CT) colonography is a relatively new technique for detecting bowel cancer or potentially precancerous polyps. CT scanning is combined with three-dimensional (3D) image reconstruction to produce a virtual endoluminal representation similar to optical colonoscopy. Because retained fluid and stool can mimic pathology, CT data are acquired with the bowel cleansed and insufflated with gas and patient in both prone and supine positions. Radiologists then match visually endoluminal locations between the two acquisitions in order to determine whether apparent pathology is real or not. This process is hindered by the fact that the colon, essentially a long tube, can undergo considerable deformation between acquisitions. The authors present a novel approach to automatically establish spatial correspondence between prone and supine endoluminal colonic surfaces after surface parameterization, even in the case of local colon collapse. METHODS: The complexity of the registration task was reduced from a 3D to a 2D problem by mapping the surfaces extracted from prone and supine CT colonography onto a cylindrical parameterization. A nonrigid cylindrical registration was then performed to align the full colonic surfaces. The curvature information from the original 3D surfaces was used to determine correspondence. The method can also be applied to cases with regions of local colonic collapse by ignoring the collapsed regions during the registration. RESULTS: Using a development set, suitable parameters were found to constrain the cylindrical registration method. Then, the same registration parameters were applied to a different set of 13 validation cases, consisting of 8 fully distended cases and 5 cases exhibiting multiple colonic collapses. All polyps present were well aligned, with a mean (+/- std. dev.) registration error of 5.7 (+/- 3.4) mm. An additional set of 1175 reference points on haustral folds spread over the full endoluminal colon surfaces resulted in an error of 7.7 (+/- 7.4) mm. Here, 82% of folds were aligned correctly after registration with a further 15% misregistered by just onefold. CONCLUSIONS: The proposed method reduces the 3D registration task to a cylindrical registration representing the endoluminal surface of the colon. Our algorithm uses surface curvature information as a similarity measure to drive registration to compensate for the large colorectal deformations that occur between prone and supine data acquisitions. The method has the potential to both enhance polyp detection and decrease the radiologist's interpretation time.

Automatic Graph Cut Segmentation of Lesions in CT Using Mean Shift Superpixels.

This paper presents a new, automatic method of accurately extracting lesions from CT data. It first determines, at each voxel, a five-dimensional (5D) feature vector that contains intensity, shape index, and 3D spatial location. Then, nonparametric mean shift clustering forms superpixels from these 5D features, resulting in an oversegmentation of the image. Finally, a graph cut algorithm groups the superpixels using a novel energy formulation that incorporates shape, intensity, and spatial features. The mean shift superpixels increase the robustness of the result while reducing the computation time. We assume that the lesion is part spherical, resulting in high shape index values in a part of the lesion. From these spherical subregions, foreground and background seeds for the graph cut segmentation can be automatically obtained. The proposed method has been evaluated on a clinical CT dataset. Visual inspection on different types of lesions (lung nodules and colonic polyps), as well as a quantitative evaluation on 101 solid and 80 GGO nodules, both demonstrate the potential of the proposed method. The joint spatial-intensity-shape features provide a powerful cue for successful segmentation of lesions adjacent to structures of similar intensity but different shape, as well as lesions exhibiting partial volume effect.