Automatic Change Detection in Sparse Repeat CT Scanning.

We describe a new method for the automatic detection of changes in repeat CT scanning with a reduced X-ray radiation dose. We present a theoretical formulation of the automatic change detection problem based on the on-line sparse-view repeat CT scanning dose optimization framework. We prove that the change detection problem is NP-hard and therefore cannot be efficiently solved exactly. We describe a new greedy change detection algorithm that is simple and robust and relies on only two key parameters. We demonstrate that the greedy algorithm accurately detects small, low contrast changes with only 12 scan angles. Our experimental results show that the new algorithm yields a mean changed region recall rate >89% and a mean precision rate >76%. It outperforms both our previous heuristic approach and a thresholding method using a low-dose prior image-constrained compressed sensing (PICCS) reconstruction of the repeat scan. The resulting changed region map may obviate the need for a high-quality repeat scan image when no major changes are detected and may streamline the radiologist's workflow by highlighting the regions of interest.

Automatic segmentation variability estimation with segmentation priors.

PURPOSE: Segmentations produced manually by experts or by algorithms are subject to variability, as they depend on many factors, e.g., the structure of interest, the resolution, contrast and quality of the images, and the expert experience or the algorithmic method. To properly assess the quality of these segmentations, it is thus essential to quantify their variability. However, obtaining reference variability ground truth requires several observers to manually delineate structures, which is time-consuming and impractical. METHODS: We describe a new comprehensive formal framework for segmentation evaluation and variability estimation without ground truth and a generic method for automatic segmentation variability estimation based on segmentation priors and multivariate sensitivity analysis. The method inputs the image scan and a user-validated segmentation of the structure of interest and uses predefined segmentation priors to compute a variability estimation around the given segmentation. The segmentation priors are combined with an integrator function whose sensitivity around the given segmentation is the segmentation variability. RESULTS: We validate our methods with two studies. The first study establishes the reference manual delineation variability. Eleven radiologists with varying levels of expertise manually delineated the contours of liver tumors, lung tumors, kidneys, and brain hematomas on 2,835 delineations from 18 CT scans. The relative delineation volume variability ranges are 51 [-24,+27]% for liver tumors, 56 [-25,+31]% for lung tumors, 25 [-12,+13]% for kidney contours, and 53 [-24,+29]% brain hematomas. The second study compares the estimated segmentation variability results to this reference data. The mean volume variability difference of the delineation is <6%, with a Dice similarity coefficient of >70% with respect to the mean manual delineation variability data. CONCLUSIONS: Reliable segmentation variability estimation with no ground truth enables the establishment of a proper observer variability reference. The segmentation variability should be taken into account when setting reference standards for clinical decisions based on volumetric measurements and when evaluating segmentation algorithms.

A fully automatic end-to-end method for content-based image retrieval of CT scans with similar liver lesion annotations.

PURPOSE: The goal of medical content-based image retrieval (M-CBIR) is to assist radiologists in the decision-making process by retrieving medical cases similar to a given image. One of the key interests of radiologists is lesions and their annotations, since the patient treatment depends on the lesion diagnosis. Therefore, a key feature of M-CBIR systems is the retrieval of scans with the most similar lesion annotations. To be of value, M-CBIR systems should be fully automatic to handle large case databases. METHODS: We present a fully automatic end-to-end method for the retrieval of CT scans with similar liver lesion annotations. The input is a database of abdominal CT scans labeled with liver lesions, a query CT scan, and optionally one radiologist-specified lesion annotation of interest. The output is an ordered list of the database CT scans with the most similar liver lesion annotations. The method starts by automatically segmenting the liver in the scan. It then extracts a histogram-based features vector from the segmented region, learns the features' relative importance, and ranks the database scans according to the relative importance measure. The main advantages of our method are that it fully automates the end-to-end querying process, that it uses simple and efficient techniques that are scalable to large datasets, and that it produces quality retrieval results using an unannotated CT scan. RESULTS: Our experimental results on 9 CT queries on a dataset of 41 volumetric CT scans from the 2014 Image CLEF Liver Annotation Task yield an average retrieval accuracy (Normalized Discounted Cumulative Gain index) of 0.77 and 0.84 without/with annotation, respectively. CONCLUSIONS: Fully automatic end-to-end retrieval of similar cases based on image information alone, rather that on disease diagnosis, may help radiologists to better diagnose liver lesions.

Automatic detection of new tumors and tumor burden evaluation in longitudinal liver CT scan studies.

PURPOSE: Radiological longitudinal follow-up of liver tumors in CT scans is the standard of care for disease progression assessment and for liver tumor therapy. Finding new tumors in the follow-up scan is essential to determine malignancy, to evaluate the total tumor burden, and to determine treatment efficacy. Since new tumors are typically small, they may be missed by examining radiologists. METHODS: We describe a new method for the automatic detection and segmentation of new tumors in longitudinal liver CT studies and for liver tumors burden quantification. Its inputs are the baseline and follow-up CT scans, the baseline tumors delineation, and a tumor appearance prior model. Its outputs are the new tumors segmentations in the follow-up scan, the tumor burden quantification in both scans, and the tumor burden change. Our method is the first comprehensive method that is explicitly designed to find new liver tumors. It integrates information from the scans, the baseline known tumors delineations, and a tumor appearance prior model in the form of a global convolutional neural network classifier. Unlike other deep learning-based methods, it does not require large tagged training sets. RESULTS: Our experimental results on 246 tumors, of which 97 were new tumors, from 37 longitudinal liver CT studies with radiologist approved ground-truth segmentations, yields a true positive new tumors detection rate of 86 versus 72% with stand-alone detection, and a tumor burden volume overlap error of 16%. CONCLUSIONS: New tumors detection and tumor burden volumetry are important for diagnosis and treatment. Our new method enables a simplified radiologist-friendly workflow that is potentially more accurate and reliable than the existing one by automatically and accurately following known tumors and detecting new tumors in the follow-up scan.

Radon Space Dose Optimization in Repeat CT Scanning.

We present a new method for on-line radiation dose optimization in repeat computer tomography (CT) scanning. Our method uses the information of the baseline scan during the repeat scanning to significantly reduce the radiation dose without compromising the repeat scan quality. It automatically registers the patient to the baseline scan using fractional scanning and detects in sinogram space the patient regions where changes have occurred without having to reconstruct the repeat scan image. It scans only these regions in the patient, thereby considerably reducing the necessary radiation dose. It then completes the missing values of the sparsely sampled repeat scan sinogram with those of the fully sampled baseline sinogram in regions where no changes were detected and computes the repeat scan image by standard filtered backprojection reconstruction. Experiments on a patient scan with simulated changes yield a mean recall of 98% using <19% of a full dose. Experiments on real CT scans of an abdomen phantom produce similar results, with a mean recall of 94.5% and only 14.4% of a full dose more than the theoretical optimum. As hardly any changed rays are missed, the reconstructed images are practically indistinguishable from a full dose scan. Our method successfully detects small, low contrast changes and produces an accurate repeat scan reconstruction using three times less radiation than an image space baseline method.

Reduced-Dose Imageless Needle and Patient Tracking in Interventional CT Procedures.

This paper describes a new method for imageless needle and patient tracking in interventional CT procedures based on fractional CT scanning. Our method accurately locates a needle with a spherical marker attached to it at a known distance from the tip with respect to the patient in the CT scanner coordinate frame with online sparse scan sampling and without reconstructing the CT image. The key principle of our method is to detect the needle and attached spherical marker in projection (sinogram) space based on the strongly attenuated X-ray signal due to the metallic composition of the needle and the needle's thin cylindrical geometry, and based on the marker's spherical geometry. A transformation from projection space to physical space uniquely determines the location and orientation of the needle and the needle tip position. Our method works directly in projection space and simultaneously performs patient registration and needle localization for every fractional CT scanning acquisition using the same sparse set of views. We performed registration and needle tip localization in five abdomen phantom scans using a rigid needle, and obtained a voxel-size tip localization error. Our experimental results indicate a voxel-sized deviation of the localization from a comparable method in 3-D image space, with the benefit of allowing X-ray dose reduction via fractional scanning at each localization. This benefit enables more frequent tip localizations during needle insertion for a similar total dose, or a reduced total dose for the same frequency of tip localization.

Sparse 3D Radon Space Rigid Registration of CT Scans: Method and Validation Study.

We present a new method for rigid registration of CT datasets in 3D Radon space based on sparse sampling of scanning projections. The inputs are the two 3D Radon transforms of the CT scans, one densely sampled and the other sparsely sampled (limited number of scan angles/ranges). The output is the rigid transformation that best matches them. The method first finds the best matching between each projection direction vector in the sparse transform and the corresponding direction vector in the dense transform. It then solves a system of linear equations derived from the direction vector pairs (parallel-beam projections) or finds a solution by non-linear optimization (fan-beam and cone-beam projections). Experimental studies show that our method for 3D parallel beam registration outperforms image space registration in terms of convergence range with significantly reduced X-ray dose compared to a full conventional CT scan.

A new method for the automatic retrieval of medical cases based on the RadLex ontology.

PURPOSE: The goal of medical case-based image retrieval (M-CBIR) is to assist radiologists in the clinical decision-making process by finding medical cases in large archives that most resemble a given case. Cases are described by radiology reports comprised of radiological images and textual information on the anatomy and pathology findings. The textual information, when available in standardized terminology, e.g., the RadLex ontology, and used in conjunction with the radiological images, provides a substantial advantage for M-CBIR systems. METHODS: We present a new method for incorporating textual radiological findings from medical case reports in M-CBIR. The input is a database of medical cases, a query case, and the number of desired relevant cases. The output is an ordered list of the most relevant cases in the database. The method is based on a new case formulation, the Augmented RadLex Graph and an Anatomy-Pathology List. It uses a new case relatedness metric [Formula: see text] that prioritizes more specific medical terms in the RadLex tree over less specific ones and that incorporates the length of the query case. RESULTS: An experimental study on 8 CT queries from the 2015 VISCERAL 3D Case Retrieval Challenge database consisting of 1497 volumetric CT scans shows that our method has accuracy rates of 82 and 70% on the first 10 and 30 most relevant cases, respectively, thereby outperforming six other methods. CONCLUSIONS: The increasing amount of medical imaging data acquired in clinical practice constitutes a vast database of untapped diagnostically relevant information. This paper presents a new hybrid approach to retrieving the most relevant medical cases based on textual and image information.

Haptic computer-assisted patient-specific preoperative planning for orthopedic fractures surgery.

PURPOSE: The aim of orthopedic trauma surgery is to restore the anatomy and function of displaced bone fragments to support osteosynthesis. For complex cases, including pelvic bone and multi-fragment femoral neck and distal radius fractures, preoperative planning with a CT scan is indicated. The planning consists of (1) fracture reduction-determining the locations and anatomical sites of origin of the fractured bone fragments and (2) fracture fixation-selecting and placing fixation screws and plates. The current bone fragment manipulation, hardware selection, and positioning processes based on 2D slices and a computer mouse are time-consuming and require a technician. METHODS: We present a novel 3D haptic-based system for patient-specific preoperative planning of orthopedic fracture surgery based on CT scans. The system provides the surgeon with an interactive, intuitive, and comprehensive, planning tool that supports fracture reduction and fixation. Its unique features include: (1) two-hand haptic manipulation of 3D bone fragments and fixation hardware models; (2) 3D stereoscopic visualization and multiple viewing modes; (3) ligaments and pivot motion constraints to facilitate fracture reduction; (4) semiautomatic and automatic fracture reduction modes; and (5) interactive custom fixation plate creation to fit the bone morphology. RESULTS: We evaluate our system with two experimental studies: (1) accuracy and repeatability of manual fracture reduction and (2) accuracy of our automatic virtual bone fracture reduction method. The surgeons achieved a mean accuracy of less than 1 mm for the manual reduction and 1.8 mm (std [Formula: see text] 1.1 mm) for the automatic reduction. CONCLUSION: 3D haptic-based patient-specific preoperative planning of orthopedic fracture surgery from CT scans is useful and accurate and may have significant advantages for evaluating and planning complex fractures surgery.

Tumor burden evaluation in NF1 patients with plexiform neurofibromas in daily clinical practice.

BACKGROUND: Existing volumetric measurements of plexiform neurofibromas (PNs) are time consuming and error prone, as they require delineation of PN boundaries, a procedure that is not practical in the typical clinical setting. The aim of this study is to assess the Plexiform Neurofibroma Instant Segmentation Tool (PNist), a novel semi-automated segmentation program that we developed for PN delineation in a clinical context. PNist was designed to greatly simplify volumetric assessment of PNs through use of an intuitive user interface while providing objectively consistent results with minimal interobserver and intraobserver variabilities in reasonable time. MATERIALS AND METHODS: PNs were measured in 30 magnetic resonance imaging (MRI) scans from 12 patients with neurofibromatosis 1. Volumetric measurements were performed using PNist and compared to a standard semi-automated volumetric method (Analyze 9.0). RESULTS: High correlation was detected between PNist and the semi-automated method (R(2) = 0.996), with a mean volume overlap error of 9.54 % and low intraobserver and interobserver variabilities. The segmentation time required for PNist was 60 % of the time required for Analyze 9.0 (360 versus 900 s, respectively). PNist was also reliable when assessing changes in tumor size over time, compared to the existing commercial method. CONCLUSIONS: Our study suggests that the new PNist method is accurate, intuitive, and less time consuming for PN segmentation compared to existing commercial volumetric methods. The workflow is simple and user-friendly, making it an important clinical tool to be used by radiologists, neurologists and neurosurgeons on a daily basis, helping them deal with the complex task of evaluating PN burden and progression.

Automatic lung tumor segmentation with leaks removal in follow-up CT studies.

PURPOSE: In modern oncology, disease progression and response to treatment are routinely evaluated with a series of volumetric scans. The number of tumors and their volume (mass) over time provides a quantitative measure for the evaluation. Thus, many of the scans are follow-up scans. We present a new, fully automatic algorithm for lung tumors segmentation in follow-up CT studies that takes advantage of the baseline delineation. METHODS: The inputs are a baseline CT scan and a delineation of the tumors in it and a follow-up scan; the output is the tumor delineations in the follow-up CT scan; the output is the tumor delineations in the follow-up CT scan. The algorithm consists of four steps: (1) deformable registration of the baseline scan and tumor's delineations to the follow-up CT scan; (2) segmentation of these tumors in the follow-up CT scan with the baseline CT and the tumor's delineations as priors; (3) detection and correction of follow-up tumors segmentation leaks based on the geometry of both the foreground and the background; and (4) tumor boundary regularization to account for the partial volume effects. RESULTS: Our experimental results on 80 pairs of CT scans from 40 patients with ground-truth segmentations by a radiologist yield an average DICE overlap error of 14.5 % ([Formula: see text]), a significant improvement from the 30 % ([Formula: see text]) result of stand-alone level-set segmentation. CONCLUSION: The key advantage of our method is that it automatically builds a patient-specific prior to the tumor. Using this prior in the segmentation process, we developed an algorithm that increases segmentation accuracy and robustness and reduces observer variability.

The role of automatic computer-aided surgical trajectory planning in improving the expected safety of stereotactic neurosurgery.

PURPOSE: Minimal invasion computer-assisted neurosurgical procedures with various tool insertions into the brain may carry hemorrhagic risks and neurological deficits. The goal of this study is to investigate the role of computer-based surgical trajectory planning tools in improving the potential safety of image-based stereotactic neurosurgery. METHODS: Multi-sequence MRI studies of eight patients who underwent image-guided neurosurgery were retrospectively processed to extract anatomical structures-head surface, ventricles, blood vessels, white matter fibers tractography, and fMRI data of motor, sensory, speech, and visual areas. An experienced neurosurgeon selected one target for each patient. Five neurosurgeons planned a surgical trajectory for each patient using three planning methods: (1) conventional; (2) visualization, in which scans are augmented with overlays of anatomical structures and functional areas; and (3) automatic, in which three surgical trajectories with the lowest expected risk score are automatically computed. For each surgeon, target, and method, we recorded the entry point and its surgical trajectory and computed its expected risk score and its minimum distance from the key structures. RESULTS: A total of 120 surgical trajectories were collected (5 surgeons, 8 targets, 3 methods). The surgical trajectories expected risk scores improved by 76% ([Formula: see text], two-sample student's t test); the average distance of a trajectory from nearby blood vessels increased by 1.6 mm ([Formula: see text]) from 0.6 to 2.2 mm (243%). The initial surgical trajectories were changed in 85% of the cases based on the expected risk score and the trajectory distance from blood vessels. CONCLUSIONS: Computer-based patient-specific preoperative planning of surgical trajectories that minimize the expected risk of vascular and neurological damage due to incorrect tool placement is a promising technique that yields consistent improvements.

Carotid vasculature modeling from patient CT angiography studies for interventional procedures simulation.

OBJECTIVE: A practical method for patient-specific modeling of the aortic arch and the entire carotid vasculature from computed tomography angiography (CTA) scans for morphologic analysis and for interventional procedure simulation. MATERIALS AND METHODS: The method starts with the automatic watershed-based segmentation of the aorta and the construction of an a-priori intensity probability distribution function for arteries. The carotid arteries are then segmented with a graph min-cut method based on a new edge weighting function that adaptively couples voxel intensity, intensity prior, and local vesselness shape prior. Finally, the same graph-cut optimization framework is used to interactively remove a few unwanted veins segments and to fill in minor vessel discontinuities caused by intensity variations. RESULTS: We validate our modeling method with two experimental studies on 71 multicenter clinical CTA datasets, including carotid bifurcation lumen segmentation on 56 CTAs from the MICCAI'2009 3D Segmentation Challenge. Segmentation results show that our method is comparable to the best existing methods and was successful in modeling the entire carotid vasculature with a Dice similarity measure of 84.5% (SD = 3.3%) and MSSD 0.48 mm (SD = 0.12 mm.) Simulation study shows that patient-specific simulations with four patient-specific models generated by our segmentation method on the ANGIO Mentor™ simulator platform are robust, realistic, and greatly improve the simulation. CONCLUSION: This constitutes a proof-of-concept that patient-specific CTA-based modeling and simulation of carotid interventional procedures are practical in a clinical environment.

Automatic segmentation, internal classification, and follow-up of optic pathway gliomas in MRI.

This paper presents an automatic method for the segmentation, internal classification and follow-up of optic pathway gliomas (OPGs) from multi-sequence MRI datasets. Our method starts with the automatic localization of the OPG and its core with an anatomical atlas followed by a binary voxel classification with a probabilistic tissue model whose parameters are estimated from the MR images. The method effectively incorporates prior location, tissue characteristics, and intensity information for the delineation of the OPG boundaries in a consistent and repeatable manner. Internal classification of the segmented OPG volume is then obtained with a robust method that overcomes grey-level differences between learning and testing datasets. Experimental results on 25 datasets yield a mean surface distance error of 0.73 mm as compared to manual segmentation by experienced radiologists. Our method exhibits reliable performance in OPG growth follow-up MR studies, which are crucial for monitoring disease progression. To the best of our knowledge, this is the first method that addresses automatic segmentation, internal classification, and follow-up of OPG.

Evaluation framework for carotid bifurcation lumen segmentation and stenosis grading.

This paper describes an evaluation framework that allows a standardized and objective quantitative comparison of carotid artery lumen segmentation and stenosis grading algorithms. We describe the data repository comprising 56 multi-center, multi-vendor CTA datasets, their acquisition, the creation of the reference standard and the evaluation measures. This framework has been introduced at the MICCAI 2009 workshop 3D Segmentation in the Clinic: A Grand Challenge III, and we compare the results of eight teams that participated. These results show that automated segmentation of the vessel lumen is possible with a precision that is comparable to manual annotation. The framework is open for new submissions through the website http://cls2009.bigr.nl.

A curvelet-based patient-specific prior for accurate multi-modal brain image rigid registration.

We present a new non-uniform sampling method for the accurate estimation of mutual information in multi-modal brain image rigid registration. Most existing density estimators used for mutual information computation incorrectly assume that the intensity of each voxel is independent from its neighborhood. Our method uses the 3D Fast Discrete Curvelet Transform to reduce the sampled voxels' interdependency by sampling voxels that are less dependent on their neighborhood, and thus provide a more accurate estimation of the mutual information and a more accurate registration. The main advantages of our method over other non-uniform sampling schemes are that: (1) it provides more accurate estimation of the image statistics with fewer samples; (2) it is less sensitive to the variability of anatomical structures shapes, orientations, and sizes, and; (3) it yields more accurate registration results. Extensive evaluation on 1000 synthetic registrations between T1 and T2-weighted clinical MRI images and 20 real clinical registrations of brain CT images to Proton Density (PD) and T1 and T2-weighted MRI images from the public RIRE database show the effectiveness of our method. Our method has the lowest mean registration errors recorded to date for CT-MR image registration in the RIRE website for methods tested on more than five datasets. These results indicate that our sampling scheme can be used to achieve more accurate multi-modal registration required for image guided therapy and surgery.

Non-parametric iterative model constraint graph min-cut for automatic kidney segmentation.

We present a new non-parametric model constraint graph min-cut algorithm for automatic kidney segmentation in CT images. The segmentation is formulated as a maximum a-posteriori estimation of a model-driven Markov random field. A non-parametric hybrid shape and intensity model is treated as a latent variable in the energy functional. The latent model and labeling map that minimize the energy functional are then simultaneously computed with an expectation maximization approach. The main advantages of our method are that it does not assume a fixed parametric prior model, which is subjective to inter-patient variability and registration errors, and that it combines both the model and the image information into a unified graph min-cut based segmentation framework. We evaluated our method on 20 kidneys from 10 CT datasets with and without contrast agent for which ground-truth segmentations were generated by averaging three manual segmentations. Our method yields an average volumetric overlap error of 10.95%, and average symmetric surface distance of 0.79 mm. These results indicate that our method is accurate and robust for kidney segmentation.

A bayesian approach for liver analysis: algorithm and validation study.

We present a new method for the simultaneous, nearly automatic segmentation of liver contours, vessels, and metastatic lesions from abdominal CTA scans. The method repeatedly applies multi-resolution, multi-class smoothed Bayesian classification followed by morphological adjustment and active contours refinement. It uses multi-class and voxel neighborhood information to compute an accurate intensity distribution function for each class. The method requires only one or two user-defined voxel seeds, with no manual adjustment of internal parameters. A retrospective study on two validated clinical datasets totaling 56 CTAs was performed. We obtained correlations of 0.98 and 0.99 with a manual ground truth liver volume estimation for the first and second databases, and a total score of 67.87 for the second database. These results suggest that our method is accurate, efficient, and robust to seed selection compared to manually generated ground truth segmentation and to other semi-automatic segmentation methods.

Classification of suspected liver metastases using fMRI images: a machine learning approach.

This paper presents a machine-learning approach to the interactive classification of suspected liver metastases in fMRI images. The method uses fMRI-based statistical modeling to characterize colorectal hepatic metastases and follow their early hemodynamical changes. Changes in hepatic hemodynamics are evaluated from T2*-W fMRI images acquired during the breathing of air, air-CO2, and carbogen. A classification model is build to differentiate between tumors and healthy liver tissues. To validate our method, a model was built from 29 mice datasets, and used to classify suspicious regions in 16 new datasets of healthy subjects or subjects with metastases in earlier growth phases. Our experimental results on mice yielded an accuracy of 78% with high precision (88%). This suggests that the method can provide a useful aid for early detection of liver metastases.