Renzo Phellan

Medical image segmentation via atlases and fuzzy object models: Improving efficacy through optimum object search and fewer models

PURPOSEnStatistical object shape models (SOSMs), known as probabilistic atlases, are popular in medical image segmentation. They register an image into the atlas coordinate system, such that a desired object can be delineated from the constraints of its shape model. While this strategy facilitates segmenting objects with even weak-boundary contrast, it tends to require more models per object to cope with possible registration errors. Fuzzy object shape models (FOSMs) gain substantial speed by avoiding image registration and placing more relaxed model constraints with optimum object search. However, they tend to require stronger object boundary contrast for effective delineation. In this work, the authors show that optimum object search, the essential underpinning of FOSMs, can improve segmentation efficacy of SOSMs with fewer models per object.nnnMETHODSnFor the sake of efficiency, the authors use three atlases per object (SOSM-3) as baseline for segmentation based on the best match with posterior probability maps. A novel strategy for SOSM with a single atlas and optimum object search (SOSM-S) is presented. When registering an image to the atlas system, one should expect that the objects boundary falls within the uncertainty region of the model-region wherein voxels show probabilities greater than 0 and less than 1 to be in the object. Since registration may fail, SOSM-S translates the atlas locally and, at each location, delineates and scores a candidate object in the uncertainty region. Segmentation is defined by the candidate with the highest score. The presented FOSM also uses a single model per object, but model construction uses only shape translations, building a fuzzy object model with larger uncertainty region. Optimum object search requires estimation of the objects location and/or optimization algorithms to speed-up segmentation.nnnRESULTSnThe authors evaluate SOSM-3, SOSM-S, and FOSM on 75 CT-images of the thorax and 35 MR T1-weighted images of the brain, with nine objects of interest. The results show that SOSM-S and FOSM can segment seven out of the nine objects with higher accuracy than SOSM-3, according to the average symmetric surface distance and statistical test. SOSM-S was consistently more accurate than FOSM, FOSM being 2-3 orders of magnitude faster than SOSM-S and SOSM-3 for model construction and hundreds of times faster than them for segmentation.nnnCONCLUSIONSnAlthough multiple models per object can usually improve segmentation efficacy, the optimum object search has shown to reduce the number of required models. The efficiency gain of FOSM over SOSM-S motivates its use for interactive applications and studies with large image data sets. FOSM and SOSM impose different degrees of shape constraints from the model, making one approach more suitable than the other, depending on contrast. This suggests the use of hybrid models that can take advantage from the strengths of fuzzy and statistical models.

Improving Atlas-Based Medical Image Segmentation with a Relaxed Object Search

Medical image segmentation using 3D probabilistic atlases has been actively pursued to avoid the time-consuming involvement of experts in manual object (organ) delineation for quantitative analysis. By mapping a new 3D image onto the reference coordinate system of the atlas, built for some organ of interest, these techniques take a binary decision based on the probability of each voxel to be part of that organ. However, image-based techniques have also been proposed to refine object delineation at the initial position given by the atlas-based segmentation. In this paper, we relax this condition for delineation refinement by moving an atlas based on the prior probability map to search for the organ around that initial position. Our method uses the multi-scale parameter search algorithm with a suitable criterion function to evaluate automatic 3D organ delineations, as obtained by the image foresting transform algorithm in an uncertainty region of the atlas. Experiments with eight organs in CT and MR images have indicated that our method can improve atlas-based segmentation with statistical significance. Moreover, the relaxed object search consistently found the organ with higher accuracy outside the position obtained by the atlas, which reinforces our claim.

Comparison of vessel enhancement algorithms applied to time‐of‐flight MRA images for cerebrovascular segmentation

Purpose: Vessel enhancement algorithms are often used as a preprocessing step for vessel segmentation in medical images to improve the overall segmentation accuracy. Each algorithm uses different characteristics to enhance vessels, such that the most suitable algorithm may vary for different applications. This paper presents a comparative analysis of the accuracy gains in vessel segmentation generated by the use of nine vessel enhancement algorithms: Multiscale vesselness using the formulas described by Erdt (MSE), Frangi (MSF), and Sato (MSS), optimally oriented flux (OOF), ranking orientations responses path operator (RORPO), the regularized Perona–Malik approach (RPM), vessel enhanced diffusion (VED), hybrid diffusion with continuous switch (HDCS), and the white top hat algorithm (WTH). Methods: The filters were evaluated and compared based on time‐of‐flight MRA datasets and corresponding manual segmentations from 5 healthy subjects and 10 patients with an arteriovenous malformation. Additionally, five synthetic angiographic datasets with corresponding ground truth segmentation were generated with three different noise levels (low, medium, and high) and also used for comparison. The parameters for each algorithm and subsequent segmentation were optimized using leave‐one‐out cross evaluation. The Dice coefficient, Matthews correlation coefficient, area under the ROC curve, number of connected components, and true positives were used for comparison. Results: The results of this study suggest that vessel enhancement algorithms do not always lead to more accurate segmentation results compared to segmenting nonenhanced images directly. Multiscale vesselness algorithms, such as MSE, MSF, and MSS proved to be robust to noise, while diffusion‐based filters, such as RPM, VED, and HDCS ranked in the top of the list in scenarios with medium or no noise. Filters that assume tubular‐shapes, such as MSE, MSF, MSS, OOF, RORPO, and VED show a decrease in accuracy when considering patients with an AVM, because vessels may vary from its tubular‐shape in this case. Conclusions: Vessel enhancement algorithms can help to improve the accuracy of the segmentation of the vascular system. However, their contribution to accuracy has to be evaluated as it depends on the specific applications, and in some cases it can lead to a reduction of the overall accuracy. No specific filter was suitable for all tested scenarios.

Vascular Segmentation in TOF MRA Images of the Brain Using a Deep Convolutional Neural Network

Cerebrovascular diseases are one of the main causes of death and disability in the world. Within this context, fast and accurate automatic cerebrovascular segmentation is important for clinicians and researchers to analyze the vessels of the brain, determine criteria of normality, and identify and study cerebrovascular diseases. Nevertheless, automatic segmentation is challenging due to the complex shape, inhomogeneous intensity, and inter-person variability of normal and malformed vessels. In this paper, a deep convolutional neural network (CNN) architecture is used to automatically segment the vessels of the brain in time-of-flight magnetic resonance angiography (TOF MRA) images of healthy subjects. Bi-dimensional manually annotated image patches are extracted in the axial, coronal, and sagittal directions and used as input for training the CNN. For segmentation, each voxel is individually analyzed using the trained CNN by considering the intensity values of neighboring voxels that belong to its patch. Experiments were performed with TOF MRA images of five healthy subjects, using varying numbers of images to train the CNN. Cross validations revealed that the proposed framework is able to segment the vessels with an average Dice coefficient ranging from 0.764 to 0.786 depending on the number of images used for training. In conclusion, the results of this work suggest that CNNs can be used to segment cerebrovascular structures with an accuracy similar to other high-level segmentation methods.

Medical image segmentation using object atlas versus object cloud models

Medical image segmentation is crucial for quantitative organ analysis and surgical planning. Since interactive segmentation is not practical in a production-mode clinical setting, automatic methods based on 3D object appearance models have been proposed. Among them, approaches based on object atlas are the most actively investigated. A key drawback of these approaches is that they require a time-costly image registration process to build and deploy the atlas. Object cloud models (OCM) have been introduced to avoid registration, considerably speeding up the whole process, but they have not been compared to object atlas models (OAM). The present paper fills this gap by presenting a comparative analysis of the two approaches in the task of individually segmenting nine anatomical structures of the human body. Our results indicate that OCM achieve a statistically significant better accuracy for seven anatomical structures, in terms of Dice Similarity Coefficient and Average Symmetric Surface Distance.

Four-Dimensional ASL MR Angiography Phantoms with Noise Learned by Neural Styling

Annotated datasets for evaluation and validation of medical image processing methods can be difficult and expensive to obtain. Alternatively, simulated datasets can be used, but adding realistic noise properties is especially challenging. This paper proposes using neural styling, a deep learning based algorithm, which can automatically learn noise patterns from real medical images and reproduce these patterns in the simulated datasets. In this work, the imaging modality to be simulated is four-dimensional arterial spin labeling magnetic resonance angiography (4D ASL MRA), a modality that includes information of the cerebrovascular geometry and blood flow. The cerebrovascular geometry used to create the simulated phantoms is obtained from segmentations of 3D time-of-flight (TOF) MRA images of healthy volunteers. Dynamic blood flow is simulated according to a mathematical model designed specifically to describe the signal generated by 4D ASL MRA series. Finally, noise is added by using neural styling to learn the noise patterns present in real 4D ASL MRA datasets. Qualitative evaluation of two simulated 4D ASL MRA datasets revealed high similarity of the blood flow dynamics and noise properties as compared to the corresponding real 4D ASL MRA datasets. These simulated phantoms, with realistic noise properties, can be useful for the development, optimization, and evaluation of image processing methods focused on segmentation and blood flow parameters estimation in 4D ASL MRA series.

Association of Cardiovascular Risk Factors With MRI Indices of Cerebrovascular Structure and Function and White Matter Hyperintensities in Young Adults.

Importance Risk of stroke and brain atrophy in later life relate to levels of cardiovascular risk in early adulthood. However, it is unknown whether cerebrovascular changes are present in young adults. Objective To examine relationships between modifiable cardiovascular risk factors and cerebrovascular structure, function, and white matter integrity in young adults. Design, Setting, and Participants A cross-sectional observational study of 125 young adults (aged 18-40 years) without clinical evidence of cerebrovascular disease. Data collection was completed between August 2014 and May 2016 at the University of Oxford, United Kingdom. Final data collection was completed on May 31, 2016. Exposures The number of modifiable cardiovascular risk factors at recommended levels, based on the following criteria: body mass index (BMI) <25; highest tertile of cardiovascular fitness and/or physical activity; alcohol consumption <8 drinks/week; nonsmoker for >6 months; blood pressure on awake ambulatory monitoring <130/80 mm Hg; a nonhypertensive diastolic response to exercise (peak diastolic blood pressure <90 mm Hg); total cholesterol <200 mg/dL; and fasting glucose <100mg/dL. Each risk factor at the recommended level was assigned a value of 1, and participants were categorized from 0-8, according to the number of risk factors at recommended levels, with higher numbers indicating healthier risk categories. Main Outcomes and Measures Cerebral vessel density, caliber and tortuosity, brain white matter hyperintensity lesion count. In a subgroup (nu2009=u200952), brain blood arrival time and cerebral blood flow assessed by brain magnetic resonance imaging (MRI). Results A total of 125 participants, mean (SD) age 25u2009(5) years, 49% women, with a mean (SD) score of 6.0 (1.4) modifiable cardiovascular risk factors at recommended levels, completed the cardiovascular risk assessment and brain MRI protocol. Cardiovascular risk factors were correlated with cerebrovascular morphology and white matter hyperintensity count in multivariable models. For each additional modifiable risk factor categorized as healthy, vessel density was greater by 0.3 vessels/cm3 (95% CI, 0.1-0.5; Pu2009=u2009.003), vessel caliber was greater by 8 &mgr;m (95% CI, 3-13; Pu2009=u2009.01), and white matter hyperintensity lesions were fewer by 1.6 lesions (95% CI, −3.0 to −0.5; Pu2009=u2009.006). Among the 52 participants with available data, cerebral blood flow varied with vessel density and was 2.5 mL/100 g/min higher for each healthier category of a modifiable risk factor (95% CI, 0.16-4.89; Pu2009=u2009.03). Conclusions and Relevance In this preliminary study involving young adults without clinical evidence of cerebrovascular disease, a greater number of modifiable cardiovascular risk factors at recommended levels was associated with higher cerebral vessel density and caliber, higher cerebral blood flow, and fewer white matter hyperintensities. Further research is needed to verify these findings and determine their clinical importance.

Vessel segmentation in 4D arterial spin labeling magnetic resonance angiography images of the brain

4D arterial spin labeling magnetic resonance angiography (4D ASL MRA) is a non-invasive and safe modality for cerebrovascular imaging procedures. It uses the patient’s magnetically labeled blood as intrinsic contrast agent, so that no external contrast media is required. It provides important 3D structure and blood flow information but a sufficient cerebrovascular segmentation is important since it can help clinicians to analyze and diagnose vascular diseases faster, and with higher confidence as compared to simple visual rating of raw ASL MRA images. This work presents a new method for automatic cerebrovascular segmentation in 4D ASL MRA images of the brain. In this process images are denoised, corresponding image label/control image pairs of the 4D ASL MRA sequences are subtracted, and temporal intensity averaging is used to generate a static representation of the vascular system. After that, sets of vessel and background seeds are extracted and provided as input for the image foresting transform algorithm to segment the vascular system. Four 4D ASL MRA datasets of the brain arteries of healthy subjects and corresponding time-of-flight (TOF) MRA images were available for this preliminary study. For evaluation of the segmentation results of the proposed method, the cerebrovascular system was automatically segmented in the high-resolution TOF MRA images using a validated algorithm and the segmentation results were registered to the 4D ASL datasets. Corresponding segmentation pairs were compared using the Dice similarity coefficient (DSC). On average, a DSC of 0.9025 was achieved, indicating that vessels can be extracted successfully from 4D ASL MRA datasets by the proposed segmentation method.