Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain.

The finite mixture (FM) model is the most commonly used model for statistical segmentation of brain magnetic resonance (MR) images because of its simple mathematical form and the piecewise constant nature of ideal brain MR images. However, being a histogram-based model, the FM has an intrinsic limitation-no spatial information is taken into account. This causes the FM model to work only on well-defined images with low levels of noise; unfortunately, this is often not the the case due to artifacts such as partial volume effect and bias field distortion. Under these conditions, FM model-based methods produce unreliable results. Here, the authors propose a novel hidden Markov random field (HMRF) model, which is a stochastic process generated by a MRF whose state sequence cannot be observed directly but which can be indirectly estimated through observations. Mathematically, it can be shown that the FM model is a degenerate version of the HMRF model. The advantage of the HMRF model derives from the way in which the spatial information is encoded through the mutual influences of neighboring sites. Although MRF modeling has been employed in MR image segmentation by other researchers, most reported methods are limited to using MRF as a general prior in an FM model-based approach. To fit the HMRF model, an EM algorithm is used. The authors show that by incorporating both the HMRF model and the EM algorithm into a HMRF-EM framework, an accurate and robust segmentation can be achieved. More importantly, the HMRF-EM framework can easily be combined with other techniques. As an example, the authors show how the bias field correction algorithm of Guillemaud and Brady (1997) can be incorporated into this framework to achieve a three-dimensional fully automated approach for brain MR image segmentation.

/pdf/segmentation-of-brain-mr-images-through-a-hidden-markov-2vxcvxn8aq.pdf

Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm

We present a technique for automatically assigning a neuroanatomical label to each location on a cortical surface model based on probabilistic information estimated from a manually labeled training set. This procedure incorporates both geometric information derived from the cortical model, and neuroanatomical convention, as found in the training set. The result is a complete labeling of cortical sulci and gyri. Examples are given from two different training sets generated using different neuroanatomical conventions, illustrating the flexibility of the algorithm. The technique is shown to be comparable in accuracy to manual labeling.

/pdf/automatically-parcellating-the-human-cerebral-cortex-16g7uhc9ep.pdf

Automatically Parcellating the Human Cerebral Cortex

▪ Abstract Image segmentation plays a crucial role in many medical-imaging applications, by automating or facilitating the delineation of anatomical structures and other regions of interest. We present a critical appraisal of the current status of semiautomated and automated methods for the segmentation of anatomical medical images. Terminology and important issues in image segmentation are first presented. Current segmentation approaches are then reviewed with an emphasis on the advantages and disadvantages of these methods for medical imaging applications. We conclude with a discussion on the future of image segmentation methods in biomedical research.

Current methods in medical image segmentation.

Describes a fully automated method for model-based tissue classification of magnetic resonance (MR) images of the brain. The method interleaves classification with estimation of the model parameters, improving the classification at each iteration. The algorithm is able to segment single- and multi-spectral MR images, corrects for MR signal inhomogeneities, and incorporates contextual information by means of Markov random Fields (MRF's). A digital brain atlas containing prior expectations about the spatial location of tissue classes is used to initialize the algorithm. This makes the method fully automated and therefore it provides objective and reproducible segmentations. The authors have validated the technique on simulated as well as on real MR images of the brain.

/pdf/automated-model-based-tissue-classification-of-mr-images-of-2l1ptkjz64.pdf

Automated model-based tissue classification of MR images of the brain

Describes a fully-automatic three-dimensional (3-D)-segmentation technique for brain magnetic resonance (MR) images. By means of Markov random fields (MRF's) the segmentation algorithm captures three features that are of special importance for MR images, i.e., nonparametric distributions of tissue intensities, neighborhood correlations, and signal inhomogeneities. Detailed simulations and real MR images demonstrate the performance of the segmentation algorithm. In particular, the impact of noise, inhomogeneity, smoothing, and structure thickness are analyzed quantitatively. Even single-echo MR images are well classified into gray matter, white matter, cerebrospinal fluid, scalp-bone, and background. A simulated annealing and an iterated conditional modes implementation are presented.

Markov random field segmentation of brain MR images

The technical characteristics of the multislice whole-body positron emission tomographic scanner (model PC4096-15WB Scanditronix) and its performance parameters are described. Spatial resolution at the center of the field of view was found to be 4.9 mm in-plane and 4.6 mm (cross slices) and 6.0 mm (direct slices) in the axial direction. The sensitivity for true and scattered coincidences is approximately 5,000 cps for direct slices and 7,100 cps for cross slices. At an activity concentration of 37 kBq/ml the system deadtime was approximately 5%. By measuring a uniform phantom with a cold cylindrical insert (5.0 cm diameter), the scatter fraction was found to be approximately 5%. The mean global uniformity over all 15 slices was 6.5%, whereas the local uniformity was found to be 4.3%. No systematic nonuniformities were observed. Finally, various methods for attenuation correction (transmission scan, contour finding, ellipse) were utilized to test their effects on the resulting reconstructed images.

Performance characteristics of an eight-ring whole body PET scanner.

Until recently, no direct comparison between [(15)O]water positron emission tomography (PET) and arterial spin labeling (ASL) for measuring cerebral blood flow (CBF) was possible. With the introduction of integrated, hybrid magnetic resonance (MR)-PET scanners, such a comparison becomes feasible. This study presents results of CBF measurements recorded simultaneously with [(15)O]water and ASL. A 3T MR-BrainPET scanner was used for the simultaneous acquisition of pseudo-continuous ASL (pCASL) magnetic resonance imaging (MRI) and [(15)O]water PET. Quantitative CBF values were compared in 10 young healthy male volunteers at baseline conditions. A statistically significant (P<0.05) correlation was observed between the two modalities; the whole-brain CBF values determined with PET and pCASL were 43.3±6.1 mL and 51.9±7.1 mL per 100 g per minute, respectively. The gray/white matter (GM/WM) ratio of CBF was 3.0 for PET and 3.4 for pCASL. A paired t-test revealed differences in regional CBF between ASL and PET with higher ASL-CBF than PET-CBF values in cortical areas. Using an integrated, hybrid MR-PET a direct simultaneous comparison between ASL and [(15)O]water PET became possible for the first time so that temporal, physiologic, and functional variations were avoided. Regional and individual differences were found despite the overall similarity between ASL and PET, requiring further detailed investigations.

https://journals.sagepub.com/doi/pdf/10.1038/jcbfm.2014.92

Comparison of cerebral blood flow acquired by simultaneous [15O]water positron emission tomography and arterial spin labeling magnetic resonance imaging.

This paper describes and compares procedures to obtain attenuation maps used for the absorption correction (AC) of PET brain scans if a transmission scan is not available as in the case of future MR-PET scanners. A previously reported approach called MBA (MRT-based attenuation correction) used Tl- weighted MR images which were segmented into four tissue types representing brain tissue, bone, other tissue and sinus to which appropriate attenuation coefficients were assigned. In this work a template-based attenuation correction (TBA) is presented which applies an attenuation template to single subjects. A common attenuation template was created from transmission scans of 10 normal volunteers and spatially normalized to the SPM2 standard brain shape. For each subject the Tl-MR template of SPM2 was warped onto the subject's individual MR image. The resulting warping matrix was applied to the common attenuation template so that an attenuation map matching the subject's brain shape was obtained. The attenuation maps of MBA and TBA were forward projected into attenuation factors which were alternatively used for AC. FDG scans of four subjects were reconstructed after AC with MBA and TBA and compared to images whose ACs were based on conventional attenuation maps (PBA=PET-based attenuation correction). Using PBA as reference in a region of interest analysis, MBA and TBA showed similar under- and overestimation of the reconstructed radioactivity up to -10% and 9%, respectively. The procedure to obtain the attenuation template needs still some improvements. Nevertheless, the TBA method of attenuation correction is a promising alternative to MBA with its still complex and not yet resolved accurate segmentation of MR images.

Alternative methods for attenuation correction for PET images in MR-PET scanners

PET studies of cerebral neuroreceptors are often recorded over periods ranging from 1 to 2 h, and head movements during the studies not only lead to blurred images but also may seriously disturb the kinetic analysis. We report the effect of motion on parametric images of the distribution volume ratio (DVR), as well as possible improvements if the dynamic PET data are corrected for head movements. Methods: The study was performed with the 5-hydroxytryptamine 2A receptor ligand 18F-altanserin. During PET scanning, which was performed in list mode for 1 h, the position of the head was monitored by an infrared motion-tracking system. The list mode data were sorted into time frames of between 10 s and 2 min. Motion was corrected using the multiple-acquisition-frame (MAF) approach, which calculates individual attenuation files for each emission frame and its corresponding head position to avoid misalignment of transmission and emission data. After reconstruction of attenuation-corrected emission frames, each image frame was realigned to match the head position of the first frame of the emission scan. The resulting motion-corrected dynamic images were evaluated using the noninvasive Logan plot to obtain parametric images of DVR. Results: DVR images of motion-affected 18F-altanserin scans showed artifacts whose extent depended on the amount of movement. The artifacts were mainly at the border between gray matter and white matter and at the outer border of gray matter. They were seen as discontinuities and small spots whose values exceeded the expected DVR values or were even negative and that disappeared when motion correction was applied. These effects in human data were also seen on simulated 18F-altanserin images that contained no statistical noise. Conclusion: Whereas the native PET images looked just blurred if the patient moved during the PET scan, parametric images of the Logan DVR, which are calculated by pixelwise linear regression, contained severe discontinuities primarily at the cortical edge. MAF-based motion correction was able to avoid these errors.

https://jnm.snmjournals.org/content/jnumed/46/6/1059.full.pdf

Elena Rota Kops

Papers

Markov random field segmentation of brain MR images

Performance characteristics of an eight-ring whole body PET scanner.

Comparison of cerebral blood flow acquired by simultaneous [15O]water positron emission tomography and arterial spin labeling magnetic resonance imaging.

Alternative methods for attenuation correction for PET images in MR-PET scanners

Motion Artifact Reduction on Parametric PET Images of Neuroreceptor Binding