This PDF file contains the front matter associated with SPIE Proceedings Volume 9681, including the Title Page, Copyright information, Table of Contents, Introduction, Authors, and Conference Committee listing.

The use of low-level visual features to assign high level labels in datasets of histopathology images is a possible solution to the problems derived from manual labeling by experts. However, in many cases, the visual cues are not enough. In this article we propose the use of features derived exclusively from the spatial distribution of the cell nuclei. These features are calculated using the weight of k-nn graphs constructed from the distances between cells. Results show that there are k values with enhanced discriminatory power, especially when comparing cancerous and non-cancerous tissue.

Convolutional neural networks (CNN) have been very successful at addressing different computer vision tasks thanks to their ability to learn image representations directly from large amounts of labeled data. Features learned from a dataset can be used to represent images from a different dataset via an approach called transfer learning. In this paper we apply transfer learning to the challenging task of medulloblastoma tumor differentiation. We compare two different CNN models which were previously trained in two different domains (natural and histopathology images). The first CNN is a state-of-the-art approach in computer vision, a large and deep CNN with 16-layers, Visual Geometry Group (VGG) CNN. The second (IBCa-CNN) is a 2-layer CNN trained for invasive breast cancer tumor classification. Both CNNs are used as visual feature extractors of histopathology image regions of anaplastic and non-anaplastic medulloblastoma tumor from digitized whole-slide images. The features from the two models are used, separately, to train a softmax classifier to discriminate between anaplastic and non-anaplastic medulloblastoma image regions. Experimental results show that the transfer learning approach produce competitive results in comparison with the state of the art approaches for IBCa detection. Results also show that features extracted from the IBCa-CNN have better performance in comparison with features extracted from the VGG-CNN. The former obtains 89.8% while the latter obtains 76.6% in terms of average accuracy.

In histology, each dye component attempts to specifically characterise different microscopic structures. In the case of the Hematoxylin-Eosin (H&E) stain, universally used for routine examination, quantitative analysis may often require the inspection of different morphological signatures related mainly to nuclei patterns, but also to stroma distribution. Nevertheless, computer systems for automatic diagnosis are often fraught by color variations ranging from the capturing device to the laboratory specific staining protocol and stains. This paper presents a novel colour normalisation method for H&E stained histopathology images. This method is based upon the opponent process theory and blindly estimates the best color basis for the Hematoxylin and Eosin stains without relying on prior knowledge. Stain Normalisation and Color Separation are transversal to any Framework of Histopathology Image Analysis.

Background and motivation: Real-time ultrasound simulation refers to the process of computationally creating fully synthetic ultrasound images instantly. Due to the high value of specialized low cost training for healthcare professionals, there is a growing interest in the use of this technology and the development of high fidelity systems that simulate the acquisitions of echographic images. The objective is to create an efficient and reproducible simulator that can run either on notebooks or desktops using low cost devices. Materials and methods: We present an interactive ultrasound simulator based on CT data. This simulator is based on ray-casting and provides real-time interaction capabilities. The simulation of scattering that is coherent with the transducer position in real time is also introduced. Such noise is produced using a simplified model of multiplicative noise and convolution with point spread functions (PSF) tailored for this purpose. Results: The computational efficiency of scattering maps generation was revised with an improved performance. This allowed a more efficient simulation of coherent scattering in the synthetic echographic images while providing highly realistic result. We describe some quality and performance metrics to validate these results, where a performance of up to 55fps was achieved. Conclusion: The proposed technique for real-time scattering modeling provides realistic yet computationally efficient scatter distributions. The error between the original image and the simulated scattering image was compared for the proposed method and the state-of-the-art, showing negligible differences in its distribution.

MRI-based bone segmentation is a challenging task because bone tissue and air both present low signal intensity on MR images, making it difficult to accurately delimit the bone boundaries. However, estimating bone from MRI images may allow decreasing patient ionization by removing the need of patient-specific CT acquisition in several applications. In this work, we propose a fast GPU-based pseudo-CT generation from a patient-specific MRI T1-weighted image using a group-wise patch-based approach and a limited MRI and CT atlas dictionary. For every voxel in the input MR image, we compute the similarity of the patch containing that voxel with the patches of all MR images in the database, which lie in a certain anatomical neighborhood. The pseudo-CT is obtained as a local weighted linear combination of the CT values of the corresponding patches. The algorithm was implemented in a GPU. The use of patch-based techniques allows a fast and accurate estimation of the pseudo-CT from MR T1-weighted images, with a similar accuracy as the patient-specific CT. The experimental normalized cross correlation reaches 0.9324±0.0048 for an atlas with 10 datasets. The high NCC values indicate how our method can accurately approximate the patient-specific CT. The GPU implementation led to a substantial decrease in computational time making the approach suitable for real applications.

Diffusion-weighted imaging is sensitive to the movement of water molecules through the tissue microstructure and can therefore be used to gain insight into the tissue cellular architecture. While the diffusion signal arising from simple geometrical microstructure is known analytically, it remains unclear what diffusion signal arises from complex microstructural configurations. Such knowledge is important to design optimal acquisition sequences, to understand the limitations of diffusion-weighted imaging and to validate novel models of the brain microstructure. We present a novel framework for the efficient simulation of high-quality DW-MRI signals based on the hybrid combination of exact analytic expressions in simple geometric compartments such as cylinders and spheres and Monte Carlo simulations in more complex geometries. We validate our approach on synthetic arrangements of parallel cylinders representing the geometry of white matter fascicles, by comparing it to complete, all-out Monte Carlo simulations commonly used in the literature. For typical configurations, equal levels of accuracy are obtained with our hybrid method in less than one fifth of the computational time required for Monte Carlo simulations.

Sickle cell disease (SCD) is a hereditary blood disorder in which the oxygen-carrying hemoglobin molecule in red blood cells is abnormal. It affects numerous people in the world and leads to a shorter life span, pain, anemia, serious infections and neurocognitive decline. Tract-Specific Analysis (TSA) is a statistical method to evaluate white matter alterations due to neurocognitive diseases, using diffusion tensor magnetic resonance images. Here, for the first time, TSA is used to compare 11 major brain white matter (WM) tracts between SCD patients and age-matched healthy subjects. Alterations are found in the corpus callosum (CC), the cortico-spinal tract (CST), inferior fronto-occipital fasciculus (IFO), inferior longitudinal fasciculus (ILF), superior longitudinal fasciculus (SLF), and uncinated fasciculus (UNC). Based on previous studies on the neurocognitive functions of these tracts, the significant areas found in this paper might be related to several cognitive impairments and depression, both of which are observed in SCD patients.

Diffusion weighted imaging (DWI) can reveal the orientation of the underlying fiber populations in the brain. High angular resolution diffusion imaging (HARDI) is increasingly used to better resolve the orientation and mixing of fibers. Here, we assessed the added value of multi-shell q-space sampling on the reconstruction of major fibers using mathematical frameworks from q-ball imaging (QBI) and generalized q-sampling imaging (GQI), as compared to diffusion tensor imaging (DTI). We scanned a healthy mouse brain using 7-Tesla 5-shell HARDI (b=1000, 3000, 4000, 8000, 12000 s/mm²), also known as hybrid diffusion imaging (HYDI). We found that QBI may provide greater reconstruction accuracy for major fibers, which improves with the addition of higher b-value shells, unlike GQI or DTI (as expected). Although QBI is a special case of GQI, the major fiber orientation in QBI was more closely related to the orientation in DTI, rather than GQI. HYDI can aid the clinical outcomes of research and especially – more advanced human and animal connectomics projects to map the brain’s neural pathways and networks.

In the present work we study a family of generative network model and its applications for modeling the human connectome. We introduce a minor but novel variant of the Mixed Membership Stochastic Blockmodel and apply it and two other related model to two human connectome datasets (ADNI and a Bipolar Disorder dataset) with both control and diseased subjects. We further provide a simple generative classifier that, alongside more discriminating methods, provides evidence that blockmodels accurately summarize tractography count networks with respect to a disease classification task.

Deficits in white matter (WM) integrity occur following traumatic brain injury (TBI), and often persist long after the visible scars have healed. Heterogeneity in injury types and locations can complicate analyses, making it harder to discover common biomarkers for tracking recovery. Here we apply a newly developed adaptive connectivity method, EPIC (evolving partitions to improve connectomics) to identify differences in structural connectivity that persist longitudinally. This data comes from a longitudinal study, in which we scanned participants (aged 8-19 years) with anatomical and diffusion MRI in both the post-acute and chronic phases (1-6 months and 13-19 months post-injury). To identify patterns of abnormal connectivity, we trained a model on data from 32 TBI patients in the post-acute phase and 45 well-matched healthy controls, reducing an initial 68x68 connectivity matrix to a 14x14 matrix. We then applied this reduced parcellation to the chronic data in participants who had returned for their chronic assessment (21 TBI and 26 healthy controls) and tested for group differences. We found significant differences in two connections, comprising callosal fibers and long anterior-posterior fibers, with the TBI group showing increased fiber density relative to controls. Longitudinal analysis revealed that these were connections that were decreasing over time in the healthy controls, as is a common developmental phenomenon, but they were increasing in the TBI group. While we cannot definitively tell why this may occur with our current data, this study provides targets for longitudinal tracking, and poses questions for future investigation.

A conventional radiology report primarily consists of a large amount of unstructured text, and lacks clear, concise, consistent and content-rich information. Hence, an area of unmet clinical need consists of developing better ways to communicate radiology findings and information specific to each patient. Here, we design a new workflow and reporting system that combines and integrates advances in engineering technology with those from the medical sciences, the Multidimensional Interactive Radiology Report and Analysis (MIRRA). Until recently, clinical standards have primarily relied on 2D images for the purpose of measurement, but with the advent of 3D processing, many of the manually measured metrics can be automated, leading to better reproducibility and less subjective measurement placement. Hence, we make use this newly available 3D processing in our workflow. Our pipeline is used here to standardize the labeling, tracking, and quantifying of metrics for renal masses.

Retinal images are used for diagnostic purposes by ophthalmologists. However, despite controlled conditions in acquisition retinal images often suffer from non-uniform illumination which hinder their clinical use. In this work we propose the compensation of the illumination by modeling the intensity as a sum of non-stationary signals using bidimensional empirical mode decomposition (BEMD). We compare the estimation and compensation of the background illumination with a widely used technique based retinal image pixel classification. The proposed method has shown to provide a better estimation of the background illumination, particularly in complicated areas such as the optic disk (usually bright) and the periphery of fundus images (usually dim).

This paper describes a prototype app to induce lifestyle modifications in newly diagnosed type 2 diabetic patients. The app design is based on International Diabetes Federation guidelines and recommendations from clinical studies related to diabetes health-care. Two main approaches, lifestyle modification and self-management education are used owing to significant benefits reported. The method used is based on setting goals under medical support related to physical activity, nutritional habits and weight loss, in addition to educational messages. This is specially implemented to address the main challenges that have limited the success of similar mobile applications already validated on diabetic patients. A traffic light is used to show the overall state of the goals compliance. This state could be understood as excellent (green), there are aspects to improve (yellow), or some individual goals are not carrying out (red). An example of how works this method is presented in results. Furthermore, the app provides recommendations to the user in case the overall state was in yellow or red. The recommendations pretend to induce the user to make changes in their eating habits and physical activity. According to international guidelines and clinical studies, a prototype of mobile application to induce a lifestyle modification in order to prevent adverse risk factors related to diabetes was presented. The resulting application is apparently consistent with clinical judgments, but a formal clinical validation is required. The effectiveness of this app is currently under consideration for the Colombian population with type 2 diabetes.

Alzheimer's disease (AD) is a neurodegenerative disease that affects higher brain functions. Initial diagnosis of AD is based on the patient's clinical history and a battery of neuropsychological tests. The accuracy of the diagnosis is highly dependent on the examiner's skills and on the evolution of a variable clinical frame. This work presents an automatic strategy that learns probabilistic brain models for different stages of the disease, reducing the complexity, parameter adjustment and computational costs. The proposed method starts by setting a probabilistic class description using the information stored in the neuropsychological test, followed by constructing the different structural class models using membership values from the learned probabilistic functions. These models are then used as a reference frame for the classification problem: a new case is assigned to a particular class simply by projecting to the different models. The validation was performed using a leave-one-out cross-validation, two classes were used: Normal Control (NC) subjects and patients diagnosed with mild AD. In this experiment it is possible to achieve a sensibility and specificity of 80% and 79% respectively.

Segmentation or delineation of specific organs and structures in medical images is an important task in the clinical diagnosis and treatment, since it allows to characterize pathologies through imaging measures (biomarkers). In brain imaging, segmentation of main tissues or specific structures is challenging, due to the anatomic variability and complexity, and the presence of image artifacts (noise, intensity inhomogeneities, partial volume effect). In this paper, an automatic segmentation strategy is proposed, based on sparse representations and coupled dictionaries. Image intensity patterns are singly related to tissue labels at the level of small patches, gathering this information in coupled intensity/segmentation dictionaries. This dictionaries are used within a sparse representation framework to find the projection of a new intensity image onto the intensity dictionary, and the same projection can be used with the segmentation dictionary to estimate the corresponding segmentation. Preliminary results obtained with two publicly available datasets suggest that the proposal is capable of estimating adequate segmentations for gray matter (GM) and white matter (WM) tissues, with an average overlapping of 0:79 for GM and 0:71 for WM (with respect to original segmentations).

Children born preterm are at risk for a wide range of neurocognitive and neurobehavioral disorders. Some of these may stem from early brain abnormalities at the neonatal age. Hence, a precise characterization of neonatal neuroanatomy may help inform treatment strategies. In particular, the ventricles are often enlarged in neurocognitive disorders, due to atrophy of surrounding tissues. Here we present a new pipeline for the detection of morphological and relative pose differences in the ventricles of premature neonates compared to controls. To this end, we use a new hyperbolic Ricci flow based mapping of the ventricular surfaces of each subjects to the Poincaré disk. Resulting surfaces are then registered to a template, and a between group comparison is performed using multivariate tensor-based morphometry. We also statistically compare the relative pose of the ventricles within the brain between the two groups, by performing a Procrustes alignment between each subject's ventricles and an average shape. For both types of analyses, differences were found in the left ventricles between the two groups.

Several developmental disorders involve shape abnormalities of the neurocranium, the most common one being craniosynostosis, that affects about 1 in 2000 infants. A key step in determining how these disorders affect neurodevelopment is to establish how the brain and neurocranium co-evolve in the normally developing child. However, due to the scarcity of normally developing infant and pediatric imaging data, there have been a lack of imaging studies pertaining to normal neurocranial development. Here, taking advantage of a large data bank of high quality brain MRI from healthy children ages 0-4 years old, and of a novel conformal geometry-based analysis pipeline, we have been determining a set of statistical atlases of the neurocranium, divided into age groups. In this first part of the study, we focus more specifically on a comparison of 1 and 2 year old infants. Characterizing neurocranium shape changes will enable us to understand how the cranial bones develop in relation to brain development. This in turn will allow a better determination of the effects of neurocranial disorders, which will help inform treatment strategies.

Modern rehabilitation protocols of most neurodegenerative diseases, in particular the Parkinson Disease, rely on a clinical analysis of gait patterns. Currently, such analysis is highly dependent on both the examiner expertise and the type of evaluation. Development of evaluation methods with objective measures is then crucial. Physical models arise as a powerful alternative to quantify movement patterns and to emulate the progression and performance of specific treatments. This work introduces a novel quantification of the Parkinson disease progression using a physical model that accurately represents the main gait biomarker, the body Center of Gravity (CoG). The model tracks the whole gait cycle by a coupled double inverted pendulum that emulates the leg swinging for the single support phase and by a damper-spring System (SDP) that recreates both legs in contact with the ground for the double phase. The patterns generated by the proposed model are compared with actual ones learned from 24 subjects in stages 2,3, and 4. The evaluation performed demonstrates a better performance of the proposed model when compared with a baseline model(SP) composed of a coupled double pendulum and a mass-spring system. The Frechet distance measured differences between model estimations and real trajectories, showing for stages 2, 3 and 4 distances of 0.137, 0.155, 0.38 for the baseline and 0.07, 0.09, 0.29 for the proposed method.

An early diagnosis of Parkinson’s Disease (PD) is crucial towards devising successful rehabilitation programs. Typically, the PD diagnosis is performed by characterizing typical symptoms, namely bradykinesia, rigidity, tremor, postural instability or freezing gait. However, traditional examination tests are usually incapable of detecting slight motor changes, specially for early stages of the pathology. Recently, eye movement abnormalities have correlated with early onset of some neurodegenerative disorders. This work introduces a new characterization of the Parkinson disease by describing the ocular motion during a common daily activity as the gait. This paper proposes a fully automatic eye motion analysis using a dense optical flow that tracks the ocular direction. The eye motion is then summarized using orientation histograms constructed during a whole gait cycle. The proposed approach was evaluated by measuring the χ² distance between the orientation histograms, showing substantial differences between control and PD patients.

A reliable diagnosis of the Parkinson Disease lies on the objective evaluation of different motor sub-systems. Discovering specific motor patterns associated to the disease is fundamental for the development of unbiased assessments that facilitate the disease characterization, independently of the particular examiner. This paper proposes a new objective screening of patients with Parkinson, an approach that optimally combines ipsilateral global descriptors. These ipsilateral gait features are simple upper-lower limb relationships in frequency and relative phase spaces. These low level characteristics feed a simple SVM classifier with a polynomial kernel function. The strategy was assessed in a binary classification task, normal against Parkinson, under a leave-one-out scheme in a population of 16 Parkinson patients and 7 healthy control subjects. Results showed an accuracy of 94;6% using relative phase spaces and 82;1% with simple frequency relations.

In this article, we show the development of a low-cost hardware/software system based on close range photogrammetry to track the movement of a person performing weightlifting. The goal is to reduce the costs to the trainers and athletes dedicated to this sport when it comes to analyze the performance of the sportsman and avoid injuries or accidents. We used a web-cam as the data acquisition hardware and develop the software stack in Processing using the OpenCV library. Our algorithm extracts size, position, velocity, and acceleration measurements of the bar along the course of the exercise. We present detailed characteristics of the system with their results in a controlled setting. The current work improves the detection and tracking capabilities from a previous version of this system by using HSV color model instead of RGB. Preliminary results show that the system is able to profile the movement of the bar as well as determine the size, position, velocity, and acceleration values of a marker/target in scene. The average error finding the size of object at four meters of distance is less than 4%, and the error of the acceleration value is 1.01% in average.

An accurate right ventricular (RV) function quantification is important to support the evaluation, diagnosis and prognosis of several cardiac pathologies and to complement the left ventricular function assessment. However, expert RV delineation is a time consuming task with high inter-and-intra observer variability. In this paper we present an automatic segmentation method of the RV in MR-cardiac sequences. Unlike atlas or multi-atlas methods, this approach estimates the RV using exclusively information from the sequence itself. For so doing, a spatio-temporal analysis segments the heart at the basal slice, segmentation that is then propagated to the apex by using a non-rigid-registration strategy. The proposed approach achieves an average Dice Score of 0:79 evaluated with a set of 48 patients.

The comparison of several Level Set algorithms is performed with respect to 2D left ventricle segmentation in Multi-Slice CT images. Five algorithms are compared by calculating the Dice coefficient between the resulting segmentation contour and a reference contour traced by a cardiologist. The algorithms are also tested on images contaminated with Gaussian noise for several values of PSNR. Additionally an algorithm for providing the initialization shape is proposed. This algorithm is based on a combination of mathematical morphology tools with watershed and region growing algorithms. Results on the set of test images are promising and suggest the extension to 3{D MSCT database segmentation.

Even though CT is an imaging technique that offers high quality images, limitations on its spatial resolution cause blurring in small objects with high contrast. This phenomenon is known as blooming artifact and affects cardiac images with small calcifications and stents. This paper describes an approach to reduce the blooming artifact and improve resolution in cardiac images using blind deconvolution and anisotropic diffusion filtering. Deconvolution increases resolution but reduces signal-to-noise ratio, and the anisotropic diffusion filter counteracts this effect without affecting the edges in the image.

We present an Eulerian motion magnification technique with a spatial decomposition based on the Hermite Transform (HT). We compare our results to the approach presented in.1 We test our method in one sequence of the breathing of a newborn baby and on an MRI left ventricle sequence. Methods are compared using quantitative and qualitative metrics after the application of the motion magnification algorithm.

Insulin sensitivity (IS) is the ability of cells to react due to insulin´s presence; when this ability is diminished, low insulin sensitivity or insulin resistance (IR) is considered. IR had been related to other metabolic disorders as metabolic syndrome (MS), obesity, dyslipidemia and diabetes. IS can be determined using direct or indirect methods. The indirect methods are less accurate and invasive than direct and they use glucose and insulin values from oral glucose tolerance test (OGTT). The accuracy is established by comparison using spearman rank correlation coefficient between direct and indirect method. This paper aims to propose a lipid-anthropometric index which offers acceptable correlation to insulin sensitivity index for different populations (DB1=MS subjects, DB2=sedentary without MS subjects and DB3=marathoners subjects) without to use OGTT glucose and insulin values. The proposed method is parametrically optimized through a random cross-validation, using the spearman rank correlation as comparator with CAUMO method. CAUMO is an indirect method designed from a simplification of the minimal model intravenous glucose tolerance test direct method (MINMOD-IGTT) and with acceptable correlation (0.89). The results show that the proposed optimized method got a better correlation with CAUMO in all populations compared to non-optimized. On the other hand, it was observed that the optimized method has better correlation with CAUMO in DB2 and DB3 groups than HOMA-IR method, which is the most widely used for diagnosing insulin resistance. The optimized propose method could detect incipient insulin resistance, when classify as insulin resistant subjects that present impaired postprandial insulin and glucose values.

People with Complex Communication Needs (CCN) face difficulties to communicate their ideas, feelings and needs. Augmentative and Alternative Communication (AAC) approaches aim to provide support to enhance socialization of these individuals. However, there are many limitations in current applications related with systems operation, target scenarios and language consistency. This work presents an AAC approach to enhance produced messages by applying elements of Natural Language Generation. Specifically, a Spanish language engine, composed of a grammar ontology and a set of linguistic rules, is proposed to improve the naturalness in the communication process, when persons with CCN tell stories about their daily activities to non-disabled receivers. The assessment of the proposed method confirms the validity of the model to improve messages quality.

In the last years, pattern recognition has been applied to several fields for solving multiple problems in science and technology as for example in protein prediction. This methodology can be useful for prediction of activity of biological molecules, e.g. for determination of antimicrobial activity of synthetic and natural peptides. In this work, we evaluate the performance of different physico-chemical properties of peptides (descriptors groups) in the presence of imbalanced data sets, when facing the task of detecting whether a peptide has antimicrobial activity. We evaluate undersampling and class weighting techniques to deal with the class imbalance with different classification methods and descriptor groups. Our classification model showed an estimated precision of 96% showing that descriptors used to codify the amino acid sequences contain enough information to correlate the peptides sequences with their antimicrobial activity by means of learning machines. Moreover, we show how certain descriptor groups (pseudoaminoacid composition type I) work better with imbalanced datasets while others (dipeptide composition) work better with balanced ones.

Recent evidence suggests that healthy brain is organized on large-scale in regions spatially distant and partially temporally synchronized. These regions commonly are called Resting State Networks (RSNs). Many RSNs has been identified in multiples spatial scales in healthy subjects and their interactions has been used to define the functional network connectivity (FNC). The main idea in FNC is that the dynamic shown in the interactions among RSNs in control subjects, can change in pathological and pharmacological conditions. However, this hypothesis assumes that functional structure of healthy brain, remains in other brain states or conditions. In this work, we proposed a novel methodology in order to find the new brain functional structure for disorders of consciousness conditions, based on multi-objective optimization approach. Particularly, we find the best partition of RSNs set, that maximize two modularity measures (Kapur and Otsu measures). Our results suggest that the brain segregation level, may be linked to consciousness level.

Functional magnetic resonance imaging in resting state (fMRI-RS) constitutes an informative protocol to investigate several pathological and pharmacological conditions. A common approach to study this data source is through the analysis of changes in the so called resting state networks (RSNs). These networks correspond to well-defined functional entities that have been associated to different low and high brain order functions. RSNs may be characterized by using Independent Component Analysis (ICA). ICA provides a decomposition of the fMRI-RS signal into sources of brain activity, but it lacks of information about the nature of the signal, i.e., if the source is artifactual or not. Recently, a multiple template-matching (MTM) approach was proposed to automatically recognize RSNs in a set of Independent Components (ICs). This method provides valuable information to assess subjects at individual level. Nevertheless, it lacks of a mechanism to quantify how much certainty there is about the existence/absence of each network. This information may be important for the assessment of patients with severely damaged brains, in which RSNs may be greatly affected as a result of the pathological condition. In this work we propose a set of changes to the original MTM that improves the RSNs recognition task and also extends the functionality of the method. The key points of this improvement is a standardization strategy and a modification of method's constraints that adds flexibility to the approach. Additionally, we also introduce an analysis to the trustworthiness measurement of each RSN obtained by using template-matching approach. This analysis consists of a thresholding strategy applied over the computed Goodness-of-Fit (GOF) between the set of templates and the ICs. The proposed method was validated on 2 two independent studies (Baltimore, 23 healthy subjects and Liege, 27 healthy subjects) with different configurations of MTM. Results suggest that the method will provide complementary information for characterization of RSNs at individual level.

Sickle cell disease may result in neurological damage and strokes, leading to morbidity and mortality. Currently, there are no dependable biomarkers to predict impending strokes. In this study, we analyzed neuronal processes at resting state and more particularly how this disease affects the default mode network. The amplitude of low frequency fluctuations was used to reflect areas of spontaneous BOLD signal across brain regions. We compared the activations of sickle cell disease patients to a control group using variance analysis and t-test. Significant regional differences among the two groups were observed, especially in the default mode network areas and cortical regions near large cerebral arteries. These findings suggest that sickle cell disease causes activation modifications near vessels, and these changes could be used as a biomarker of the disease.

The removal of non-brain regions in neuroimaging is a critical task to perform a favorable preprocessing. The skull-stripping depends on different factors including the noise level in the image, the anatomy of the subject being scanned and the acquisition sequence. For these and other reasons, an ideal brain extraction method should be fast, accurate, user friendly, open-source and knowledge based (to allow for the interaction with the algorithm in case the expected outcome is not being obtained), producing stable results and making it possible to automate the process for large datasets. There are already a large number of validated tools to perform this task but none of them meets the desired characteristics. In this paper we introduced an open source brain extraction tool (OSBET), composed of four steps using simple well-known operations such as: optimal thresholding, binary morphology, labeling and geometrical analysis that aims to assemble all the desired features. We present an experiment comparing OSBET with other six state-of-the-art techniques against a publicly available dataset consisting of 40 T₁-weighted 3D scans and their corresponding manually segmented images. OSBET gave both: a short duration with an excellent accuracy, getting the best Dice Coefficient metric. Further validation should be performed, for instance, in unhealthy population, to generalize its usage for clinical purposes.

This article presents the development of a Cardiology Electronic Health Record (CEHR) system. Software consists of a structured algorithm designed under Health Level-7 (HL7) international standards. Novelty of the system is the integration of high resolution ECG (HRECG) signal acquisition and processing tools, patient information management tools and telecardiology tools. Acquisition tools are for management and control of the DIGICARDIAC electrocardiograph functions. Processing tools allow management of HRECG signal analysis searching for indicative patterns of cardiovascular pathologies. Telecardiology tools incorporation allows system communication with other health care centers decreasing access time to the patient information. CEHR system was completely developed using open source software. Preliminary results of process validation showed the system efficiency.

Grid computing has been successfully applied on teleradiology, leading to the creation of important platforms such as MEDICUS, VirtualPACS and mantisGRID, among others. These platforms are studied on the basis of their available documentation in order to compare and discuss differences and similarities, advantages and disadvantages between them. Then, a grid platform architecture is proposed, based on the best features of the surveyed platforms with an additional emphasis on general federated queries involving CBIR (Content-Based Image Retrieval) and Semantic Annotations.

Advances in medical imaging have fostered medical diagnosis based on digital images. Consequently, the number of studies by medical images diagnosis increases, thus, collaborative work and tele-radiology systems are required to effectively scale up to this diagnosis trend. We tackle the problem of the collaborative access of medical images, and present WebMedSA, a framework to manage large datasets of medical images. WebMedSA relies on a PACS and supports the ontological annotation, as well as segmentation and visualization of the images based on their semantic description. Ontological annotations can be performed directly on the volumetric image or at different image planes (e.g., axial, coronal, or sagittal); furthermore, annotations can be complemented after applying a segmentation technique. WebMedSA is based on three main steps: (1) RDF-ization process for extracting, anonymizing, and serializing metadata comprised in DICOM medical images into RDF/XML; (2) Integration of different biomedical ontologies (using L-MOM library), making this approach ontology independent; and (3) segmentation and visualization of annotated data which is further used to generate new annotations according to expert knowledge, and validation. Initial user evaluations suggest that WebMedSA facilitates the exchange of knowledge between radiologists, and provides the basis for collaborative work among them.

Ultrasound (US) images of the fetal brain provide the experts with valuable indicators of the fetal development. However as the skull thickens, it obstructs the transmission of the acoustic waves, which in turn occludes the anatomy behind the thickened fetal skull. A viable option to improve the visibility of the fetal brain, before complete calcification of the skull, is the calculation of a compounded image made of different views of the same anatomical plane. In this work we report a new method for the composition of ultrasound images based on the Weighted Mean of the pixels, from different views, which correspond to each position (x, y) in the final compounded image. A support vector machine (SVM) is used to calculate the weights of each pixel from a different view, based on intensity, entropy and variance features. We present the initial test results of our method on synthetic US images of a head phantom, contaminated with speckle noise; we report the signal to noise ratio (SNR) and the normalized mutual information (NMI), for different number of views (2, 3, and 5), and compare the results against images compounded using the Mean, Root Mean Square (RMS), and Geometrical Mean composition methods. With our scheme we were able to recover the occluded information to increase the NMI from 16% to 26%, representing a 58% improvement.

According to World Health Organization, pneumonia is the respiratory disease with the highest pediatric mortality rate accounting for 15% of all deaths of children under 5 years old worldwide. The diagnosis of pneumonia is commonly made by clinical criteria with support from ancillary studies and also laboratory findings. Chest imaging is commonly done with chest X-rays and occasionally with a chest CT scan. Lung ultrasound is a promising alternative for chest imaging; however, interpretation is subjective and requires adequate training. In the present work, a two-class classification algorithm based on four Gray-level co-occurrence matrix texture features (i.e., Contrast, Correlation, Energy and Homogeneity) extracted from lung ultrasound images from children aged between six months and five years is presented. Ultrasound data was collected using a L14-5/38 linear transducer. The data consisted of 22 positive- and 68 negative-diagnosed B-mode cine-loops selected by a medical expert and captured in the facilities of the Instituto Nacional de Salud del Niño (Lima, Peru), for a total number of 90 videos obtained from twelve children diagnosed with pneumonia. The classification capacity of each feature was explored independently and the optimal threshold was selected by a receiver operator characteristic (ROC) curve analysis. In addition, a principal component analysis was performed to evaluate the combined performance of all the features. Contrast and correlation resulted the two more significant features. The classification performance of these two features by principal components was evaluated. The results revealed 82% sensitivity, 76% specificity, 78% accuracy and 0.85 area under the ROC.

Cardiovascular events are one of the main causes of death in the world. Considering that most of these events are produced by asymptomatic lesions, early and noninvasive analysis and detection of hints for such afflictions are quite desirable. In this work, we study a method known as Lagrangian Speckle Model Estimator (LSME) for assessing the full strain tensor that describes tissue motion between two states during vascular pulse. We then evaluate the results of the LSME method using computer generated images, varying certain mechanical and computational parameters. The strain maps obtained were then assessed, showing percent variation within range of the exact solution. Qualitatively, strain patterns were similar to the analytical solution. Finally, a discussion about further improvements and error causes is carried out.

The cerebellum is an important structure to determine the gestational age, cerebellar diameter obtained by ultrasound volumes of the fetal brain has shown a high correlation with gestational age, therefore is useful to determine fetal growth restrictions. The manual annotation of 3D surfaces from the fetal brain is time consuming and needs to be done by a highly trained expert. To help with the annotation in the evaluation of cerebellar diameter, we developed a new automatic scheme for the segmentation of the 3D surface of the cerebellum in ultrasound volumes, using a spherical harmonics model and the optimization of an objective function based on gray level voxel profiles. The results on 10 ultrasound volumes of the fetal brain show an accuracy in the segmentation of the cerebellum (mean Dice coefficient of 0.7544). The method reported shows potential to effectively assist the experts in the assessment of fetal growth in ultrasound volumes. We consider the proposed cerebellum segmentation method a contribution for the SPHARM segmentations models, because it is capable to run without hardware restriction, (GPU), and gives adequate results in a reasonable amount of time.

One of the limitations of Encephalography (EEG) data is its quality, as it is usually contaminated with electric signal from muscle. This research intends to study results of two EEG source analysis methods applied to scalp recordings taken in paralysis and in normal conditions during the performance of a cognitive task. The aim is to determinate which types of analysis are appropriate for dealing with EEG data containing myogenic components. The data used are the scalp recordings of six subjects in normal conditions and during paralysis while performing different cognitive tasks including the oddball task which is the object of this research. The data were pre-processed by filtering it and correcting artefact, then, epochs of one second long for targets and distractors were extracted. Distributed source analysis was performed in BESA Research 6.0, using its results and information from the literature, 9 ideal locations for source dipoles were identified. The nine dipoles were used to perform discrete source analysis, fitting them to the averaged epochs for obtaining source waveforms. The results were statistically analysed comparing the outcomes before and after the subjects were paralysed. Finally, frequency analysis was performed for better explain the results. The findings were that distributed source analysis could produce confounded results for EEG contaminated with myogenic signals, conversely, statistical analysis of the results from discrete source analysis showed that this method could help for dealing with EEG data contaminated with muscle electrical signal.

When children are taken apart from their parents because of many violence situations, they are taken to foster homes, where they share place with kids who have lived similar situations. United Nations Children’s Fund (2014) refer that Children who have been abused or neglected, often may have low self-esteem and other emotional problems, which can lead, at worst, to risky behaviors and self-harm . They also could tend to internalize that behavior, repeating the pattern of violence and abuse as a response to their environment. In this line, the latest estimates provided by SOS Children's Village International show a global complex picture: around 24 million of children in the world live in foster homes, one billion of children live in conflict-affected areas; and, furthermore, there is a lack of mental health professionals in most of the countries. On those grounds, in this paper we propose an intelligent ecosystem to provide support for psychologists during the psychodiagnosis and intervention with children, especially the ones who are in foster homes. Currently, the system is able to automatically determine some psychological traits, according to responses provided by each patient. One part of the diagnostic system is based on two psychological tests: the Draw-A-Person test and the Draw-A-Family test. The results obtained on the first stage let the system establish different challenges according to the skills that the evaluated child needs to develop. Our proposed approach was tested in a population of 124 children (93 school students, and 31 living in shelters), and has achieved encouraging results (80% of precision in patient's profile determination).

This paper reports a comparison between three fetal ECG (fECG) detectors developed during the CinC 2013 challenge for fECG detection. Algorithm A1 is based on Independent Component Analysis, A2 is based on fECG detection of RS Slope and A3 is based on Expectation-Weighted Estimation of Fiducial Points. The proposed methodology was validated using the annotated database available for the challenge. Each detector was characterized in terms of its performance by using measures of sensitivity, (Se), positive predictive value (P+) and delay time (t_d). Additionally, the database was contaminated with white noise for two SNR conditions. Decision fusion was tested considering the most common types of combination of detectors. Results show that the decision fusion of A1 and A2 improves fQRS detection, maintaining high Se and P+ even under low SNR conditions without a significant t_d increase.

Heart diseases are the main cause of death worldwide. The first step in the diagnose of these diseases is the analysis of the electrocardiographic (ECG) signal. In turn, the ECG analysis begins with the detection of the QRS complex, which is the one with the most energy in the cardiac cycle. Numerous methods have been proposed in the bibliography for QRS complex detection, but few authors have analyzed the possibility of taking advantage of the information redundancy present in multiple ECG leads (simultaneously acquired) to produce accurate QRS detection. In our previous work we presented such an approach, proposing various data fusion techniques to combine the detections made by an algorithm on multiple ECG leads. In this paper we present further studies that show the advantages of this multi-lead detection approach, analyzing how many leads are necessary in order to observe an improvement in the detection performance. A well known QRS detection algorithm was used to test the fusion techniques on the St. Petersburg Institute of Cardiological Technics database. Results show improvement in the detection performance with as little as three leads, but the reliability of these results becomes interesting only after using seven or more leads. Results were evaluated using the detection error rate (DER). The multi-lead detection approach allows an improvement from DER = 3:04% to DER = 1:88%. Further works are to be made in order to improve the detection performance by implementing further fusion steps.

This paper reports an automatic method for characterizing the quality of the RR-time series in the stress test database known as DICARDIA. The proposed methodology is simple and consists in subdividing the RR time series in a set of windows for estimating the quantity of artifacts based on a threshold value that depends on the standard deviation of RR-time series for each recorded lead. In a first stage, a manual annotation was performed considering four quality classes for the RR-time series (Reference lead, Good Lead, Low Quality Lead and Useless Lead). Automatic annotation was then performed varying the number of windows and threshold value for the standard deviation of the RR-time series. The metric used for evaluating the quality of the annotation was the Matching Ratio. The best results were obtained using a higher number of windows and considering only three classes (Good Lead, Low Quality Lead and Useless). The proposed methodology allows the utilization of the online available DICARDIA Stress Test database for different types of research.

This paper introduces a robust method based on the Support Vector Machine (SVM) algorithm to detect the presence of Fetal QRS (fQRS) complexes in electrocardiogram (ECG) recordings provided by the PhysioNet/CinC challenge 2013. ECG signals are first segmented into contiguous frames of 250 ms duration and then labeled in six classes. Fetal segments are tagged according to the position of fQRS complex within each one. Next, segment features extraction and dimensionality reduction are obtained by applying principal component analysis on Haar-wavelet transform. After that, two sub-datasets are generated to separate representative segments from atypical ones. Imbalanced class problem is dealt by applying sampling without replacement on each sub-dataset. Finally, two SVMs are trained and cross-validated using the two balanced sub-datasets separately. Experimental results show that the proposed approach achieves high performance rates in fetal heartbeats detection that reach up to 90.95% of accuracy, 92.16% of sensitivity, 88.51% of specificity, 94.13% of positive predictive value and 84.96% of negative predictive value. A comparative study is also carried out to show the performance of other two machine learning algorithms for fQRS complex estimation, which are K-nearest neighborhood and Bayesian network.