This PDF file contains the front matter associated with SPIE Proceedings Volume 11583 including the Title Page, Copyright information, and Table of Contents.

Automatic segmentation of white matter hyperintensities in magnetic resonance images is of paramount clinical and research importance. Quantification of these lesions serve as a predictor for risk of stroke, dementia and mortality. During the last years, convolutional neural networks (CNN) specifically tailored for biomedical image segmentation have outperformed all previous techniques in this task. However, they are extremely data dependent, and maintain a good performance only when data distribution between training and test datasets remains unchanged. When such distribution changes but we still aim at performing the same task, we incur in a domain adaptation problem (e.g. using a different MR machine or different acquisition parameters for training and test data). In this work, we explore the use of cycle-consistent adversarial networks (CycleGAN) to perform unsupervised domain adaptation on multicenter MR images with brain lesions. We aim at learning a mapping function to transform volumetric MR images between domains, which are characterized by different medical centers and MR machines with varying brand, model and configuration parameters. Our experiments show that CycleGAN allows us to reduce the Jensen-Shannon divergence between MR domains, enabling automatic segmentation with CNN models on domains where no labeled data was available.

BrainAge (a subject’s apparent age predicted from neuroimaging data) is an important biomarker of brain aging. The deviation of BrainAge from true age has been associated with psychiatric and neurological disease, and has proven effective in predicting conversion from mild cognitive impairment (MCI) to dementia. Conventionally, 3D convolutional neural networks and their variants are used for brain age prediction. However, these networks have a larger number of parameters and take longer to train than their 2D counterparts. Here we propose a 2D slice-based recurrent neural network model, which takes in an ordered sequence of sagittal slices as input to predict the brain age. The model consists of two components: a 2D convolutional neural network (CNN), which encodes the relevant features from the slices, and a recurrent neural network (RNN) that learns the relationship between slices. We compare our method to other recently proposed methods, including 3D deep convolutional regression networks, information theoretic models, and bag-of-features (BoF) models (such as BagNet) - where the classification is based on the occurrences of local features, without taking into consideration their global spatial ordering. In our experiments, our proposed model performs comparably to, or better than, the current state of the art models, with nearly half the number of parameters and a lower convergence time.

Collectively, vast quantities of brain imaging data exist across hospitals and research institutions, providing valuable resources to study brain disorders such as Alzheimer’s disease (AD). However, in practice, putting all these distributed datasets into a centralized platform is infeasible due to patient privacy concerns, data restrictions and legal regulations. In this study, we propose a novel federated feature selection framework that can analyze the data at each individual institution without data-sharing or accessing private patient information. In this framework, we first propose a federated group lasso optimization method based on block coordinate descent. We employ stability selection to determine statistically significant features, by solving the group lasso problem with a sequence of regularization parameters. To accelerate the stability selection, we further propose a federated screening rule, which can identify and exclude the irrelevant features before solving the group lasso. Here, we use this framework for patch based feature selection on hippocampal morphometry. Shape is characterized through two different kinds of local measures, the radial distance and the surface area determined via tensor-based morphometry (TBM). The method is tested on 1,127 T1-weighted brain magnetic resonance images (MRI) of AD, mild cognitive impairment (MCI) and elderly control subjects, randomly assigned to five independent hypothetical institutions for testing purpose. We examine the association of MRI-based anatomical measures with general cognitive assessment and amyloid burden to identify the morphometry changes related to AD deterioration and plaque accumulation. Finally, we visualize the significance of the association on the hippocampal surfaces. Our experimental results successfully demonstrate the efficiency and effectiveness of our method.

This work explores the spatial coherence of magnetic patterns at a microanatomical scale, using a texture descriptor (LBP) which captures local spatial relationships between levels of intensity in Magnetic Resonance Images (MRI). The relative frequency of these texture micropatterns was estimated and used as a regional characterization according to the Harvard Oxford anatomical parcellation. This characterization was compared for different groups of age in healthy and patients with Alzheimer’s disease. A statistical analysis was conducted by partitioning the total sample age in sub intervals of fixed length. Subsequently, an ANOVA analysis was performed to determine the existence of statistically significant differences between these age ranges, as well as a multiple comparison of these age ranges characterization using a Tukey Test. Finally, based on similarity of these comparisons, a grouping of the considered regions was achieved. The method was also applied to group of subjects with Alzheimer’s Disease. A different distribution of these texture patterns was observed between cortical and subcortical areas while patients with Alzheimer’s disease showed unique differences at the amygdala level while the hippocampus appeared as part of a group that also expressed differences.

Modeling of structural brain variation over the lifespan is important to better understand factors contributing to healthy aging and risk for neurological conditions such as Alzheimer’s disease. Even so, we lack normative data on brain morphometry across the adult lifespan in large, well-powered samples. Here, in a large population-based sample of 26,440 adults from the UK Biobank (age: 44-81 yrs.), we created normative percentile charts for MRI-derived subcortical volumes. Next, we investigated associations between these morphometric measures and the strongest known genetic risk factor for late-onset Alzheimer’s disease (APOE genotype) and mapped the spatial distribution of age-by-sex interactions using computational surface mesh modeling and shape analysis. Vertex-wise shape mapping supplements traditional gross volumetric approaches to reveal finer-grained variations across functionally important brain subcompartments. Normative curves revealed volumetric loss with age, as expected, for all subcortical brain structures except for the lateral ventricles, which expanded with age. Surprisingly, no volumetric associations with APOE genotype were detected, despite the very large sample size. Age-related trajectories for volumes differed in women versus men, and surface-based statistical maps revealed the spatial distribution of the age-by-sex interaction. Subcortical volumes declined faster in men than women over the full age range, but after age 60, fewer structures showed sex-dependent trajectories, indicating similar volumetric changes in older men and women. Large-scale statistical modeling of age effects on brain structures may drive new insights into individual differences in brain aging and help to identify factors that promote healthy brain aging and risk for disease.

The brain’s white matter microstructure, as assessed using diffusion-weighted MRI (DWI), changes significantly with age and also exhibits significant sex differences. Here we examined the ability of a traditional diffusivity metric (fractional anisotropy derived from diffusion tensor imaging, DTI-FA) and advanced diffusivity metrics (fractional anisotropy derived from the tensor distribution function, TDF-FA; neurite orientation dispersion and density imaging measures of intracellular volume fraction, NODDI-ICVF; orientation dispersion index, NODDI-ODI; and isotropic volume fraction, NODDI-ISOVF) to detect sex differences in white matter aging. We also created normative aging reference curves based on sex. Diffusion tensor imaging (DTI) applies a single-tensor diffusion model to single-shell DWI data, while the tensor distribution function (TDF) fits a continuous distribution of tensors to single-shell DWI data. Neurite orientation dispersion and density imaging (NODDI) fits a multi-compartment model to multi-shell DWI data to distinguish intra- and extracellular contributions to diffusion. We analyzed these traditional and advanced diffusion measures in a large population sample available through the UK Biobank (15,394 participants; age-range: 45-80 years) by using linear regression and fractional polynomials. Advanced diffusivity metrics (NODDI-ODI, NODDI-ISOVF, TDF-FA) detected significant sex differences in aging, whereas a traditional metric (DTI-FA) did not. These findings suggest that future studies examining sex differences in white matter aging may benefit from including advanced diffusion measures.

This study used advanced diffusion-weighted MRI (dMRI) to examine the association between exogenous sex-hormone exposure and the brain’s white matter aging trajectories in a large population-based sample of women. To investigate the effect of pre- and post-menopausal sex hormones on brain aging, cross-sectional brain dMRI data from the UK Biobank was analyzed using 3 diffusion models: conventional diffusion tensor imaging (DTI), the tensor distribution function (TDF), and neurite orientation dispersion and density imaging (NODDI). Mean skeletonized diffusivity measures were extracted and averaged across the whole brain, including fractional anisotropy, isotropic volume fraction, intracellular volume fraction and orientation dispersion index. We used general linear models and fractional polynomial regressions to characterize age-related trajectories in white matter measures following hormone therapy (HT) and oral contraceptive (OC) use in women (HT analysis: N=8,301; OC analysis: N=8,913). Sex hormone treatment (HT and OC) was statistically associated with the aging trends in white matter measures. Estrogen therapy alone appeared to exert a neuroprotective effect on age-related white matter processes, compared to HT containing both estrogen and progestin therapy - which was associated with accelerated aging-related processes in women. These results support the hypothesis that exogenous sex hormone exposure may impact white matter aging; white matter metrics may also be sensitive to sex hormone levels in women. Furthermore, we discuss the necessity to test alternative models for lifespan trajectories beyond popular linear and quadratic models, especially when dealing with large samples. Fractional polynomial models may provide a more adaptive alternative to linear or quadratic models.

Prior studies show that obesity is associated with accelerated brain aging and specific patterns of brain atrophy. Finerscale mapping of the effects of obesity on the brain would help to understand how it promotes or interacts with disease effects, but so far, the influence of the obesity on finer-scale maps of anatomy remains unclear. In this study, we propose a deep transfer learning network based on Optimal Mass Transport (OMTNet) to classify individuals with normal versus overweight/obese body mass index (BMI) using vertex-wise brain shape metrics extracted from structural MRI scans from the UK Biobank study. First, an area-preserving mapping was used to project 3D brain surface meshes onto 2D planar meshes. Vertex-wise maps of brain metrics such as cortical thickness were mapped into 2D planar images for each brain surface extracted from each person’s MRI scan. Second, several popular networks pretrained on the ImageNet database, i.e., VGG19, ResNet152 and DenseNet201, were used for transfer learning of brain shape metrics. We combined all shape metrics and generated a metric ensemble classification, and then combined all three networks and generated a network ensemble classification. The results reveal that transfer learning always outperforms direct learning, and we obtained accuracies of 65.6±0.7% and 62.7±0.7% for transfer and direct learning in the network ensemble classification, respectively. Moreover, surface area and cortical thickness, especially in the left hemisphere, consistently achieved the highest classification accuracies, together with subcortical shape metrics. The findings suggest a significant and classifiable influence of obesity on brain shape. Our proposed OMTNet method may offer a powerful transfer learning framework that can be extended to other vertex-wise brain structural and functional imaging measures.

Generation of white matter (WM) tractography for neonates primarily depends on a successful development of a diffuse tensor imaging (DTI)-based ATLAS. In this study, we present a deep-learning framework for WM tractography of neonates’ brain that is independent of any specific ATLAS. A convolutional neural network (CNN)-based deep-learning architecture is proposed for automated generation of WM tractography. Our dataset consists of DTI scan of 37 neonates (18 preterm and 19 term-born) that can be used to train the model. Although the proposed model is adopted for WM tractography, it can generally be applied for subcortical structures and cerebellum.

The need to make timely and accurate diagnoses of brain diseases has posed challenges to computer-aided diagnosis systems. In this field, advances in deep learning techniques play an important role, as they carry out processes to extract relevant anatomical and functional characteristics of the tissues to classify them. In this paper, the study of various architectures of convolutional neural networks (CNN) is presented, with the aim of classifying three types of brain tumors in high-contrast magnetic resonance (MR) images. The architectures of the present study were VGG16, ResNet50, Xception, whose implementations are defined in the Keras framework. The evaluation of these architectures were preceded by data augmentation techniques and transfer learning, which improved the effectiveness of the training process, thanks to the use of pre-trained models with the ImageNet dataset. The VGG16 architecture was the one with the best performance, with an accuracy of 98.04%, followed by ResNet50 with 94.89%, and finally, Xception with 92.18%.

This paper presents a new method for coronary artery segmentation in X-ray angiograms based on deep learning and a patch-based training. The blood vessel segmentation is performed using the U-Net convolutional neural network, which has been trained using patches extracted from the original angiograms instead of using complete images. The publicly available database of coronary angiograms DCA1 containing 130 angiograms with their respective ground-truth has been used to generate the training patterns and subsequently to evaluate and compare the segmentation performance of the proposed method. The hyper-parameter configuration used for training the U-Net parameters has been selected from 90 possible combinations according to five binary classification metrics. Each combination involving the selection of a patch size, weight assigned to the blood vessel class, and learning rate used by the optimization method, has been used in order to train the U-Net parameters with patterns extracted from a set of 100 images. The segmentation performance of the proposed method is compared with five specialized blood vessel segmentation methods from the state of the art using a test set of 30 images, achieving the highest accuracy (0.977) and Dice similarity coefficient (0.779). Moreover, the experimental results have also shown that the proposed method is suitable to be integrated into a computer-aided system to support decision making in medical practice.

Medical ultrasound has gained relevance in medical diagnosis, procedure guidance and emergentology. It is a low cost, safe and non-invasive practice, while obtaining real-time results. Limitations on working hours and ultrasound equipment have made simulation the main alternative for training, since it eliminates the need for a real ultrasound machine and allows the modeling of a large number of case studies. Patient 11 from the 3D-IRCADb-01 CT-scans data-set was used to generate corresponding OBJ mesh files. A JSON file carries relevant information of the scene and tissues involved in it. Monte-Carlo simulation over GPU is used to improve ray tracing performance on a deterministic surface model. An improvement in ray tracing performance and throughput is achieved by the presented model. In execution cases presented, a minimum global acceleration of 30% is reached, while reducing ray tracing times by at least 80%. Monte-Carlo simulation allows a more realistic image generation than the previous deterministic model. Low memory usage gives room to implement the remaining stages of the process on GPU. A correct parametrization of tissues is required in order to achieve improved image quality.

Accurate automatic segmentation of the prostate in ultrasound images is still a challenging research problem. In this work, we propose the use of gray level images, constructed with a sample of gray level profiles perpendicular to the contour of the prostate. A two dimensional principal component analysis (2D PCA) was performed on a set of training contour images. The reconstruction error from the 2D PCA was used as an objective function for automatic adjustment of a point distribution model of the prostate. Our method was validated on 9 ultrasound images of the prostate and compared to the optimization of an objective function based on the mean Mahalanobis distance of a sampled gray level profile to the corresponding statistical profile model. Our new method based on a 2D PCA shows improved prostate segmentation results.

Ultrasound has been considered a safe and accurate alternative to radiography and computerized tomography to diagnose lung diseases such as pneumonia. However, speckle noise, artifacts or certain conditions can difficult image interpretation. For example, in some cases, the pleura line cannot be observed. This work proposes an approach for discriminating between lung and muscular tissues in ultrasound images. We evaluated the symlet and daubechies wavelets for feature extraction, principal component analysis and recursive backward elimination for feature selection, and supervised learning methods for classification. Statistical moments and the energy of the second horizontal coefficient and peak-to-peak root mean squared ratio were the features more outstanding over the rest. The best model was obtained with recursive backward elimination for feature selection and knearest neighbor for classification. Tissue classification was possible with a mean accuracy of 97.5% and area under the curve of 99%. These results offer great insights on the recognition of lung and muscular tissues, which could improve the effectiveness of automatic segmentation and analysis algorithms.

Pancreatic Cancer (PC) is a very aggressive cancer, with a mortality of 0.98 and a 5-year survival rate of 6.7%.^1–3 Endoscopic ultrasonography (EUS) is the imaging modality to early detection of PC. Its reported diagnosis sensitivity for an experienced gastroenterologist ranges from 87 to 100%.^3–5 Computational strategies, as Elastography, have been developed to support mass malignancy diagnosis. However, most studies evaluate their strategies using private datasets, making results incomparable. This work presents an annotated open access database of Endoscopy Ultrasound videos obtained in the Gastroenterology Unit of the Hospital Universitario Nacional de Colombia and the Unidad de Gastroenterolog´ıa y Ecoendoscopia. The dataset consists in a set of 55 cases acquired in B-mode Ultrasound image, composed of 18 cases with pancreatic cancer, 5 cases with pancreatitis, and 32 cases that include healthy pancreas, liver and gallbladder. Cases were confirmed and staged by pathological examination from biopsy samples and manually annotated per each video frame. Additionally, herein it is presented a preprocessing methodology aimed to highlight the useful echo patterns to differentiate pancreatic diseases.

Retinal diseases are a common cause of blindness around the world, early detection of clinical findings can help to avoid vision loss in patients. Optical coherence tomography images have been widely used to diagnose retinal diseases, due to the capacity to show in detail findings as drusen, hyperreflective foci, and intraretinal and subretinal fluids. The location of findings is vital to identify and follow-up the retinal disease. However, the detection and segmentation of these findings is not an easy task due to artifacts noise, and the time consuming even to experts ophthalmologist. This paper proposes a computational method based on deep learning to automatically identify fluids and hyperreflective foci as a tool to identify retinal diseases through the use of OCT images. The method was evaluated on a set of OCT images manually annotated by experts. The experimental results present a Dice coefficient of 0,4437 and 0,6245 in the segmentation task of fluids (intrarretinal fluids and subretinal fluids), and hyperreflective foci respectively.

Diabetic retinopathy is the most common cause of blindness in population in developed countries. However, with early diagnosis of this asymptomatic disease, it could be prevented 80 percent of cases. Diabetic retinopathy is detected in fundus images, also called retinal digital angiography. In most cases, due to the concave geometry of the eye, these images have a high variability in local contrast and luminance. This lack of uniformity may mask signs of this disease as ocular hemorrhages, microaneurysms, hard exudates and cotton wool spots affecting the diagnostic quality. This paper presents an automatic method for the segmentation of hard exudates and cotton wool in fundus images based on Mathematical Morphology in color spaces. The aim of this development is to assist the expert in the diagnosis of disease. To validate the proposed method, the Diabetic Retinopathy Database and the Evaluation Protocol diaRetdB0 and diaRetdB1 were used. The proposed method archived values above 96.5%. This performance was compared with a technique developed by other authors, obtaining a difference of above 26.5% of true positives in favor of our method.

In this paper, we aim to assist the measurement of the volume of lower-limb muscles from 3D freehand ultrasound images. Volume estimation typically requires a very time-consuming manual segmentation step. To facilitate and speed-up the volume measurements, in this paper, we propose a non-learning based approach that, starting from sparse annotations in 2D images, propagates the labels to the full volume. Furthermore, we rely on 3D3D image-based registration to combine labels from different ultrasound acquisitions of the same muscle with different qualities. Our goal here is to provide a simple and low-cost solution, relying mainly on open-source software for the processing steps. The proposed approach effectively reduces the manual interaction time while providing reasonable estimations for the segmentation and volume calculation. We achieve a mean dice score of 0.89±0.03 and a mean volumetric measure error of 4,18%. The resultant volumes may also be useful for building augmented annotated databases to develop automatic learning-based segmentation approaches.

This study presents a novel strategy to characterize and remove non-nuclei signal (noise) in histopathological images stained with hematoxylin and eosin (H and E), a preprocessing step to improve traditional nuclei segmentation methods. Any non nuclei structure is mapped to a noiselet space at different resolution levels where a classic classifier is trained to recognize the noiselet coefficients of this projection. The proposed approach was evaluated with two multi-organ datasets manually annotated, comparing the nuclei segmentation obtained by a watershed algorithm plus the presented approach against the watershed method alone. An average Dice Score in these datasets (MICCAI Challenge and TCIA) of 70.2 and 59.6 was obtained by applying the herein introduced method, while the obtained Dice Score with only the watershed method was of 66.7 and 55.5.

The variation in quantitative measures extracted from computed tomography (CT) perfusion parametric maps due to changes in dose was evaluated. A CT perfusion phantom was scanned on a Philips CT scanner using AAPM recommendations at 2 different speeds and varying x-ray exposure. The acquired images were post-processed using the TeraRecon software. The software outputted Cerebral Blood Flow (CBF), Cerebral Blood Volume (CBV), Mean Transit Time (MTT) maps and Time Attenuation Curves (TAC) of the artery and the vein rods of the phantom. Measurements were made in regions of interest (ROIs) in the two tissue rods (foreground) and 5 regions in the background, across the different parametric maps, respectively. Mixed effect model with AR (1) covariance structure was used to compare measurements across different dose levels as repeated measured random effect. Dunnet adjustment was used for posthoc pairwise comparisons. For the foreground ROI, no significant changes in the measured mean CBF, CBV, and MTT values were observed with changes in dose. As expected, the standard deviation (SD) of CBF and CBV decreased as dose increased. At each dose, higher speed settings were consistently associated with higher SD of CBF and lower MTT. For the background ROI, the measured mean CBF and CBV were significantly higher at lower dose levels, and the SD of CBF decreased as the dose increased. The MTT of the background did not vary with dose. We conclude that radiation dose affects perfusion metrics especially for low or no flow conditions.

Currently, cancer is the leading cause of death worldwide, making millions of deaths annually in developing countries due to a shortage of detection and treatment. Early detection of cancer neoantigens is useful for specialists because they can help in the development of more successful treatments. Based on this problem, the objective of this work is to carry out a comparative process between machine learning models, to determine which of them allows an adequate prediction of the data, and thus determine the carcinogenic neoantigens. For this, information extracted from protein sequences was employed. The preliminary results show sensitivity and specificity of 1.0 and 0.98 respectively.

This work presents a watermarking algorithm applied to medical images of COVID-19 patients, intending to preserve its diagnose and that it not be modified when watermark will be inserted. Besides, we tried to protect the information of the patient using an imperceptible watermarking. Our technique is based on a perceptive approach to insert the watermark by decomposing the medical image using the Hermite transform. We use as watermark two image logos, including text strings to demonstrate that the watermark can contain relevant information of the patient. Some metrics were applied to evaluate the performance of the algorithm. Finally, we present some results about robustness with some attacks applied to watermark images.

COVID-19 disruptions to the global supply chain has caused a shortage of personal protective equipment (PPE) at healthcare facilities world-wide. Traditional manufacturers are unable to fill the demand for equipment. Additive manufacturing has been able to help slacken the need by creating novel PPE. Over 8000 reusable filtered masks, over 10000 face shields, and testing equipment were created with the combined efforts of healthcare organizations, academic institutions, community maker organizations, and creative industry partners.

Atrial fibrillation (AF) is the most common cardiac arrhythmia in the world which shows rising prevalence leading to increased comorbidities, such as, Ischemic heart disease and Stroke that the main cause of deaths in the world. Since AF and most of the arrhythmias are generated due to electrical problems at the heart, electrocardiography provides the best noninvasive method to diagnose and QRS complex play an important role as a benchmark. In this paper, a novel methodology for QRS complex detection is presented. The algorithm introduces a modification of the well known Pan Tompkins approach, performing a multi channel detection, based on the signal to noise ratio of every channel. After application of the squaring operation in the channels with the highest signal to noise ratio a new single channel is created with improved quality, allowing the accurate detection of the QRS complexes in signals with atrial arrhythmias. The approach was tested in electrocardiography records from the Hospital Universitario de Valencia in Spain, showing an average positive predictive value of 99.6% and an average sensitivity of 99.9%.

In the field of continuous vital-sign monitoring in critical care settings, it has been observed that the “earlywarning signs” of impending physiological deterioration can fail to be detected timely and sometimes by resourceconstrained clinical staff. This effect may be escalated by the “data deluge” caused by acquisition of more complex patient data during routine care. The objective of this study is to develop a probabilistic model for predicting the future clinical episodes of a patient using observed vital sign values prior to a clinical event. Vital signs (e.g. heart rate, blood pressure) are used to monitor a patient’s physiological functions and their simultaneous changes indicate transitions between patient’s health states. If such changes are abnormal then it may lead to serious physiological deterioration. The CRISP-DM (CRoss-Industry Standard Process for Data Mining) methodology was used as a data mining process and then we use Markov chains to identify the clinical states through which the patient passes. Then, a Hidden Markov model (HMM) based approach is applied for classification and prediction of patient’s deterioration by computing the probability of future clinical states. Both learning models were trained and evaluated using six vital signs data from 94,678 patient records, collected from the database of real patients who were in the Pediatric Intensive Care Unit of the Central Military Hospital in the city of Bogot´a, Colombia. The proposed technique based on monitoring multiple physiological variables showed promising results in early identifying the deterioration of critically ill patients.

In this paper, we present a new non-contact strategy to estimate the breathing rate based on the Eulerian motion video magnification technique and an Artificial Hydrocarbon Networks (AHN) as classifier. After the magnification procedure, a AHN is trained to detect the inhalation and exhalation frames in the video. From this classification, the respiratory rate is estimated. The magnification procedure was carried out using the Hermite decomposition. The respiratory rate (RR) is estimated from the classified frames. We have tested the method on 10 healthy subjects in different positions. To compare performance of methods to respiratory rate the mean average error and a Bland and Altman analysis is used to investigate the agreement of the methods. The mean average error for our strategy is 4.46 ± 3.68% with and agreement with respect of the reference of ≈ 98%.

Attention deficit and hyperactivity disorder (ADHD) is a medical condition that affects approximately 7% of children worldwide. The diagnosis of ADHD can be done using psychological tests and electroencephalography (EEG). However, the variability and complexity of EEG signals affects its diagnostic utility. The purpose of this work is to identify relevant features of EEG signals from children diagnosed with ADHD and control cases for their classification. A total of 47 children were included in the study (22 with ADHD and 25 in the control group). EEG of cognitive evoked potentials were preprocessed using wavelet filtering and synchronized averaging. Then, fourteen features were calculated in signals from four channels (F3, AF3, F4 and AF4), including evoked potentials, power spectrum, entropy, chaos, bicoherence measures, and prominent peaks. For feature selection, the algorithms PCA, hybrid stepwise regression, ridge regression, and correlation values were evaluated. It was evidenced that evoked potentials have a relative high level of importance, as well as the prominent peaks. On the other hand, the values of chaos and bicoherence measures, along with the gender, are the least representative features. These results are consistent among the four feature selection algorithms. In conclusion, 9 of the 14 features are representative of the data set and were used for the classification stage of this work.

Cross-modality synthesis represent nowadays a promising application in medical image processing to manage the problem of paired data scarcity. In this work we designed and trained a CycleGAN model to generate PET/CT data from 2D slices collected from the liver body region of twelve patients. The results obtained from the six test patients show how our model can outperform baseline CycleGAN framework and effectively be used for synthesizing artificial images to be used for data augmentation or dataset completion.

Although musical neurofeedback is used in multiple works, few systems have been developed for stress regulation, and no systems have been developed for memory stimulation. For this reason, a music-based neurofeedback system for stress regulation and memory stimulation is proposed. This system was designed as a response to a previous research called “Neurophysiology of Emotions and Intimate Partner Violence (IPV) against Women”. The designed system uses 8 EEG channels to analyze alpha and theta brain-waves from 4 areas of the brain: prefrontal, frontal, temporal and central. By recording a 30 seconds baseline, the system is capable to detect changes in the EEG signal that can be used for the interaction. For feedback, three musical features are modified depending on the EEG analysis: tempo, loudness and loudness of the voice of the singer. For testing the system, two protocols were designed, these protocols focused on memory stimulation and stress regulation. They were designed specifically for each one of the three types of feedback. These protocols were applied on two women (43 and 52 years old), both had been part of the previous project. Results are promising, showing changes in the EEG signals of the participants when comparing the first session and the last one. Changes in performance of some specific tasks in the protocols, show an adequate usability of the system. Further studies will be carry on in order to evaluate long-term effects of the system with more activities.

Reverberant shear wave elastography (R-SWE) is a novel technique that takes advantage of reflected waves, applying multiple vibration sources near the region of interest, in order to create a diffuse field. Different studies show encouraging results assuming a full development of the desired field. However, the presence of the optimal diffuse field it is not guaranteed. Moreover, previous studies have performed numerical simulations and phantom experiments, considering the reverberant field generation through the summation of plane shear waves propagating with random amplitude, direction and phase. In contrast, these assumption is not accurate, since physical experiments could be performed using normal load sources with a reduce contact surface, propagating spherical shear waves. This work is based on a simulation study with four different scenarios with aleatory mini-surface sources considering their position, degree of similarity with theoretical models and shear wave speed (SWS) estimation. The results suggest that similar SWS maps can be reconstructed independently of the range of sources used. Furthermore, there is a significant bias (around 25%) for SWS estimation obtained by curve fitting estimation since this technique is optimized for plane waves.

Human pose detection is defined as the process of locating the joints of a person or a crowd given an image or video. Currently, pose detection is widely used for the evaluation of athletes, workers, and the monitoring of patients in clinical settings. However, human pose estimation and fall detection are not easy tasks as it requires experts to manually assess the person’s position by using specialized equipment such as e-health devices (watches, bands, handles), markers and high-cost cameras to monitor a limited scenario. The main goal of this article is to implement a marker-less low-cost computer vision system to get the automatic estimation of poses and falls detection recorded on video by calculating the person’s joint angle with a high level of adaptability to any space. This proposed model is the first step in the construction of a system that allows monitoring and generating alerts to prevent falls at home and clinical settings.