This PDF file contains the front matter associated with SPIE Proceedings Volume 11330, including the Title Page, Copyright information, Table of Contents, Author and Conference Committee lists.

Purpose: Evaluate the feasibility of using a Nakagami model to create an accurate parametric image from ultrasound imaging data for the differentiation of homogenous and heterogeneous texture phantoms. Analysis was done on the raw data i.e., radiofrequency (RF) data collected before any post processing that can affect the images. Materials and methods: The Nakagami parametric image was constructed on demodulated RF data with the sliding window technique to create a map of local parameters. The Nakagami parameter (m) for the entire image was found by averaging all values. By design, when m is greater than 1, the distribution is post-Rayleigh. When m is equal to 1, the distribution is Rayleigh. To test the technique, two agar phantoms were constructed, using varying amounts of flour as the scatterer. The higher amount of flour scatterer was meant to mimic heterogeneous texture and the lesser amount meant to mimic homogeneous texture. Scans were done on each phantom and analyzed for differences in the Nakagami parameter. Results: Phantom 1 displayed a post-Rayleigh distribution (m = 36.1±7.0), while phantom 2 did so, to a lesser extent (m = 1.64±0.12). As the distribution transitions from Rayleigh to post Rayleigh, the scatterers in the sample go from being periodically located/randomly distributed to large numbers of randomly distributed scatterers. Conclusion: Our study suggests that Nakagami parametric based metrics may be used to increase robustness of texture analysis, considering the analysis is done on the raw data before any post processing that can affect the images is introduced.

Pneumonia is an infection of the lungs caused by virus, bacteria or fungi. It affects mainly children under five and can be life-threatening. Diagnosis of pneumonia is usually performed using imaging techniques such as chest radiography, ultrasound, and CT. Several studies have shown that ultrasound is an effective, safe and cost-efficient technique for pneumonia detection. However, due to the low signal-to-noise ratio of the images, this technique is highly dependent on the experience of the practitioner. This paper proposes an approach for pneumonia detection from image texture features. We used empirical mode decomposition for feature extraction, principal component analysis for dimensionality reduction and supervised learning methods for classification. Results show that features of the first mode present large differences between healthy and pneumonia patients according to the Cohen’s d index. Pneumonia detection was possible with a rotation forest model with a mean accuracy of 83.33%.

The growing need to perform surgical procedures, monitoring, and intervention of greater precision have led to the development of multimodal medical imaging systems. Multimodal images are a strategy to overcome the limitations of medical imaging technologies by combining the strengths of individual modalities or technologies. In this work, we propose a low-cost multimodal system that combines 3D freehand ultrasound with fringe projection profilometry to obtain information from the external and the internal structure of an object of interest. Both modalities are referred to a single coordinate system defined in the calibration to avoid post-processing and registration of the acquired images. The freehand ultrasound calibration results are similar to those previously reported in the literature using more expensive infrared tracking systems. The calibration reproducibility at the center point of the ultrasound image was 0.6202 mm for 8 independent calibrations. We tested our system on a breast phantom with tumors. Encouraging results show the potential of the system for applications in intraoperative settings.

Speckle noise filtering has been investigated since at least fifty years, this multiplicative and granular interference may be found in any image, i.e, Synthetic Aperture Radar (SAR), optical coherence tomography, and of course, medical ultrasound imaging. Speckle noise is produced by structural characteristics of materials, in case of the ultrasound imaging case, by small structural irregularities. This work proposes a novel speckle noise filtering strategy using a bank of morphological multi-scale filters that captures anisotropic information and additionally preserves cardiac structures. This method is compared against commonly used filters, namely: Anisotropic Diffusion Filter (ADMSS), Non-Local Means Filter (NLMF) and Detail Preserving Anisotropic Diffusion Filter (DPAD).

The interpretation of tissue images is the first step performed by the pathologist before render an accurate diagnosis. Digital Pathology is challenging to implement in developing countries since, due to the acquisition process, these images usually become low illuminated and contrasted, which significantly affects their analysis. Transformation of these low-quality images would be needed to improve the pathologist diagnosis process. This article aims to show the results obtained from a test carried out by three pathologists on a series of classic color manipulation algorithms that improve the information (i.e., nuclear shape, stroma, and gland formation) contained in prostate images. For this study, images from prostatic tissue supplied by the Instituto Caldense de Patologia (ICP) were used. Seven biopsy samples were used, each of them captured with 10x, 20x, and 40x magnifications, forming a total of 50 images. Then, a subjective quality test was performed by three pathologists. This test consisted of grading a series of 42 classical image processing algorithms on a scale of 0 to 4. The value 0 represents the worst rating (the new image did not provide any information), and 4 represents the best rating (the original image highlights exciting diagnosis features and offers essential information). 52.27% of the transformations were classified as useful (scores between 3-4) by the pathologists (23 of 44 in total). The color transformations succeeded in highlighting the different structures such as the cytoplasm, stroma, and nucleus, improving the perception of contours and shapes, which will help to render an accurate diagnosis.

Histopathological evaluation plays a crucial role in the process of understanding lung cancer biology. Such evaluation consists in analyzing patterns related with tissue structure and cell morphology to identify the presence of cancer and the associated subtype. This investigation presents a multi-level texture approach to differentiate the two main lung cancer subtypes, adenocarcinoma (ADC) and squamous cell carcinoma (SCC), by estimating global spatial patterns in terms of cell density. Such patterns correspond to texture features computed from cell density distribution in a co-occurrence frame. Results using the proposed approach achieved an accuracy of 0.72 and F-score of 0.72.

Age-related macular degeneration (AMD) is the leading cause of irreversible blindness in industrialized countries. It is estimated that AMD affects at least 1 in 10 Hispanics. Previous reports have shown that AMD has multiple risk factors. Recently, we demonstrated that some genetic variants in the SGCD gene are involved in AMD developments, especially in early-stage (geographic atrophy, GA). Therefore, to evaluate the relationship between SGCD's absence and the loss of photoreceptors in GA, we worked with a genetically modified mouse model, SGCD deficient (Sgcd^−/−) and a control mouse C57BL/6J (Sgcd^+/+). First, we obtained hematoxylin and eosin (H&E) retina staining microscopic images. Then, we coarsely selected the outer and inner nuclear retinal layer (ONL and INL respectively) and finally, we applied an automatic nuclei segmentation to calculate the nuclear density in each region. Our results showed that Sgcd absence does not result in photoreceptor loss, on the contrary, it promotes an increment in nuclear density by 8.7% in ONL and 20.1% in INL compared with control eyes (p = 0.0033 and p < 0.0001 respectively). This could be explained by the fact that SGCD codifies the delta-sarcoglycan protein and there is evidence that showed a relationship between the absence of this protein with the activation of a cell proliferation signaling pathway. Finally, our results show that the delta-sarcoglycan protein could play an important role in the pathogenesis of the geographic atrophy. Moreover, there are promising perspectives for the systematic approach applied for cell image analysis, as an important tool to determine the nuclear density for assessing the progression of AMD.

Clinical imaging techniques have low accuracy in differentiating malignant tumors such as clear cell Renal Cell Carcinoma (ccRCC) and benign tumors such as oncocytoma. Texture metrics i.e., metrics assessing the variations in grey-levels of intensity making up a region of interest extracted from routine clinical images have shown promising results in achieving this objective. To explore the relationship between tumor behavior and texture metrics from images, we test the effectiveness of 2D Curvelet Transform-based texture analysis in differentiating between ccRCC and Oncocytoma using contrast-enhanced computed tomography (CECT) images. Whole lesions were manually segmented on the nephrographic phase using Synapse 3D (Fujifilm, CT) and co-registered to other phases of multiphase CT acquisitions for each tumor. A first-generation curvelet transform code was used to apply forward, inverse transform to segmented images, and texture metrics were extracted from each CT phase. Histopathological diagnosis was obtained following surgical resection. A Wilcoxon rank-sum test showed that curvelet-based metric: energy on corticomedullary phase was significantly (p <0.005) higher in oncocytoma (0.06±0.04) than ccRCC (0.04±0.05). Higher values of energy are associated with homogenous textures. A supportive receiver operator characteristics analysis based on energy metric revealed reasonable discrimination (AUC>0.7, p <0.05) between ccRCC and oncocytoma. We conclude based on these preliminary results that curvelet- based energy metric can differentiate between ccRCC and oncocytoma based on their CECT data. In combination with other metrics, curvelet metrics may advance radiomic analysis in evaluating clinical imaging data.

Convolutional Neural Networks have been successfully used in several tasks in the last decade, but this kind of supervised method requires a large number of annotated examples to obtain good results. However, in biomedical domains, such as digital pathology, the amount of annotated samples is very reduced and imbalanced. On the other hand, models trained using supervised methods need to be retrained when a new class (with its samples) is introduced. This process is computationally expensive and can be difficult when you don’t have a lot of examples for the same class. In order to address these issues, there are novel approaches that use a few number of annotated examples to achieve an acceptable or good generalization capability of the model. One of these methods is One-shot Learning, which is intended to classify a set of data from different classes from one or a few number of annotated examples, and which allows to incorporate new data from new classes without re-training the model. This work explores a method to classify samples of tissues in breast cancer histopathology images by means of similarity between pairs of image samples using a Siamese Convolutional Neural Networks, which achieved a 90.83% of accuracy test.

This article introduces a novel adaptive frequency saliency model (AFSM) that selects relevant information by filtering an image with a set of band pass filters optimally placed in the frequency space using an auto- encoder CNN. The obtained images show a higher signal-to-noise ratio and therefore they improve a classifier performance. The proposed method is challenged by a classification task: prostate magnetic resonance imaging (MRI) to be labeled as cancerous or non-cancerous tissue. Evaluation in this case was carried out by training a convolutional neural network (CNN) with a prostate dataset but at the testing phase, the trained model is assessed with non-filtered and filtered images. The classifier tried with filtered images outperformed the results obtained with the non filtered ones (classification accuracy scores of 0.792± 0.016 and 0.776± 0.036 respectively), demonstrating better overall performance and the importance of using filtering processes.

Magnetic resonance imaging (MRI) plays a valuable role in many task related with characterization of prostate cancer lesions. Recently, the DCE-MRI (Dynamic contrast Enhanced) has allowed to visualize and localize potential tumor regions. Specifically, K^trans, from DCR-MRI, has shown to be a powerful pharmacokinetic parameter that allows to characterize tumor biology and to detect treatment responses from reconstructed coefficient maps of capillary permeability. Nevertheless, even expert-based analysis of K^trans sequences are subject to a large false positive findings (FPF). In much of such cases, the prostate angiogenesis, or benign prostatic hyperplasia (BPH) regions are misclassified as cancer findings. This work introduces a robust deep convolutional strategy that characterizes K^trans regions and allows an automatic prediction of cancer findings. The proposed strategy was validated over the SPIE-AAPM-NCI PROSTATEx public dataset with 320 multimodal images on peripheral, transitional and anterior fibromuscular stroma regions. The best configuration of proposal strategy achieved an area under the ROC curve (AUC) of 0.74. Additionally, the proposed strategy achieved a proper characterization by using mainly K^trans information that together with T2-MRI-transaxial overcome baseline strategies that use additional modalities of MRI.

Histopathological tissue analysis is the most effective and definitive method to prognosis cancer and stratify the aggressiveness of the disease. The Gleason Score (GS) is the most powerful grading system based on architectural tumor pattern quantification. This score characterizes cancer tumor tissue, such as the level of cell differentiation on histopathological images. The reported GS is described as the sum of two principal grades present in a particular image, and ranged from 6 (cancer grow slowly) to 10 (cancer cells spread more rapidly). A main drawback of GS is the pathological dependency on histopathological region stratification, which strongly impacts the clinical procedure to treat the disease. The agreement among experts has been quantified with a kappa index of: ~0.71. Even worse, a higher uncertainty is reported for intermediate grade stratification. This work presents a like-inception deep architecture that is able to differentiate between intermediate and close GS grades. Each image herein evaluated was split-up into regional patches that correspond to a single GS grade. A set of training patches were augmented according to appearance image variations of each grade. Then, a transfer learning scheme was implemented to adapt a bi-Gleason tumor patterns prediction among close levels. The proposed approach was evaluated on public set of 886 tissue H&E stained images with different GS grades, achieving an average accuracy of 0.73% between grades three and four.

Bone Age Assessment (BAA) is a task performed by physicians to estimate the skeletal development of a pediatric patient. Tipically physicians perform this exam by doing a manual analysis of the X-ray image of the non-dominant hand of a child, either by taking the image as a whole or paying attention to certain anatomical Regions Of Interest (ROIs). Over the years, several datasets have been proposed in order to generate automated methods to perform this task. Most notably, in 2017 the Radiological Society of North America (RSNA)1 created the Pediatric Bone Age Challenge, which encouraged the development of machine learning approaches for this task. In this paper, we present GPNet a convolutional neural network capable of performing BAA precisely and effectively by analyzing the whole hand in a single forward pass. We train GPNet using the training data available from the dataset created in the RSNA challenge and evaluate our method using the validation set. We use the testing set to compare our performance with the current state-of-the-art and find that GPNet significantly outperforms previous methods. During our architecture search we perform several experiments to demonstrate the effect of different layers, proving that some blocks do not contribute to the performance of the network, but instead they affect it. As a result, we are able to develop a method that reduces the number of trainable parameters by nearly 82.15 M in comparison to the state-of-the-art, while improving the performance.

Human Pose Estimation (HPE) is a Computer Vision problem that has become increasingly popular over the last few years, with multiple applications in the medical field such as therapy using virtual and augmented reality, robot caregivers, virtual physical therapy and kinematic analysis. Nevertheless, all the machine learning algorithms developed for these applications are trained in small datasets with images captured in constrained scenarios and with information given by sensors, bounding the applicability of these methods. We developed a simple yet useful deep learning algorithm for Human Pose Estimation that uses as input only an image of a scene with people. The estimated position of the joints and body parts can be used to retrieve basic kinematic information from the people on the image that can be applied to the aforementioned medical applications. We focus on overcoming the limit of Human Pose Estimation algorithms due to jittering, aiming to preserve more precise pixel location. Thus, we explore different novel approaches to improve the precision of the existing state- of-the-art algorithms in keypoint estimation and evaluate them on COCO keypoint dataset, outperforming the current top methods. We hope our algorithm encourages the academic community to develop simpler but precise HPE algorithms for medical applications based on RGB images.

Automatic segmentation of brain tumors is a challenging problem with many inherent difficulties, such as restricted training data, great intra-class variance, and volumetric images with large computational requirements in terms of processing. To overcome these difficulties, we propose Brain Tumor Parser (BTP), a novel convolutional neural network that takes advantage of a refinement module and global 3D information to perform semantic segmentation of brain structures in volumetric images with multiple modalities. We draw inspiration from recent breakthroughs in edge detection and semantic segmentation in natural images, and we build an accurate and effective three-dimensional network that segments small structures while refining large instances in multi-modal Magnetic Resonance Imaging (MRI). We evaluate our approach on the data from the Brain Tumor segmentation (BraTS) 2017 challenge, obtaining comparable results with the best performing algorithms, while using a single yet efficient architecture.

High-dimensional manifold modeling increases the precision and performance of cortical morphometry analysis by densely sampling on the grey matters. But this also brings redundant information and increased computational burden. Gaussian process regression has been used to tackle this problem by learning a mapping to a low-dimensional subspace. However, current methods may not take relevant morphometric properties, usually measured by geometric features, into account, and as a result, may generate morphometrically insignificant selections. In this paper, we propose a morphometric Gaussian process (M-GP) as a novel Bayesian model on the gray matter tetrahedral meshes. We also implement an M-GP regression landmarking algorithm as a manifold learning method for non-linear dimensionality reduction. The definition of M-GP involves a scale-invariant wave kernel signature distance map measuring the local similarities of geometric features, and a heat flow entropy which implicitly embeds the global curvature flow. With such a design, the prior knowledge fully encodes the geometric information so that a posterior predictive inference is morphometrically significant. In experiments, we use 518 grey matter tetrahedral meshes generated from structural magnetic resonance images of a publicly available Alzheimer’s disease imaging cohort to empirically and numerically evaluate our method. The results verify that our method is theoretically and experimentally valid in selecting a representative subset from the original massive data. Our work may benefit any studies involving large-scale or iterative computations on extensive manifold-valued data, including morphometry analyses and general medical data processing.

Diffusion-weighted imaging (DWI) is a magnetic resonance imaging technique commonly used to infer tissue microstructure, however, acquisition time requirements affect the effective spatial resolution for DWI high quality. This paper presents a novel super-resolution strategy to reconstruct high-resolution DW images by linearly combining information from different gradient acquisitions. The strategy comprises two main stages, a representation learning and a high-resolution mapping. In the former stage, information from different gradients is grouped by patch-wise statistical similarities. Representative coefficients are then estimated to represent each group. In the latter stage, adapted patch coefficients predict the high-resolution image while a regularization method eliminates possible reconstruction overlapping effects. Several tests evaluate the method ability to pre- dict high resolution information, PSNR and SSIM metrics were applied to quantitatively measure the quality improvement. Results demonstrate that quality reconstruction outperforms state of art methods in about 0.3 dB for PSNR and 1 % for SSIM.

Hypothalamus is a small structure of the brain with an important role in sleep, appetite, body temperature regulation and emotion. Some neurological diseases, such as Schizophrenia, Alzheimer and Amyotrophic Lateral Sclerosis (ALS) may be related to hypothalamic volume variation. However, hypothalamic morphological landmarks are not always clear on magnetic resonance (MR) images and manual segmentation can become variable, leading to inconsistent findings in the literature. In this work, we propose a fully automatic segmentation method, with no human interaction, to segment hypothalamus in MR images using convolutional neural networks (CNNs). The best performance was obtained by a consensus model using the majority voting from three 2D-CNNs trained on axial, coronal and sagittal MRI slices, achieving a DICE coefficient of 0.77. To the best of our knowledge, this is the first work to fully automatically segment the hypothalamus.

Image acquisition and automatic quality analysis are fundamental stages and tasks to support an accurate ocular diagnosis. In particular, when eye fundus image quality is not appropriate, it can hinder the diagnosis task performed by experts. Portable, smart-phone-based eye fundus image acquisition devices have the advantage of their low cost and easy deployment, however, their main disadvantage is the sacrifice of image quality. This paper presents a deep-learning-based model to assess the eye fundus image quality which is small enough to be deployed in a smart phone. The model was evaluated in a public eye fundus dataset with two sets of annotations. The proposed method obtained an accuracy of 0.911 and 0.856, in the binary classification task and the three-classes classification task respectively. Besides, the presented method has a small number of parameters compared to other state-of-the-art models, being an alternative for a mobile-based eye fundus quality classification system.

Pleomorphic adenomas (PA) and Warthin tumors (WT) are the most common salivary gland tumors. It exhibits very similar characteristics, which make very difficult to distinguish them. To differentiate these it is necessary due that it strongly influences the surgical procedure and treatment. In this paper, we propose the use of radiomic analysis for discriminating PA and WT based on T1 and T2 Magnetic Resonance Images. The proposed approach starts with the extraction of a large number of quantitative features of delineated tumors from both T1 and T2 sequences, then a feature selector is used for selecting the most discriminative, which are used for training an SVM classification model, whose parameters are optimally selected using the Particle Swarm Optimization algorithm. The proposed approach is able to discriminate between PA and WT tumors with a sensitivity of 91.57± 11.07 and specificity of 89.50 ±11.07, making this a promising method to characterizing and classifying parotid gland tumors.

Patients who develop Alzheimer’s disease (AD) have passed through a prodromal stage called mild cognitive impairment (MCI) but, not every MCI patient will develop AD. Structural magnetic resonance imaging (MRI) has the potential to capture the differences between the groups of MCI: patients who converted to dementia due to AD (MCIc) and those who remained stable (MCInc). Sulci and folding structures are affected due to the neurodegenerarive nature of AD. A characterization method for these structures is proposed by performing a multi-spherical mapping and a curvelet decomposition at different levels of the brain. 829 MCI subjects from the ADNI initiative were studied at 24, 36 and 60 months of conversion. A random forest classification model was used to validate the methodology achieving an AUC = 0.77 with high-dimensional feature vectors and AUC = 0.70 when 5% of the features are selected. This suggests discriminating differences in sulci and gyri depth between non-converters and converters to AD.

As improved treatments have increased survival rates for pediatric brain tumor patients, there is an urgent need to study the long-term sequelae of treatment. Central nervous system (CNS) germ cell tumors constitute a form of neoplasm that is normally treated with chemotherapy and/or cranial radiation therapy (CRT). Starting from an MRI dataset of patients with such tumors at Children's Hospital Los Angeles, we aim to understand the morphological effects of CRT on brain structures. The current study examines the long-term effects of radiation dosage on the brainstem, corpus callosum, and ventricles post- radiation therapy. The results of the analysis showed a significant association among brain structure areas and radiation dosage received. These preliminary findings point to deleterious effects of radiation treatment on brain anatomy.

The purpose of this work is to characterize the process of natural neuronal death associated with brain aging in healthy subjects by characterizing local relationships in nuclear magnetic resonances. The whole method focuses on describing local neighborhoods with a texture descriptor, specifically the Linear Binary Pattern invariant to rotation, aiming to capture local tissue relationships by clustering this texture descriptor in brain regions previously parcelled. This approach has been tested in a group of control subjects extracted from the ADNI database project, dividing the experimental group after subjects’ chronological age. Results show differences concentrated in certain regions, i.e., the caudate, the hippocampus, and the thalamus, all of them usually reported as deteriorating with ageing process.

In this paper, we present a new non-contact strategy to estimate the breathing rate based on the Eulerian motion video magnification technique and a system based on a Convolutional Neural Network (CNN). After the magnification procedure, a CNN 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. Two strategies are used as input to the CNN. A CNN-ROI proposal where a region of interest is selected manually on the image frame and in the second case, a CNN-Whole-Image proposal where the entire image frame is selected. Finally, the RR is estimated from the classified frames. The CNN-ROI proposal is tested on five subjects in lying face down position and it is compared to a procedure using different image processing steps to tag the frames as inhalation or exhalation. The mean average error in percentage obtained for this proposal is 2.326±1.144%. The CNN-whole-image proposal is tested on eight subjects in lying face down position. The mean average error in percentage obtained for this proposal is 2.115 ± 1.135%.

In the up-sampling process may occur effects like aliasing, blurring or noise addition which mainly affect the edges of the images. For those reasons is necessary to choose a method that preserves images quality so that these problems are minimized. In this paper, we present an alternative method to restore blurred images using linear programming to solve a minimization problem stated in the L₁ norm. The model requires the blurred image and some prior knowledge about the blurring function type (Point spread function). In the proposed method we obtain a PSNR of 30 dB overcoming a classic bi-linear method by 4 dB in a set of thirty images from a cardiac MRI data set.

Attention-deficit/hyperactivity disorder (ADHD) is a neurodevelopmental condition whose mechanism is not completely understood. However, it has been consistently found that regions within the Default Mode Network have an important role in the dysfunctional brain activity. The objective of this study was to use a dynamic brain approach for the analysis of resting-state functional magnetic resonance imaging data from subjects with typical development and subjects with ADHD, and to compare it with the static approach. In the current study we computed the amplitude of low frequency fluctuations in the 0.01 to 0.08 Hz frequency band within the Default Mode Network regions. Statistically significant differences between typical development and ADHD subjects were found, where more regions revealed a differential behavior between groups when performing the dynamic approach. Results indicated that the dynamic approach provides complementary explanation of brain function, compared to the static analysis, and allows to access information concerning temporal behavior by looking at each window.

The detection of the heartbeat from electrocardiographic (ECG) and arterial blood pressure (ABP) signals, either exploited individually or jointly, has been carried out successfully using different approaches that range from the use of simple digital signal processing techniques until the use of more advanced techniques based on machine learning. In this paper, we employed a heartbeat detector that uses two hidden Markov models (HMM) to characterize the dynamics of the presence and the absence of heartbeats in ECG and ABP signals. The HMM-based detector can exploit univariate observations (ECG or ABP signals) or bivariate observations (ECG an ABP signals jointly, in a centralized manner). Two transformations of the signals were applied as a preprocessing step: absolute value and squared functions. In this sense, six detectors based on univariate observations and nine detectors based on bivariate observations were conceived and validated in ten records of the MGH/MF Waveform Database. The detection performance when the absolute value of ECG and the absolute value of ABP are jointly exploited by the HMM produced T P = 58736, FN = 631, FP = 788, sensitivity = 98.73%, positive predictivity = 98.22%).

Quantitative ultrasound has shown potential of improving medical diagnosis. In this work, a regularized power law (RPL) for the joint estimation of backscatter coefficient (BSC) and attenuation coefficient (AC) parameters was derived and tested with simulated phantoms and in vivo. For the RPL method, a total variation regularization term regarding the BSC and AC parameters were used. The results were compared with ground truth simulated values. An improvement of precision can be appreciated without compromising the accuracy by factor of 80% and 54% according to bias and coefficient of variation, respectively. The in vivo experiments showed comparable results of the algorithm with the literature (i.e. fibroadenoma:1.90 ± 0.31dB.cm⁻¹ .MHz⁻¹ , normal tissue: 0.62 ± 0.20dB.cm⁻¹ .MHz⁻¹ ). The results suggest the RPL method has the potential to accurately and precisely estimating BSCs and ACs.

Crawling Waves Sonoelastography (CWS) is an elastography technique based on an interference pattern produced by the application of two external vibration sources. In this study, a 3D printed holder was designed to overcome the compatibility limitations of normal excitation modules for different ultrasound systems, previously developed in the literature. This holder was assembled with a transducer and two voice coil actuators, and its performance was tested in homogeneous and inclusion tissue-mimicking phantom. A phase derivative estimator was used to reconstruct the shear wave speed (SWS) map. The coefficient of variation (CV), bias and contrast- to-noise ratio (CNR) were used as quantitative metrics for comparison. These metrics have shown a comparable SWS estimation in the background (SWS_b) of the homogeneous phantom with previous studies (e.g. SWS_b = 3.58± 0.42 m/s). In the same way, the SWS of the inclusion (SWS_i) was distinguished from the background with an accurate SWS value (e.g. SWS_i = 5.64 ±0.42 m/s and SWS_b = 3.70 ±0.16 m/s). The results suggest that the portable device is capable of achieving a good performance according to their values of bias, CV and CNR. The extension of this work is oriented to ex vivo and in vivo experiments.

Skin diseases cause changes in the elasticity of the skin. Recently, a new ultrasound method which combines crawling waves elastography with a high-frequency ultrasound (HF-US) system was reported for the estimation of skin elasticity. This study applies crawling waves method with an HF-US system in homogeneous and heterogeneous phantoms with similar shear wave speeds found in the literature to validate this technique. In addition, estimation performance was compared between the Phase Derivative (PD) estimator and the Regularized Wave- length Average Velocity Estimator (R-WAVE). Results corroborate the feasibility of this method. Additionally, it suggests that R-WAVE estimator can provide better measurements than PD with a lower coefficient of variation and fewer artifacts.

Latin America and other places in the world have tropical regions, from which multiple diseases emerge. Cutaneous Leishmaniasis (CL) is one of the most common pathologies that attack the populations on these areas, generating an ulcer and leaving marks for life. The difficult access to appropriate medical attention of affected populations raises the need for new instruments and techniques that facilitate the diagnosis and monitoring of treatments for this disease. This article proposes a methodology of analysis of golden hamster skin with cutaneous ulcer caused by leishmaniasis through a three-layer inverse reflectance model, genetic algorithms and the Hyperspectral image acquisition system ASCLEPIOS. As a result, the concentration maps of: thickness of the epidermis (epi), epidermis plus thickness of the dermis (epiPder ), volumetric fraction of blood (Fblood), diameter of the Collagen particles (Coll), volumetric fraction of collagen (Fcoll), diameter of the Fibroblast particles (Fibro), volumetric fraction of melanin (Fmel), diameter of Keratynocytes particles (Kera), diameter of Macrophages particles (Macro), and oxygen saturation (OS) were obtained.

Chagas disease (American trypanosomiasis) is an endemic parasitic disease in some areas of Latin America, about 16-18 million people are infected with the etiology agent of Chagas disease , Trypanosoma cruzi, and is transmitted to humans through triatomine insects commonly known as kissing bugs. One of the standard laboratory diagnosis during acute phase of the disease is by direct visualization of the parasite, the most common methods is the visualization in blood smear stained with some colorant. Trypanosoma cruzi uses several strategies to survive in different hosts which involves various morphological, biochemical, and genetic changes. Trypanosoma cruzi displays distinct morphology changes, which have not been fully characterized. The objective of this work is the morphological characterization of shape structures on blood smears. We proposed a high resolution chain code algorithm in bi-dimensional curves, which allows to discretize the contour with greater approximation to its real shape, and consequently obtain features in a objective way.

Image-based diagnosis is able to spot several diseases and clinical conditions faster and more accurately than traditional manual ones, becoming also an alternative in monitoring and predicting patients responses to specific health treatments. In this work, we present a supervised learning approach to segment pixel-wise parts of spermatozoa using a random forest (RF) classifier. The framework created a multi-channel image combining intensity RGB bands with three neighborhood based bands. The last neighborhood based bands were Sobel’s magnitude and orientation and Shannon’s entropy. A RF was trained using labeled pixels provided by expert andrologists, biochemists and specialists in reproductive health. We compared results with a simple model on the RGB only. The whole automatic process (segmentation and classification) achieved an average precision of 98%, recall of 98% and F-Score of 98%. Highest improvement in comparison to the RGB model was shown on the segmentation of the tail. We provided a fully automatic spermatozoa semantic segmentation based on local and non-local information. The results are aimed to develop a CASA (Computer Assisted Sperm Analysis) system that can provide results over the Internet. The experiment was conducted on normalized images of a specific microscope. We are planning to extend the experiment in future work to more realistic conditions including different stainings, microscopes and resolutions.

Lower limb prosthesis has the purpose of recovering mobility in amputees, giving autonomy to patients to do several activities. Mobility degree quantification and correct use of the prosthesis is necessary to reduce the risk of desertion. An adequate measurement of movements when patients are walking can help the physiotherapists evaluate the performance. For that reason, this work presents a new tracking method based on the extraction of texture and shape features that feed the retraining Random Forest classifier. The aim is to use a depth camera to track people with lower limb prosthesis when walking between parallel bars. Two experiments were performed with the proposed system: the first one under three patients with lower limb prostheses in order to apply the tracking algorithm. The second was carried out in three healthy control subjects with the purpose of validating the proposed algorithm and comparing the results with a motion capture system (MoCap). In this test the participants carried out two different activities; the results present errors from 3.3 to 4.9 mm according to the root mean square error. This suggests that the system can be used to track human joints under different conditions; however, it is necessary to solve the problem of occlusion artifacts by using human body models or by employing several depth cameras.

Observation of Gait patterns is the available evaluation in clinical routine of the motor manifestations in Parkinson’s Disease (PD). Lately, different investigations have attempted to quantitatively analyze gait patterns by linear methods facing several limitations since the non-stationary nature of the gait patterns. This study presents a non-linear characterization of the Parkinson's disease gait by a deterministic chaotic analysis which represents the temporal gait dynamics with a minimum set of parameters. Specifically, delay and embedding dimension parameters were obtained for reconstructing the phase space and its characteristic coefficients, namely Lyapunov, correlation dimension, and approximate entropy. Statistical differences (p < 0.05, Mann-Whitney test) were found for the Lyapunov exponent and the approximated entropy when describing two gait patterns, i.e., control and PD groups.

Foot drop, or drop foot, is a difficulty in lifting the front part of the foot, which can make walking challenging. This study analyzes the gait of a 52-year-old patient diagnosed with drop foot caused by a surgery at the lumbar level. Baropodometric tests were conducted with and without the cane as assistive device to identify significant changes in load distribution and stabilometry. An inertial sensor was also used to get information about gait quality, symmetry and propulsion index, and pelvic angles when said aid was implemented in order to confirm its influence on the patient’s gait. In obtained results with use of cane, there is found an improvement of 34.4% in lateral distribution of loads and 4.7% in quality index of left foot gait, this confirms positive influence that has this device to rehabilitation process of the patient.

Linear morphea is a disease that affects the subcutaneous tissue and the skin generating its hardening due to the excess of collagen production, causing joint contracture that can involve muscle and bone preventing the growth of the affected limb. This study analyzes the gait of a 13-year-old patient who was diagnosed with this pathology, a biomechanical analysis was performed using baropodometric tests and gait analysis. For this, conventional tests were carried out using specialized sensors, how the test results were designed and the development of an orthotic support device, as an actuator to improve the patient's condition. Subsequently, the use of the device in the patient was implemented to evaluate its effectiveness by comparing the tests performed before and after use. It was identified that improves the quality of the patient gait and in the parameters such as the distribution of body weight and its stability.

Visual disorders are one of the common side problems in Cerebral Palsy (CP) being reported with an incidence between 50% to 90% of the cases. These visual disorders may interfere with the developmental process and motor learning of these children. This work presents a robust method to spatio-temporally characterize the ocular motion in CP. Smooth pursuit and saccadic eye tasks were assessed by using simple visual stimuli and recording the eye trajectories during the task. A dense optical flow estimation of the ocular movement is modulated by a visual attention model which extracts relevant eye motion information. Analyses in time and frequency domain were performed suggesting statistical differences (p-value < 0.01) in Signal to Noise Ratio (SNR) and Wavelet coefficients for both saccadic and smooth pursuit tasks.

Photoplethysmography (PPG) monitoring with wrist-wearable devices appears as a non-invasive method, enabling longterm monitoring of the heart rate and thus, improving arrhythmia detection. The motion artifacts have been shown as a harsh, but an avoidable drawback for heart disease diagnosis, since they distort the signal and lead to a misinterpretation of the results, then missing or overestimating arrhythmia events. Noise extraction without removing essential data from the signal is challenging, mainly because of the overlapping between the spectra of both signal and noise. Also, the nonexistence of public and available PPG signals datasets with motion artifacts and arrhythmias included at once, do not allow other studies to prove and contribute with the dealing of both these problems. Therefore, five approaches to recreate noise due to motion artifacts were proposed, characterizing six activities with different intensity levels and movements, using information from real patients. To evaluate the performance of each noise model, these were used to subtract noise from the same dataset of PPG signals with six physical activities, hoping they can resemble the behavior of the movement artifacts presented. Subsequently, a peak detector was used to perform classification tests using ECG signal as the gold standard. This test showed a better performance of the Dynamic Variance Moving Average method, increasing the sensitivity and specificity by 2%. As a result, a model for noise components of motion artifacts using an open database of PPG corrupted data was created, looking forward to use this as a contribution for future works on noise removal algorithms validation.

Organizational behavior and organizational cognitive neuroscience are newly established interdisciplinary fields that use scientific techniques to answer questions about behaviors within organizations. Clarifying the more precise role of emotions and their regulation in forming a judgment in different managerial decision-making contexts, exploring how team members function synchronously, and the links between psychophysiological traits and how they relate to leadership, have been recurring subjects of interest. In this document, we present the psychophysiological response of eight senior project managers and eight junior project managers as they are faced with a challenging situation designed to assess some of the most relevant personal skills in project management: teamwork, management, problem, and conflicts resolution. Noninvasive electroencephalography (EEG), voice recording (VoR) and video recording (ViR) were performed during the application of an Assessment Center evaluation in addition to before and after heart rate (HR), blood pressure (BP) and pulse oximetry (SpO₂) measurements. The results suggest that the voice frequency is related to the task that a subject performs. In this case, an activity that requires active discussion causes a person to talk at a higher pitch, which is the mean value of the fundamental frequency of the voice. In relation to the analysis of the voice depending on the project management expertise, no significant differences were found for calculated voice features. These results suggest that the voice features are not related with the project management expertise. On the other hand, the coefficient of variation of facial redness for junior project managers groups was higher than those of the senior manager groups. This indicates that even though facial readiness was not universally higher for the junior project manager, it is much more prevalent in those with the shorter project management expertise. With respect to the estimation of subject’s task engagement, measured through head movement cue analysis, it can be said that this method of engagement estimation proved to be in accordance with the expert psychology’s opinion. In concern with EEG, no statistically significant differences were found in frontal asymmetry value between the discussion stage and the presentation and ideation stage and between Senior and Junior managers in the discussion stage. Regarding the before and after heart rate (HR), blood pressure (BP) and pulse oximetry (SpO₂) measurements, even though, in general all three variables had a discernible behavior (HR and BP increased while SpO₂ decreased), no statistical relevant differences were found in the pre and pose measurements. However, it can be said that the Assessment Center evaluation induced a physiological response on the subjects since, at least one of three variables varied for each subject.

The geographical conditions of Colombia favor the re-emergence and propagation of infectious tropical diseases. Among them, the Dengue virus is highly endemic throughout the country, thereby locating this arbovirus as one of the major pathologies of the public health system. Therefore, there is a global challenge to generate novel strategies to predict and control dengue virus transmission. In particular, during the Colombian 2016 Dengue outbreak, more than 17 thousand Dengue cases were reported by the health authorities of Medellin, Antioquia. In this paper, we present a machine learning approach for the early detection of dengue outbreaks in the city of Medellin. We use an artificial neural network as the core of the machine learning algorithm, with environmental, meteorological and epidemiological data from the National Institute of Health –SIVIGILA– and the Aburra Valley Early Warning System –SIATA–. Our objective is to identify possible Dengue outbreaks, i.e. to create an early warning system, to provide a preventive timeline to the health authorities to design contingency plans and to mitigate the impact of dengue on the population of Medellin. Our results indicate that a artificial neural network forecasting for time series shows a trend for the correct prediction of dengue cases up to the first four weeks with a deterioration in precision as the forecast is pushed for additional ten to twenty weeks.

Colorectal cancer (CRC) is a major public health issue by its high incidence and mortality rate. CRC appears as premalignant lesions growing in the endoluminal wall, called polyps. Currently, a regular screening of CRC during a colonoscopy is the standard procedure to localize and treat polyps. However, evidence suggests 20% - 24% of adenomatous polyps may be missed during a routine colonoscopy. A limited adenoma detection (ADRs) is obtained because colon exploration is a very challenging task: it is highly dependent on the expert training and colon preparation. Hence, a second reader is required to support CRC screening. This paper presents a novel automatic computer-aided method to localize polyps in colonoscopy images. The method starts by segmenting an input frame into a set of superpixels, each of them characterized by concatenating color, texture, and shape features computed either locally, i.e., basic local statistics, or regionally, i.e., any measure is modulated by information in neighboring superpixels. Afterward, this representation feeds a classifier which sets a probability and a polyp is a group of superpixels with high assigned probability. Finally, the resultant groups were enclosed by a bounding box which corresponds to the colorectal polyp localization. The proposed approach was trained with 200 polyps (350 images) and tested with 86 polyps (236 images) of different size. Performance of our method was compared with a baseline based on deep CNN obtaining an average of Annotated Area Covered of 0.90 vs 0.89 and a precision of 0.96 vs 0.95 respectively.

In this work, we present a shape-based approach to automatic skin lesion segmentation and classification in dermoscopic images. We aim to differentiate three types of lesion 1) common nevi, 2) atypical nevi, and 3) melanomas by exploring the morphology features of segmented skin lesions. Our method is an attempt to design a computer-aided ABCDEs of melanoma, where the Asymmetry and Border components are estimated using morphological features. The lesions are first segmented using a super-pixel merging strategy with an RGB criterion. Later, the segmentation method was evaluated on the PH² dataset, and compared with other state-of- the-art skin segmentation methods. The classification was also conducted on the PH² dataset through a 10-fold cross-validation set-up with a training and testing set partition of 90% and 10% respectively. We employed logistic regression, SVM and a neural network as classification algorithms. The best performances was 86.5% on average with the neural network.

Cutaneous ulcer caused by Leishmaniasis is a neglected disease which is more common in low-income areas. The main challenge in this disease is its diagnosis; the lack of specialized physicians makes its diagnosis difficult since quite often it can be miss-diagnosed with other type of skin ulcers such us venous, diabetes, and others. Given the previous mentioned facts, cutaneous ulcers caused by cutaneous Leishmaniasis require for the creation of novel tools that could assist its diagnosis. Hyperspectral and multispectral images measure the radiance reflected and emitted by a surface in hundreds or tens of spectral bands along the electromagnetic spectrum. This type of systems has been used for the analysis of cutaneous pathologies such us cancer, vitiligo, melasma, among others. With a set of classified hyperspectral images of cutaneous ulcers caused by different pathologies, it is possible to create an algorithm based on a multilayer neural network in order to achieve a classification of different types of ulcers. In this article we present the design of a feed-forward artificial neural network for the classification of cutaneous ulcers’ hyperspectral images in 4 kind of causes: occlusive vasculopathy, venous, Leishmaniasis, and diabetic. As result, a neural network structure is obtained that achieves a percentage of success higher than 72% in the classification of data.

Nowadays magnetic resonance images (MRI) are being used to calculate important clinical parameters such as ejection fraction (EF), left ventricle myocardium mass (MM), and stroke volume (SV) which are crucial to estimate the cardiac function, surgical planning and create patient-specific heart models, therefore for quantifying accurately these parameters it is also necessary a good delimitation of cardiac structures. The proposed approach presents an automatic segmentation of the left ventricle (LV) and basically is composed by three steps: first, heart structure localization with template matching technique in coronal and sagittal view that is used to restrict axial analysis. Second, ellipsoidal approximation using the axial projection of previous coarse segmentation. Third, a conventional snake algorithm is performed to refine external myocardium boundaries in axial view. The strategy was evaluated using 100 cardiac MRI volumes provided by the ACDC 2017 MICCAI challenge which is composed of 4 different heart diseases, the strategy had an average Dice Score of 0.79.

This paper presents a novel method for the automatic design of convolutional gray-level templates for detecting coronary arteries in X-ray angiograms. The proposed method uses the metaheuristic of iterated local search (ILS) to address the high-dimensional problem (O(256ⁿ)) involved in the design of convolutional templates. This automatically generated template is convolved in the spatial domain at different orientations to form a directional filter bank in order to detect coronary arteries at different angular resolutions. The vessel detection results are compared with those obtained by four state-of-the-art vessel enhancement methods in terms of the area (A_z) under the receiver operating characteristic (ROC) curve. The proposed method achieved the highest detection results with A_z = 0.9405 using a training set of 50 angiograms. Moreover, the convolutional gray-level template obtained from the training step, it was directly evaluated with an independent test set of 50 X-ray angiograms obtaining an A_z = 0.9565, which is the highest performance according to the comparative analysis. In addition to the experimental results, the use of metaheuristics for designing convolutional gray-level templates obtains suitable results to be considered in systems that perform computer-aided diagnosis, and it also represents an encouraging area for future research.

Attention-deficit/ hyperactivity disorder (ADHD) is the most common neurodevelopment disorder in children, and many genetic markers have been linked to the behavioral phenotypes of this highly heritable disease. The neuroimaging correlates are similarly complex, with multiple combinations of structural and functional alterations associated with the disease presentations of hyperactivity and inattentiveness. Thus far, neuroimaging studies have provided mixed results in ADHD patients, particularly with respect to the laterality of findings. It is possible that hemispheric asymmetry differences may help reconcile the variability of these findings. We recently reported that inter-hemispheric asymmetry differences were more sensitive descriptors for identifying differences between ADHD and typically developing (TD) brains (n=849) across volumetric, morphometric, and white matter neuroimaging metrics. Here, we examined the replicability of these findings across a new data set (n=202) of TD and ADHD subjects at the time of diagnosis (medication naive) and after a six week course of either stimulant drugs, nonstimulant medications, or combination therapy. Our findings replicated our earlier work across a number of volumetric and white matter measures confirming that asymmetry is more robust at detecting differences between TD and ADHD brains. However, the effects of medication failed to produce significant alterations across either lateralized or symmetry measures. We suggest that the delay in brain volume maturation observed in ADHD youths may be hemisphere dependent. Future work may investigate the extent to which these inter-hemispheric asymmetry differences are causal or compensatory in nature.