Andrés M. Álvarez

Nerve Structure Segmentation from Ultrasound Images Using Random Under-Sampling and an SVM Classifier.

The identification of nerve structures is a crucial issue in the field of anesthesiology. Recently, ultrasound images have become relevant for performing Peripheral Nerve Blocking (PNB) procedures since it offers a non-invasive visualization of the nerve and the anatomical structures around it. However, the location of nerve structures from ultrasound images is a difficult task for the specialist due to the artifacts, i.e., speckle noise, which affect the intelligibility of a given image. Here, we proposed an automatic nerve structure segmentation approach from ultrasound images based on random under-sampling (RUS) and a support vector machine (SVM) classifier. In particular, we use a Graph Cuts-based technique to define a region of interest (ROI). Then, such an ROI is split into several correlated areas (superpixels) using the well-known Simple Linear Iterative Clustering algorithm. Further, a nonlinear Wavelet transform is applied to extract relevant features. Afterward, we use a classification scheme based on RUS and SVM to predict the label of each parametrized superpixel. Thus, our approach can deal with the imbalance issues when classifying a superpixel as nerve or non-nerve. Attained results on a real-world dataset demonstrate that our method outperforms similar works regarding both the dice segmentation coefficient and the geometric mean-based classification assessment.

Shape Classification Using Hilbert Space Embeddings and Kernel Adaptive Filtering

Shape classification is employed for realizing image object identification and classification tasks. Most of the state-of-the-art approaches use sequential features extracted from contours to classify shapes, either directly, i.e., k-nearest neighbors (KNN), or through stochastic models, i.e., hidden Markov models (HMMs). Here, inspired by probability based metrics using Hilbert space embedding (HSE), we introduce a novel scheme for efficient shape classification. To this end, we highlight relevant curvature patterns from binary images towards a Kernel Adaptive Filtering (KAF)-based enhancement of the maximum mean discrepancy metric. Namely, we test the performance of our approach on the well-known MPEG-7 and 99-Shapes databases. Results show that our strategy can code relevant shape properties from binary images achieving competitive classification results.

A Kernelized Morphable Model for 3D Brain Tumor Analysis

Abnormal tissue analysis in brain volumes is a difficult task, due to the shape variability that the brain tumors exhibit between patients. The main problem in these processes is that the common techniques use linear representations of the input data which makes unsuitable to model complex shapes as brain tumors. In this paper, we present a kernelized morphable model (3D-KMM) for brain tumor analysis in which the model variations are captured through nonlinear mappings by using kernel principal component analysis. We learn complex shape variations through a high-dimensional representation of the input data. Then from the trained model, we recover the pre-images from the features vectors and perform a non-rigid matching procedure to fit the modeled tumor to a given brain volume. The results show that by using a kernelized morphable model, the non-rigid properties (i.e., nonlinearities and shape variations) of the abnormal tissues can be learned. Finally, our approach proves to be more accurate than the classic morphable model for shape analysis.

Student Desertion Prediction Using Kernel Relevance Analysis

This paper presents a kernel-based relevance analysis to support student desertion prediction. Our approach, termed KRA-SD, is twofold: (i) A feature ranking based on centered kernel alignment to match demographic, academic, and biopsychosocial measures with the output labels (deserter/not deserter), and (ii) classification stage based on k-nearest neighbors and support vector machines to predict the desertion. For concrete testing, the student desertion database of the Universidad Tecnologica de Pereira is employed to assess the KRA-SD under a training, validation, and testing scheme. Attained results show that the proposed approach can recognize the main features related to the student desertion achieving an 85.64% of accuracy. Therefore, the proposed system aims to serve as a handy tool for planning strategies to prevent students from leaving the university without finishing their studies.

Multilayer-Based HMM Training to Support Bearing Fault Diagnosis

The bearings are among the most critical components in rotating machinery. For this reason, fault diagnosis in those elements is essential to avoid economic losses and human casualties. Traditionally, the automatic bearing fault diagnosis has been addressed by approaches based on Hidden Markov Models (HMM). However, the efficiency and reliability of the HMM-based diagnostic systems are still relevant topics for many researchers. In this paper, we present a modified training approach based on multilayer partition to support bearing fault diagnosis, that we called MHMM. The proposed strategy seeks to increase the system efficiency by reducing the number of HMM required to perform a proper diagnosis, making it more intelligent and suitable for this application. For concrete testing, the bearing fault databases from the Western Case Reserve University and the Politecnica Salesiana University were employed to assess the MHMM under a training and testing scheme. Attained results show that the proposed approach can effectively reduce the number of models required to perform the diagnosis while keeping high accuracy ratings when we compare the MHMM with the benchmarks. Also, the diagnosis process time is reduced as well.

3D Probabilistic Morphable Models for Brain Tumor Segmentation.

Segmenting abnormal areas in brain volumes is a difficult task, due to the shape variability that the brain tumors exhibit between patients. The main problem in these processes is that the common segmentation techniques used in these tasks, lack of the property of modeling the shape structure that the tumor presents, which leads to an inaccurate segmentation. In this paper, we propose a probabilistic framework in order to model the shape variations related to abnormal tissues relevant in brain tumor segmentation procedures. For this purpose the database of the Brain Tumor Image Segmentation Challenge (Brats) 2015 is used. We use a Probabilistic extension of the 3D morphable model to learn those tumor variations between patients. Then from the trained model, we perform a non-rigid matching to fit the deformed modeled tumor in the medical image. The experimental results show that by using Probabilistic morphable models, the non-rigid properties of the abnormal tissues can be learned and hence improve the segmentation task.

Automatic Peripheral Nerve Segmentation in Presence of Multiple Annotators.

Peripheral Nerve Blocking (PNB) is a technique commonly used to perform regional anesthesia. The success of PNB procedures lies of the accurate location of the target nerve. The ultrasound images (UI) have frequently been used aiming to locate nerve structures in the context of PNB procedures. This type of images allows a direct visualization of the target nerve, and the anatomical structures around it. Notwithstanding, the nerve segmentation in UI by an anesthesiologist is not straightforward since these images are affected by several artifacts; hence, the accuracy of nerve segmentation depends on the anesthesiologist expertise. In this sense, we face a scenario where we have manual multiple nerve segmentations performed by several anesthesiologists with different levels of expertise. In this paper, we propose a nerve segmentation approach based on supervised learning. For the classification step, we compare two schemes based on the concepts “Learning from crowds” aiming to code the information of multiple manual segmentations. Attained results show that our approach finds a suitable UI approximation by ensuring the identification of discriminative nerve patterns according to the opinions given by multiple specialists.

Clustering-Based Undersampling to Support Automatic Detection of Focal Cortical Dysplasias

Focal Cortical Dysplasias (FCDs) are cerebral cortex abnormalities that cause epileptic seizures. Recently, machine learning techniques have been developed to detect FCDs automatically. However, dysplasias datasets contain substantially fewer lesional samples than healthy ones, causing high order imbalance between classes that affect the performance of machine learning algorithms. Here, we propose a novel FCD automatic detection strategy that addresses the class imbalance using relevant sampling by a clustering strategy approach in cooperation with a bagging-based neural network classifier. We assess our methodology on a public FCDs database, using a cross-validation scheme to quantify classifier sensitivity, specificity, and geometric mean. Obtained results show that our proposal achieves both high sensitivity and specificity, improving the classification performance in FCD detection in comparison to the state-of-the-art methods.

Non-stationary Multi-output Gaussian Processes for Enhancing Resolution over Diffusion Tensor Fields

Diffusion magnetic resonance imaging (dMRI) is an advanced technique derived from magnetic resonance imaging (MRI) that allows the study of internal structures in biological tissue. Due to acquisition protocols and hardware limitations of the equipment employed to obtain the data, the spatial resolution of the images is often low. This inherent lack in dMRI is a considerable difficulty because clinical applications are affected. The scientific community has proposed several methodologies for enhancing the spatial resolution of dMRI data, based on interpolation of diffusion tensors fields. However, most of the methods have considerable drawbacks when they interpolate strong transitions, such as crossing fibers. Also, relevant clinical information from tensor fields is modified when interpolation is performed. In this work, we propose a probabilistic methodology for interpolation of diffusion tensors fields using multi-output Gaussian processes with non-stationary kernel function. First, each tensor is decomposed in shape and orientation features. Then, the model interpolates the features jointly. Results show that proposed approach outperforms state-of-the-art methods regarding resolution enhancement accuracy on synthetic and real data, when we evaluate interpolation quality with Frobenius and Riemann metrics. Also, the proposed method demonstrates an adequate characterization of both stationary and non-stationary fields, contrary to previous approaches where performance is seriously reduced when complex fields are interpolated.

A Kernel-Based Approach for DBS Parameter Estimation

The volume of tissue activated (VTA) is commonly used as a tool to explain the effects of deep brain stimulation (DBS). The VTA allows visualizing the anatomically accurate reconstructions of the brain structures surrounding the DBS electrode as a 3D high-dimensional activate/non-activate image, which leads to important clinical applications, e.g., Parkinson’s disease treatments. However, fixing the DBS parameters is not a straightforward task as it depends mainly on both the specialist expertise and the tissue properties. Here, we introduce a kernel-based approach to learn the DBS parameters from VTA data. Our methodology employs a kernel-based eigendecomposition from pair-wise Hamming distances to extract relevant VTA patterns into a low-dimensional space. Further, DBS parameters estimation is carried out by employing a kernel-based multi-output regression and classification. The presented approach is tested under both isotropic and anisotropic conditions to validate its performance under realistic clinical environments. Obtained results show a significant reduction of the input VTA dimensionality after applying our scheme, which ensures suitable DBS parameters estimation accuracies and avoids over-fitting.