Alejandro Veloz

Self-Organizing Neuro-Fuzzy Inference System

The architectural design of neuro-fuzzy models is one of the major concern in many important applications. In this work we propose an extension to Rogerss ANFIS model by providing it with a selforganizing mechanism. The main purpose of this mechanism is to adapt the architecture during the training process by identifying the optimal number of premises and consequents needed to satisfy a users performance criterion. Using both synthetic and real data, our proposal yields remarkable results compared to the classical ANFIS.

Fuzzy spatial growing for glioblastoma multiforme segmentation on brain magnetic resonance imaging

Image segmentation is a fundamental technique in medical applications. For example, the extraction of biometrical parameter of tumors is of paramount importance both for clinical practice and for clinical studies that evaluate new brain tumor therapies. Tumor segmentation from brain Magnetic Resonance Images (MRI) is a difficult task due to strong signal heterogeneities and weak contrast at the boundary delimitation. In this work we propose a new framework to segment the Glioblastoma Multiforme (GBM) from brain MRI. The proposed algorithm was constructed based on two well known techniques: Region Growing and Fuzzy C-Means. Furthermore, it considers the intricate nature of the GBM in MRI and incorporates a fuzzy formulation of Region Growing with an automatic initialization of the seed points. We report the performance results of our segmentation framework on brain MRI obtained from patients of the chilean Carlos Van Buren Hospital and we compare the results with Region Growing and the classic Fuzzy C-Means approaches.

Modified expectation maximization algorithm for MRI segmentation

Magnetic Resonance Image segmentation is a fundamental task in a wide variety of computed-based medical applications that support therapy, diagnostic and medical applications. In this work, spatial information is included for estimating paramaters of a finite mixture model, with gaussian distribution assumption, using a modified version of the well-know Expectation Maximization algorithm proposed in [3]. Our approach is based on aggregating a transition step between E-step and M-step, that includes the information of spatial dependences between neighboring pixels. Our proposal is compared with other approaches proposed in the image segmentation literature using the size and shape test, obtaining accurate and robust results in the presence of noise.

A Flexible Neuro-Fuzzy Autoregressive Technique for Non-linear Time Series Forecasting

The aim of this paper is to simultaneously identify and estimate a non-linear autoregressive time series using a flexible neuro-fuzzy model. We provide a self organization and incremental mechanism to the adaptation process of the neuro-fuzzy model. The self organization mechanism searches for a suitable set of premises and consequents to enhance the time series estimation performance, while the incremental method selects influential lags in the model description. Experimental results indicate that our proposal reliably identifies appropriate lags for non-linear time series. Our proposal is illustrated by simulations on both synthetic and real data.

Brain Tumors: How Can Images and Segmentation Techniques Help?

These days, cancer is one of the diseases that scares people the most. Brain cancer may be considered among the most difficult cancers to treat, as it involves the organ which is not only in control of the body, but is also responsible for the self-definition of the person. During surgery or any kind of treatment, eloquent areas must not be affected in order to minimize iatrogenic risks. Therefore good diagnosis and planning of treatment choices is essential. This is why images are now of paramount importance in the evaluation of brain tumors: oncologists, neurosurgeons and the entire medical team need to know how to understand them and how to use the current tools provided by computational techniques to take advantage of the information retrieved from them. A wide variety of images is available to support the physician’s actions at different levels, including diagnosis, treatment election, interventional support, and follow-up. Investigation in this area is very active; attempts are being made to go beyond the current pixel resolution, and to gain information with “molecular images”; not only in nuclear medicine but also in magnetic resonance images. Everybody agrees that images are now an invaluable service in the practice of medicine. However, the present and future use of images is intrinsically associated with larger numbers of images, which are not easily manageable by either radiologists or surgeons. Neuroradiology is conceived as a discipline in which the health status of a patient is inferred according to the visual inspection of images taken from different modalities. This implies that the success of the clinical diagnosis depends on the physician’s particular skills, and also on the information that the clinical team can handle. In addition, numerous image modalities are used frequently at different time points; therefore there is also a need for integration of the features reflected by these different sources of images. In order to provide support for this integration, automatic processing methods have been developed. Many Computer Aided Diagnostic (CAD) software packages have been developed, in particular to provide second readings in mammography, lung or brain cancer (Doi, 2007). These developments have motivated several clinical applications. Regarding brain tumor image processing, what is usually expected is to detect the localization and extension of the tumor, in other words to segment the tumor in the image.

Neuro-fuzzy-based Arrhythmia Classification Using Heart Rate Variability Features

Arrhythmia diagnosis is commonly conducted through visual analysis of human electrocardiograms, a very resource consuming task for physicians. In this paper we present a computational approach for arrhythmia detection based on heart rate variability signal analysis and the application of a neuro-fuzzy classification model called SONFIS. The aforementioned method generates a set of linguistically interpretable inference rules for pattern classification and outperforms artificial neural networks and support vector machines in accuracy and several other performance indicators.

Identification of Lags in Nonlinear Autoregressive Time Series Using a Flexible Fuzzy Model

This work proposes a method to find the set of the most influential lags and the rule structure of a Takagi–Sugeno–Kang (TSK) fuzzy model for time series applications. The proposed method resembles the techniques that prioritize lags, evaluating the proximity of nearby samples in the input space using the closeness of the corresponding target values. Clusters of samples are generated, and the consistency of the mapping between the predicted variable and the set of candidate past values is evaluated. A TSK model is established, and possible redundancies in the rule base are avoided. The proposed method is evaluated using simulated and real data. Several simulation experiments were conducted for five synthetic nonlinear autoregressive processes, two nonlinear vector autoregressive processes and eight benchmark time series. The results show a competitive performance in the mean square error and a promising ability to find a proper set of lags for a given autoregressive process.

SONFIS: Structure Identification and Modeling with a Self-Organizing Neuro-Fuzzy Inference System

AbstractThis paper presents a new adaptive learning algorithm to automatically design a neural fuzzy model. This constructive learning algorithm attempts to identify the structure of the model based on an architectural self-organization mechanism with a data-driven approach. The proposed training algorithm self-organizes the model with intuitive adding, merging and splitting operations. Sub-networks compete to learn specific training patterns and, to accomplish this task, the algorithm can either add new neurons, merge correlated ones or split existing ones with unsatisfactory performance. The proposed algorithm does not use a clustering method to partition the input-space like most of the state of the art algorithms. The proposed approach has been tested on well-known synthetic and real-world benchmark datasets. The experimental results show that our proposal is able to find the most suitable architecture with better results compared with those obtained with other methods from the literature.

SIFAR: Self-Identification of Lags of an Autoregressive TSK-based Model

In this work, a Takagi-Sugeno-Kang (TSK) model is used for time series analysis and some important questions about the identification of this kind of models are addressed: the identification of the model structure and the set of the most influential regressors or lags. The main idea behind of the proposed method resembles to those techniques that prioritize lags evaluating the proximity of nearby samples in the input space in relation to the closeness of the corresponding target values. Clusters of samples are generated and the consistence of the mapping between the predicted variable and the set of candidate past values is evaluated. Afterwards, a TSK model is established and the redundancies in the rule base are avoided. Simulation experiments were conducted for 2 synthetic nonlinear autoregressive processes and for 4 benchmark time series. Results show a promising performance in terms of forecasting error and in terms of ability to find a proper set of lags of a given autoregressive process.

Photoplethysmogram Fits Finger Blood Pressure Waveform for non-Invasive and minimally-Intrusive Technologies - Evaluation of Derivative Approaches.

The purpose of this work is to fit Photoplethysmography (PPG) to finger Arterial Pressure (fiAP) waveform using derivative approaches. Derivative approaches consider using Linear Combination of Derivatives (LCD) and Fractional Derivatives (FDPα). Four informed healthy subjects, aging 35.8± 11.0 years old, agreed to perform Handgrip maneuvers. Signals are recorded continually; a Finapres NOVA device is used for fiAP, while a BIOPAC System is used for PPG and ECG. PPG is smoothed and segmented by heartbeat; recording sections interfered with spiky blocking noise, are eliminated. Finally, PPG is processed using LCD and FDPα and their results are enriched using Lasso technique. Twenty records per subject at rest and twenty at raised BP are analyzed. Results show PPG to fiAP fitting errors 5.38%± 0.91 at resting fiAP and 5.86%± 1.21 at raised fiAP, being always lower than 15%, suggesting that derivative approaches could be suitable for medical