Figure 1 - uploaded by Alfredo Vellido
Content may be subject to copyright.
Schematic illustration of the Moving Window technique.
Source publication
The diagnosis of brain tumours is an extremely sensitive and complex clinical task that must rely upon information gathered through non-invasive techniques. One such technique is magnetic resonance, in the modalities of imaging or spectroscopy. The latter provides plenty of metabolic information about the tumour tissue, but its high dimensionality...
Context in source publication
Context 1
... mathematical function can be used to find optimal sub-regions within informative frequency intervals, and also to perform extraction and/or selection of features. Figure 1 illustrates this idea. When w = 1, the MW moves along the signal, travelling from the first sample to the last, creating m-w+1 outlets (m = number of samples). ...
Similar publications
![Characteristics of the dataset [36]](profile/Ammar-Altaher/publication/371947097/figure/tbl1/AS:11431281171123953@1688031884486/Characteristics-of-the-dataset-36_Q320.jpg)
The high human labor demand involved in collecting paired medical imaging data severely impedes the application of deep learning methods to medical image processing tasks such as tumor segmentation. The situation is further worsened when collecting multi-modal image pairs. However, this issue can be resolved through the help of generative adversari...
Citations
... All rights reserved. types) often complicate their diagnostic-oriented classification (Arizmendi, Vellido, & Romero, 2009). This paper aims to contribute new tools for the automated classification analysis of brain tumours from MRS data. ...
... These can be associated to specific metabolites to improve the interpretability of the results, which is a key goal in medical practice (Vellido, Martín-Guerrero, & Lisboa, 2012). The second step of the method involves dimensionality reduction in the form of a feature selection filter method known as Moving Window with Variance Analysis (MWVA) (Arizmendi et al., 2009). In the third and final step, the remaining relevant information is fed to an Artificial Neural Network (ANN) classifier with Bayesian regularization (MacKay, 1992). ...
... In this paper, we use MWVA: a feature selection filter method proposed in Arizmendi et al. (2009), which consists of the combination of the Moving Window technique in conjunction with the calculation of a standard ratio X, defined as the quotient between the between-groups variance (BGV) and the within-groups variance (WGV) for a particular width w of the window. The reader is referred to Arizmendi et al. (2009) for more details on this method. ...
... This decomposition process by itself does not reduce the high dimensionality of the data. For this reason, the DWT process is followed by dimensionality reduction using Moving Window with Variance Analysis (MWVA: Arizmendi, Vellido, & Romero, 2009 ) for feature selection or Principal Component Analysis (PCA) for feature extraction. Diagnostic classification is then accomplished using Artificial Neural Networks (ANN) with Bayesian regularization. ...
... The MWVA is a feature selection filter method proposed in Arizmendi, et al. (2009), which consists of the combination of the Moving Window technique in conjunction with the calculation of a standard ratio X, defined as the quotient between the between-groups variance (BGV) and the within-groups variance (WGV) for a particular width w of the window:Fig. 6. Mean ± standard deviation of O, in descending order, as a function of the number of variables. ...
... 5 displays a graphic illustration of the DIM matrix for the specific experiment gl vs. no (see tumour labels inTable 1), where good values for the width of the window can be visually inspected by looking for large values of X(w, t). In our study, and following the procedure described in (Arizmendi, et al. 2009 ), the value of the optimal width w was found to be 1. Therefore , every window corresponds to a single variable. ...
... All rights reserved. types) often complicate their diagnostic-oriented classification (Arizmendi, Vellido, & Romero, 2009). This paper aims to contribute new tools for the automated classification analysis of brain tumours from MRS data. ...
... These can be associated to specific metabolites to improve the interpretability of the results, which is a key goal in medical practice (Vellido, Martín-Guerrero, & Lisboa, 2012). The second step of the method involves dimensionality reduction in the form of a feature selection filter method known as Moving Window with Variance Analysis (MWVA) (Arizmendi et al., 2009). In the third and final step, the remaining relevant information is fed to an Artificial Neural Network (ANN) classifier with Bayesian regularization (MacKay, 1992). ...
... In this paper, we use MWVA: a feature selection filter method proposed in Arizmendi et al. (2009), which consists of the combination of the Moving Window technique in conjunction with the calculation of a standard ratio X, defined as the quotient between the between-groups variance (BGV) and the within-groups variance (WGV) for a particular width w of the window. The reader is referred to Arizmendi et al. (2009) for more details on this method. ...
... In this study, we analyze a set of MRS data from the multi-centre, international INTERPRET database [2]. We do so using several methodologies that involve signal processing, feature selection and classification, namely and in turn: Gaussian Decomposition (GD) to transform the signal in terms of coefficients of amplitude, standard deviation, and translation [3]; moving window with variance analysis (MWVA) [4] [7].The proposed combination of techniques is shown to yield high diagnostic classification accuracy for a broad range of brain tumour pathologies, some of which have seldom been analyzed in this setting. ...
... When one such a repetition in translation was identified, the lowest of both translations was moved the closest lower integer value, whereas the highest of both was moved to the closest higher integer value, thus avoiding the overlapping. Having re-arranged the data this way, feature selection was carried out using the MWVA technique [4], [11], from the re-scaled amplitude and standard deviation vectors. A modification of MWVA was implemented, with the dissimilarity index matrix (DIM) obtained in each experiment. ...
... This was accomplished by concatenating the amplitude and standard deviation DIM's. Once the new DIM was obtained, feature selection was carried out using the energy criterion described in [4]. ...
The development, implementation and use of computer-based medical decision support systems (MDSS) based on pattern recognition techniques holds the promise of substantially improving the quality of medical practice in diagnostic and prognostic tasks. In this study, the core of a decision support system for brain tumour classification from magnetic resonance spectroscopy (MRS) data is presented. It combines data pre-processing using Gaussian decomposition, dimensionality reduction using moving window with variance analysis, and classification using artificial neural networks (ANN). This combination of techniques is shown to yield high diagnostic classification accuracy in problems concerning diverse brain tumour pathologies, some of which have received little attention in the literature.
... Dimensionality reduction was implemented according to two strategies: feature selection using MWVA and feature extraction using PCA (of common use in radiology data analysis). MWVA is a feature selection filter method first proposed in [6]. It is based on a combination of the moving window technique and the analysis of between/within group variance. ...
The accurate diagnosis of human brain tumours is a sensitive medical task, for which radiology experts often must rely on indirect signal measurements. There is thus a need for developing computer-based decision support tools to assist doctors in their diagnostic task. The experiments in this brief paper address such problem in the form of binary classification, for which the pre-processing of the Magnetic Resonance Spectroscopy (MRS) signal is a most relevant data analysis stage. A combination of the Discrete Wavelet Transform (DWT) for signal decomposition and an energy criterion for signal reconstruction is used to pre-process the MRS data prior to the feature selection and classification with Bayesian Neural Networks.
... This decomposition by itself does not alleviate the high dimensionality of the data. For this reason, dimensionality reduction is implemented using Moving Window with Variance Analysis (MWVA) [3] for feature selection and Principal Components Analysis (PCA) for feature extraction. The processed data are classified using Artificial Neural Networks (ANN) with Bayesian regularization. ...
... After processing the MR spectra with wavelet Biortogonal (3.3) and filtering it with the combination Rigsure-Sln-Hard, we proceeded to reduce the dimensionality of the data using MWVA and PCA, taking as input variables the decomposition coefficients. The MWVA is a feature selection filter method proposed in [3]. For PCA, principal components were added one at a time until the differential cumulative variance between two consecutive components was less than 1%. ...
The diagnosis of human brain tumours from noninvasive signal measurements is a sensitive task that requires specialized expertise. In this task, radiology experts are likely to benefit from the support of computer-based systems built around robust classification processes. In this brief paper, a method that combines data pre-processing using wavelets with classification using Artificial Neural Networks is shown to yield high diagnostic classification accuracy for a broad range of brain tumour pathologies.
