Sequential Forward Floating Feature Selection Research Articles

A condition-independent framework for the classification of error-related brain activity.

The cognitive processing and detection of errors is important in the adaptation of the behavioral and learning processes. This brain activity is often reflected as distinct patterns of event-related potentials (ERPs) that can be employed in the detection and interpretation of the cerebral responses to erroneous stimuli. However, high-accuracy cross-condition classification is challenging due to the significant variations of the error-related ERP components (ErrPs) between complexity conditions, thus hindering the development of error recognition systems. In this study, we employed support vector machines (SVM) classification methods, based on waveform characteristics of ErrPs from different time windows, to detect correct and incorrect responses in an audio identification task with two conditions of different complexity. Since the performance of the classifiers usually depends on the salience of the features employed, a combination of the sequential forward floating feature selection (SFFS) and sequential forward feature selection (SFS) methods was implemented to detect condition-independent and condition-specific feature subsets. Our framework achieved high accuracy using a small subset of the available features both for cross- and within-condition classification, hence supporting the notion that machine learning techniques can detect hidden patterns of ErrP-based features, irrespective of task complexity while additionally elucidating complexity-related error processing variations. Graphical abstract A schematic of the proposed approach. (a) EEG recordings in an auditory experiment in two conditions of different complexity. (b) Characteristic event related activity feature extraction. (c) Selection of feature vector subsets for easy and hard conditions corresponding to correct (Class1) and incorrect (Class2) responses. (d) Performance for individual and cross-condition classification.

Radiomic features to predict microvascular invasion in hepatocellular carcinoma based on conventional MRI: preliminary findings

Objective To identify the preoperative MRI findings for predicting microvascular invasion (MVI) using texture analysis (TA) on multiple MRI sequences. Methods Two hundred and fifty patients with HCC pathologically confirmed by surgery in Zhongshan Hospital from October 2015 to October 2016 were analyzed retrospectively. All patients underwent conventional MRI plain scan and dynamic contrast-enhanced examination within 2 weeks before operation. According to the ratio of 1∶1, the patients were divided into a training set (125 cases) and a test set (125 cases).The training set was used to establish a classifier to predict MVI of HCCs via the TA, and the test set was used to evaluate the performance of the classifier. An image analysis was performed using an in-house software contained a set of 2 415 features which were generated from all conventional axial sequences, including the T2WI, DWI, ADC map, and dynamic enhancement images.. A four-fold cross validation (FFCV) and sequential forward floating feature selection strategy (SFFS) were employed to select an optimal subset of features and a linear discriminant analysis (LDA) was employed to establish a classifier. The clinical laboratory examination, morphologic characteristics and quantitative analysis of conventional MR were used to compare the performance of predicting MVI with the classifier. A Chi-squared test or Fisher exact probablities test were used for categorical variables, and independent t test or Mann-Whitney U test were used for used for continuous variables. Factors with a P value less than 0.05 at univariate analyses were entered into the multivariate model to identify independent predictors. The Hosmer-Lemeshow test was performed to explain the goodness of fit of the multivariate logistic model. A receiver operating characteristic curve analysis was performed to evaluate the diagnostic performance. Results The classifier set up by the training set consists of 13 texture features. When conventional MRI texture features of test set were used to judge whether there was MVI or not, the AUC of all texture features of arterial phase (AP) was the highest (0.506 3). Univariate regression analysis showed that there were significant differences in pathological grade (P=0.026), AFP level (P=0.033), lesion edge shape (P=0.038), AP enhancement (P=0.038), and AP peritumoral enhancement (P=0.008). Multivariate binary logistic regression analysis showed that peritumoral enhancement and texture classifier assessed MVI with P values of 0.005 and 0.001,which were independent risk factors for MVI. The significance level of Hosmer Lemeshow test was 0.796, indicating the goodness of fit of acceptable models. The AUCs of single variable, combined variable (including of AFP level, irregular tumor margin, enhancement intensity in AP and peritumoral enhancement in AP) and texture classifier for MVI were 0.588 to 0.627, 0.798 and 0.733, respectively. When compared the AUC of the combination features (including of AFP level, irregular tumor margin, enhancement intensity in AP and peritumoral enhancement in AP) with the classifier to identify MVI of HCC in the test set, no significant difference was found(P=0.108 6). However, although the sensitivity of them were same as 70.73%, the specificity of the combination features was mildly higher than that of classifier (82.14% vs. 78.57%). Conclusions Combination features of AFP level, tumor margin, enhancement intensity in AP and peritumoral enhancement in AP can be used to predict MVI of HCCs. It is a new method of noninvasive evaluation of MVI before operation. The performance of the classifier made by TA was not superior to that of combination features based on clinic and conventional MR sequences. Key words: Liver neoplasms; Magnetic resonance imaging; Radiomics; Microvascular invasion

Instantaneous fMRI based cerebral parameters for automatic Alzheimer, mild cognitive impairment and healthy subject classification.

Automatic identification and categorization of Alzheimer's patients and the ability to distinguish between different levels of this disease would be very helpful to the research community studying this disease since non-automatic approaches are both very time-consuming and highly dependent upon the experience of experts. Here, it is proposed that instantaneous cerebral phase and envelope information from functional magnetic resonance imaging data is of use to discriminate between Alzheimer's patients, mild cognitively impaired subjects and healthy individuals. Following a region-of-interest analysis of functional magnetic resonance imaging data, different features including power, entropy, and coherency features are derived from the instantaneous phase and envelope signal sequences. Various sets of features are calculated and fed to a sequential forward floating feature selection algorithm to identify the most discriminative and informative feature sets. A Student's t-test was employed to select the most relevant features from the sets. Finally, a K-nearest neighbor classifier is used to distinguish between classes in a three-class categorization problem. The reported performance in overall accuracy using functional magnetic resonance imaging data of 111 combined participants is 80.1% with 80.0% sensitivity for the distinction of both Alzheimer's and healthy categories. This is comparable to the state-of-the-art approaches recently proposed for this task. The significance of obtained results was statistically confirmed by the evaluation of standard classification performance indicators. Results illustrate that the analytic phase and envelope feature indices derived from the region of interest signals described here are significant discriminators suited to distinguish between Alzheimer patients and healthy subjects.

Development of a Computer-Aided Differential Diagnosis System to Distinguish Between Usual Interstitial Pneumonia and Non-specific Interstitial Pneumonia Using Texture- and Shape-Based Hierarchical Classifiers on HRCT Images.

A computer-aided differential diagnosis (CADD) system that distinguishes between usual interstitial pneumonia (UIP) and non-specific interstitial pneumonia (NSIP) using high-resolution computed tomography (HRCT) images was developed, and its results compared against the decision of a radiologist. Six local interstitial lung disease patterns in the images were determined, and 900 typical regions of interest were marked by an experienced radiologist. A support vector machine classifier was used to train and label the regions of interest of the lung parenchyma based on the texture and shape characteristics. Based on the regional classifications of the entire lung using HRCT, the distributions and extents of the six regional patterns were characterized through their CADD features. The disease division index of every area fraction combination and the asymmetric index between the left and right lungs were also evaluated. A second SVM classifier was employed to classify the UIP and NSIP, and features were selected through sequential-forward floating feature selection. For the evaluation, 54 HRCT images of UIP (n=26) and NSIP (n=28) patients clinically diagnosed by a pulmonologist were included and evaluated. The classification accuracy was measured based on a fivefold cross-validation with 20 repetitions using random shuffling. For comparison, thoracic radiologists assessed each case using HRCT images without clinical information or diagnosis. The accuracies of the radiologists' decisions were 75 and 87%. The accuracies of the CADD system using different features ranged from 70 to 81%. Finally, the accuracy of the proposed CADD system after sequential-forward feature selection was 91%.

DAG-SVM based infant cry classification system using sequential forward floating feature selection

Crying is the infant’s first verbal communication. Before learning how to express the emotions or physiological/psychological requirements with language, infants usually express how they feel through crying. Crying is a response to a stimulus such as hunger, pain, or discomfort. However, it is sometimes difficult to figure out why an infant is crying. This can certainly be frustrating or even frightening for a caretaker, and so we in this paper have proposed an infant cry classification system to categorize the types of infant crying to help parents and nursing staffs attending to the needs of the infants. Currently, three kinds of distinct infant cries have been identified: hunger; pain; and feeling sleepy. Fifteen features are extracted from each crying frame and the sequential forward floating selection is then adopted to pick out high discriminative features. The directed acyclic graph support vector machine is finally used to classify infant crying. Experimental results have revealed the good performance of the proposed system and the classification accuracy is up to 92.17 %.