Linear Discriminant Analysis-based Classifier Research Articles

MEMRI-based imaging pipeline for guiding preclinical studies in mouse models of sporadic medulloblastoma.

Genetically engineered mouse models of sporadic cancers are critical for studying tumor biology and for preclinical testing of therapeutics. We present an MRI-based pipeline designed to produce high resolution, quantitative information about tumor progression and response to novel therapies in mouse models of medulloblastoma (MB). Sporadic MB was modeled in mice by inducing expression of an activated form of the Smoothened gene (aSmo) in a small number of cerebellar granule cell precursors. aSmo mice were imaged and analyzed at defined time-points using a 3D manganese-enhanced MRI-based pipeline optimized for high-throughput. A semi-automated segmentation protocol was established that estimates tumor volume in a time-frame compatible with a high-throughput pipeline. Both an empirical, volume-based classifier and a linear discriminant analysis-based classifier were tested to distinguish progressing from nonprogressing lesions at early stages of tumorigenesis. Tumor centroids measured at early stages revealed that there is a very specific location of the probable origin of the aSmo MB tumors. The efficacy of the manganese-enhanced MRI pipeline was demonstrated with a small-scale experimental drug trial designed to reduce the number of tumor associated macrophages and microglia. Our results revealed a high level of heterogeneity between tumors within and between aSmo MB models, indicating that meaningful studies of sporadic tumor progression and response to therapy could not be conducted without an imaging-based pipeline approach.

Identification of molecular genetic contributants to canine cutaneous mast cell tumour metastasis by global gene expression analysis.

Cutaneous mast cell tumours are one of the most common canine cancers. Approximately 25% of the tumours metastasise. Activating c-kit mutations are present in about 20% of tumours, but metastases occur in the absence of mutations. Tumour metastasis is associated with significantly diminished survival in spite of adjuvant chemotherapy. Available prognostic tests do not reliably predict whether a tumour will metastasise. In this study we compared the global expression profiles of 20 primary cutaneous mast cell tumours that metastasised with those of 20 primary tumours that did not metastasise. The objective was to identify genes associated with mast cell tumour metastatic progression that may represent targets for therapeutic intervention and biomarkers for prediction of tumour metastasis. Canine Gene 1.1 ST Arrays were employed for genome-wide expression analysis of formalin-fixed, paraffin-embedded biopsies of mast cell tumours borne by dogs that either died due to confirmed mast cell tumour metastasis, or were still alive more than 1000 days post-surgery. Decreased gene expression in the metastasising tumours appears to be associated with a loss of cell polarity, reduced cell-cell and cell-ECM adhesion, and increased cell deformability and motility. Dysregulated gene expression may also promote extracellular matrix and base membrane degradation, suppression of cell cycle arrest and apoptosis, and angiogenesis. Down-regulation of gene expression in the metastasising tumours may be achieved at least in part by small nucleolar RNA-derived RNA and microRNA-effected gene silencing. Employing cross-validation, a linear discriminant analysis-based classifier featuring 19 genes that displayed two-fold differences in expression between metastasising and non-metastasising tumours was estimated to classify metastasising and non-metastasising tumours with accuracies of 90–100% and 70–100%, respectively. The differential expression of 9 of the discriminator genes was confirmed by quantitative reverse transcription-PCR.

A Housekeeping Prognostic Health Management Framework for Microfluidic Systems

Micro-electro-mechanical systems and microfluidics are becoming popular solutions for sensing, diagnostics, and control applications. Reliability and validation of function is of increasing importance in the majority of these applications. On-line testing strategies for these devices have the potential to provide real-time condition monitoring information. It is shown that this information can be used to diagnose and prognose the health of the device. This information can also be used to provide an early failure warning system by predicting the remaining useful life. Diagnostic and prognostic outcomes can also be leveraged to improve the reliability, dependability, and availability of these devices. This paper has delivered a methodology for a “lightweight” prognostics solution for a microfluidic device based on real-time diagnostics. An oscillation-based test methodology is used to extract diagnostic information that is processed using a linear discriminant analysis-based classifier. This enables the identification of current health-based on pre-defined health levels. As the deteriorating device is periodically classified, the rate at which the device degrades is used to predict the devices remaining useful life.

Detection and Classification of Nasalized Vowels in Noise Based on Cepstra Derived from Differential Product Spectrum

In this paper, a method based on cepstra derived from the differential product spectrum is developed for the detection and classification of nasalized vowels with varying degree of nasalization. Conventionally, features for detecting and classifying nasalized vowels are derived considering magnitude spectrum only, ignoring the phase spectrum. Exploiting the power spectrum and the group delay function of a band-limited vowel, the product spectrum is defined thus incorporating the information of both magnitude and phase spectra. The product spectrum is then differentiated with respect to frequency to obtain differential product spectrum (DPrS) that is argued to provide more noise robustness in the presence of noise. Unlike conventional mel-frequency cepstral coefficient (MFCC), MFCCs computed from the differential product spectrum, namely MFDPrSCCs, are fed to a linear discriminant analysis-based classifier for the detection and classification of nasalized vowels. Detailed simulation results on TIMIT database show that the proposed cepstral features derived from the differential product spectrum are capable of outperforming the cepstral features derived from the conventional power spectrum in the task of detecting and classifying nasalized vowel not only in clean condition but also in different noisy condition with varying signal to noise ratio.

Analyzing EEG signals to detect unexpected obstacles during walking.

BackgroundWhen an unexpected perturbation in the environment occurs, the subsequent alertness state may cause a brain activation responding to that perturbation which can be detected and employed by a Brain-Computer Interface (BCI). In this work, the possibility of detecting a sudden obstacle appearance analyzing electroencephalographic (EEG) signals is assessed. For this purpose, different features of EEG signals are evaluated during the appearance of sudden obstacles while a subject is walking on a treadmill. The future goal is to use this procedure to detect any obstacle appearance during walking when the user is wearing a lower limb exoskeleton in order to generate an emergency stop command for the exoskeleton. This would enhance the user-exoskeleton interaction, improving the safety mechanisms of current exoskeletons.MethodsIn order to detect the change in the brain activity when an obstacle suddenly appears, different features of EEG signals are evaluated using the recordings of five healthy subjects. Since the change in the brain activity occurs in the time domain, the features evaluated are: common spatial patterns, average power, slope, and the coefficients of a polynomial fit. A Linear Discriminant Analysis-based classifier is used to differentiate between two conditions: the appearance or not of an obstacle. The evaluation of the performance to detect the obstacles is made in terms of accuracy, true positive (TP) and false positive (FP) rates.ResultsFrom the offline analysis, the best performance is achieved when the slope or the polynomial coefficients are used as features, with average detection accuracy rates of 74.0 and 79.5 %, respectively. These results are consistent with the pseudo-online results, where a complete EEG recording is segmented into windows of 500 ms and overlapped 400 ms, and a decision about the obstacle appearance is made for each window. The results of the best subject were 11 out of 14 obstacles detected with a rate of 9.09 FPs/min, and 10 out of 14 obstacles detected with a rate of 6.34 FPs/min using slope and polynomial coefficients features, respectively.ConclusionsAn EEG-based BCI can be developed to detect the appearance of unexpected obstacles. The average accuracy achieved is 79.5 % of success rate with a low number of false detections. Thus, the online performance of the BCI would be suitable for commanding in a safely way a lower limb exoskeleton during walking.

Comparison of speech parameterization techniques for the classification of speech disfluencies

Stuttering assessment through the manual classification of speech disfluencies is subjective, inconsistent, time-consuming, and prone to error. The aim of this paper is to compare the effectiveness of the 3 speech feature extraction methods, mel-frequency cepstral coefficients, linear predictive coding (LPC)-based cepstral parameters, and perceptual linear predictive (PLP) analysis, for classifying 2 types of speech disfluencies, repetition and prolongation, from recorded disfluent speech samples. Three different classifiers, the k-nearest neighbor classifier, linear discriminant analysis-based classifier, and support vector machine, are employed for the classification of speech disfluencies. Speech samples are taken from the University College London Archive of Stuttered Speech and stuttered events are identified through manual segmentation. A 10-fold cross-validation method is used for testing the reliability of the classifier results. The effect of the 2 parameters (LPC order and frame length) in the LPC- and PLP-based methods on the classification results is also investigated. The experimental results reveal that the proposed method can be used to help speech language pathologists in classifying speech disfluencies.

Application of recurrence quantification analysis to automatically estimate infant sleep states using a single channel of respiratory data

Previous work has identified that non-linear variables calculated from respiratory data vary between sleep states, and that variables derived from the non-linear analytical tool recurrence quantification analysis (RQA) are accurate infant sleep state discriminators. This study aims to apply these discriminators to automatically classify 30 s epochs of infant sleep as REM, non-REM and wake. Polysomnograms were obtained from 25 healthy infants at 2 weeks, 3, 6 and 12 months of age, and manually sleep staged as wake, REM and non-REM. Inter-breath interval data were extracted from the respiratory inductive plethysmograph, and RQA applied to calculate radius, determinism and laminarity. Time-series statistic and spectral analysis variables were also calculated. A nested cross-validation method was used to identify the optimal feature subset, and to train and evaluate a linear discriminant analysis-based classifier. The RQA features radius and laminarity and were reliably selected. Mean agreement was 79.7, 84.9, 84.0 and 79.2 % at 2 weeks, 3, 6 and 12 months, and the classifier performed better than a comparison classifier not including RQA variables. The performance of this sleep-staging tool compares favourably with inter-human agreement rates, and improves upon previous systems using only respiratory data. Applications include diagnostic screening and population-based sleep research.

Automatic detection of a prefrontal cortical response to emotionally rated music using multi-channel near-infrared spectroscopy

Emotional responses can be induced by external sensory stimuli. For severely disabled nonverbal individuals who have no means of communication, the decoding of emotion may offer insight into an individual’s state of mind and his/her response to events taking place in the surrounding environment. Near-infrared spectroscopy (NIRS) provides an opportunity for bed-side monitoring of emotions via measurement of hemodynamic activity in the prefrontal cortex, a brain region known to be involved in emotion processing. In this paper, prefrontal cortex activity of ten able-bodied participants was monitored using NIRS as they listened to 78 music excerpts with different emotional content and a control acoustic stimuli consisting of the Brown noise. The participants rated their emotional state after listening to each excerpt along the dimensions of valence (positive versus negative) and arousal (intense versus neutral). These ratings were used to label the NIRS trial data. Using a linear discriminant analysis-based classifier and a two-dimensional time-domain feature set, trials with positive and negative emotions were discriminated with an average accuracy of 71.94% ± 8.19%. Trials with audible Brown noise representing a neutral response were differentiated from high arousal trials with an average accuracy of 71.93% ± 9.09% using a two-dimensional feature set. In nine out of the ten participants, response to the neutral Brown noise was differentiated from high arousal trials with accuracies exceeding chance level, and positive versus negative emotional differentiation accuracies exceeded the chance level in seven out of the ten participants. These results illustrate that NIRS recordings of the prefrontal cortex during presentation of music with emotional content can be automatically decoded in terms of both valence and arousal encouraging future investigation of NIRS-based emotion detection in individuals with severe disabilities.

Audio Signal Feature Extraction and Classification Using Local Discriminant Bases

Audio feature extraction plays an important role in analyzing and characterizing audio content. Auditory scene analysis, content-based retrieval, indexing, and fingerprinting of audio are few of the applications that require efficient feature extraction. The key to extract strong features that characterize the complex nature of audio signals is to identify their discriminatory subspaces. In this paper, we propose an audio feature extraction and a multigroup classification scheme that focuses on identifying discriminatory time-frequency subspaces using the local discriminant bases (LDB) technique. Two dissimilarity measures were used in the process of selecting the LDB nodes and extracting features from them. The extracted features were then fed to a linear discriminant analysis-based classifier for a three-level hierarchical classification of audio signals into ten classes. In the first level, the audio signals were grouped into artificial and natural sounds. Each of the first level groups were subdivided to form the second level groups viz. instrumental, automobile, human, and nonhuman sounds. The third level was formed by subdividing the four groups of the second level into the final ten groups (drums, flute, piano, aircraft, helicopter, male, female, animals, birds and insects). A database of 213 audio signals were used in this study and an average classification accuracy of 83% for the first level (113 artificial and 100 natural sounds), 92% for the second level (73 instrumental and 40 automobile sounds; 40 human and 60 nonhuman sounds), and 89% for the third level (27 drums, 15 flute, and 31 piano sounds; 23 aircraft and 17 helicopter sounds; 20 male and 20 female speech; 20 animals, 20 birds and 20 insects sounds) were achieved. In addition to the above, a separate classification was also performed combining the LDB features with the mel-frequency cepstral coefficients. The average classification accuracies achieved using the combined features were 91% for the first level, 99% for the second level, and 95% for the third level

Melanoma Metastases in Regional Lymph Nodes Are Accurately Detected by Proton Magnetic Resonance Spectroscopy of Fine-Needle Aspirate Biopsy Samples

Nonsurgical assessment of sentinel nodes (SNs) would offer advantages over surgical SN excision by reducing morbidity and costs. Proton magnetic resonance spectroscopy (MRS) of fine-needle aspirate biopsy (FNAB) specimens identifies melanoma lymph node metastases. This study was undertaken to determine the accuracy of the MRS method and thereby establish a basis for the future development of a nonsurgical technique for assessing SNs. FNAB samples were obtained from 118 biopsy specimens from 77 patients during SN biopsy and regional lymphadenectomy. The specimens were histologically evaluated and correlated with MRS data. Histopathologic analysis established that 56 specimens contained metastatic melanoma and that 62 specimens were benign. A linear discriminant analysis-based classifier was developed for benign tissues and metastases. The presence of metastatic melanoma in lymph nodes was predicted with a sensitivity of 92.9%, a specificity of 90.3%, and an accuracy of 91.5% in a primary data set. In a second data set that used FNAB samples separate from the original tissue samples, melanoma metastases were predicted with a sensitivity of 87.5%, a specificity of 90.3%, and an accuracy of 89.1%, thus supporting the reproducibility of the method. Proton MRS of FNAB samples may provide a robust and accurate diagnosis of metastatic disease in the regional lymph nodes of melanoma patients. These data indicate the potential for SN staging of melanoma without surgical biopsy and histopathological evaluation.