Linear Discriminant Analysis Model Research Articles

Imaging Mass Spectrometry for the Classification of Melanoma Based on BRAF/NRAS Mutational Status

Mutations of the oncogenes v-raf murine sarcoma viral oncogene homolog B1 (BRAF) and neuroblastoma RAS viral oncogene homolog (NRAS) are the most frequent genetic alterations in melanoma and are mutually exclusive. BRAF V600 mutations are predictive for response to the two BRAF inhibitors vemurafenib and dabrafenib and the mitogen-activated protein kinase kinase (MEK) inhibitor trametinib. However, inter- and intra-tumoral heterogeneity and the development of acquired resistance to BRAF inhibitors have important clinical implications. Here, we investigated and compared the molecular profile of BRAF and NRAS mutated and wildtype melanoma patients' tissue samples using imaging mass spectrometry-based proteomic technology, to identify specific molecular signatures associated with the respective tumors. SCiLSLab and R-statistical software were used to classify peptide profiles using linear discriminant analysis and support vector machine models optimized with two internal cross-validation methods (leave-one-out, k-fold). Classification models showed molecular differences between BRAF and NRAS mutated melanoma, and identification of both was possible with an accuracy of 87-89% and 76-79%, depending on the respective classification method applied. In addition, differential expression of some predictive proteins, such as histones or glyceraldehyde-3-phosphate-dehydrogenase, correlated with BRAF or NRAS mutation status. Overall, these findings provide a new molecular method to classify melanoma patients carrying BRAF and NRAS mutations and help provide a broader view of the molecular characteristics of these patients that may help understand the signaling pathways and interactions involving the altered genes.

Visible-Near-Infrared Spectroscopy and Chemometrics for Authentication Detection of Organic Soybean Flour

Organic and non-organic soybean flours, although visually indifferent, have a significant difference in price and nutrition content. Therefore, the accurate authentication detection of organic soybean flour is necessary. Visible-near-infrared (Vis-NIR) spectroscopy coupled with chemometric methods is a non-destructive technique applied to detect authentic or adulterated organic soybean flour. The spectra of organic, adulterated organic, and non-organic soybean flours were captured using a Vis-NIR spectrometer at 350–1000 nm. The spectra were analyzed using partial least squares (PLS), principal component analysis (PCA), and the combination of these two with discriminant analysis (DA). The results showed that PCA using PC1 and PC2 could differentiate organic and non-organic soybean flours, whereas PC1 and PC4 can detect pure and adulterated organic soybean flours. The PCA–linear DA models showed 98.5% accuracy (Acc) for predicting pure organic and adulterated soybean flours and 100% Acc for predicting organic and non-organic flours. Moreover, PLS regression models resulted in a high R² of >95% for predicting organic and non-organic flours and pure and adulterated soybean flours. In addition, the PLS-DA models can differentiate organic from non-organic soybean flour and distinguish pure and adulterated soybean flours with 100% Acc and reliability.

Identifying type 1 / type 2 diabetes in medico-administrative database to improve health surveillance, medical research and prevention in diabetes: Algorithm development and application

IntroductionBig data sources represent an opportunity for diabetes research. One example is the French national health data system (SNDS), gathering information on medical claims of out-of-hospital health care and hospitalizations for the entire French population (66 million). Currently, a validated algorithm based on antidiabetic drug reimbursement is able to identify people with pharmacologically-treated diabetes in the SNDS. But it cannot distinguish type 1 from type 2 diabetes. Differentiating type 1 and type 2 diabetes is crucial in diabetes surveillance, because they carry differences in their prevention, populations at risk, disease natural history, pathophysiology, management and risk of complications.This article investigates the development of a type 1/type 2 diabetes classification algorithm using artificial intelligence and its application to estimate the prevalence of type 1 and type 2 diabetes in France. MethodsThe final data set comprised all diabetes cases from the CONSTANCES cohort (n = 951). A supervised machine learning method based on eight steps was used: final data set selection, target definition (type 1), coding features, final data set splitting into training and testing data sets, feature selection and training and validation and selection of algorithms. The selected algorithm was applied to SNDS data to estimate the type 1 and type 2 diabetes prevalence among adults 18–70 years of age. ResultsAmong the 3481 SNDS features, 14 were selected to train the different algorithms. The final algorithm was a linear discriminant analysis model based on the number of reimbursements for fast-acting insulin, long-acting insulin and biguanides over the previous year (specificity 97% and sensitivity 100%). In 2016, after adjusting for algorithm performance, type 1 and type 2 diabetes prevalence in France was estimated to be 0.3% and 4.4%, respectively. ConclusionOur type 1/type 2 classification algorithm was found to perform well and to be applicable to any prescription or medical claims database from other countries. Artificial intelligence opens new possibilities for research and diabetes prevention.

Chemical characterisation of bitumen type and ageing state based on FTIR spectroscopy and discriminant analysis integrated with variable selection methods

The chemical characterization of bitumen type and ageing state are fundamental in determining structural and mechanical properties of bitumen. This work aims to classify various bitumen types at different ageing states and to identify the primary chemical differences relevant to the classification. Fourier transform infrared (FTIR) spectral data of eight bitumen types at five ageing states were analyzed using a chemometric procedure that incorporates principal component analysis (PCA), linear discriminant analysis (LDA) models, variable selection methods. The models presented results of high accuracy in differentiating bitumen type and ageing state. The results show that the spectral regions that describe the aliphatic and aromatic bonds are critical to the identification of bitumen types. The chemical changes due to bitumen ageing are mainly revealed at the region of 1800 – 900 cm-1. This chemometric method is instructive for the characterization of chemical bitumen properties.

Real-Time Estimation of Support Provision Capability for Poor-Observable Distribution Networks

An indispensable step towards coordinating the actions of distribution and transmission system operators ( <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">dso</small> s- <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">tso</small> ) is to estimate the range of flexibility that can be offered to the <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">tso</small> by <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">dso</small> s. Within this context, a data-driven probabilistic approach is proposed to evaluate the capability of a distribution network in providing active and reactive power support in real-time. To this end, in an offline phase, a linear discriminant analysis model, together with a piecewise linear model of the distribution network are trained to delineate the boundary of a region representing the adherence to distribution network operational constraints. In the implementation phase, this region comprises the feasible set of a series of optimization problems, formulated to determine the support provision capability. These optimization problems are of iterative linear programming type, which allows for real-time applicability. The evaluated support capability can be deemed as the available reserve in real-time transmission operation, which enables providing a coordinated response towards unexpected events, and facilitates the participation of distributed resources in the balancing market by granting an up-to-date estimation of available supports. This approach is tested on the IEEE 123-node system and verified through comparison with an AC optimal power flow technique.

오류 라벨을 포함한 분류모형의 강건성 모의실험연구

In order to increase the accuracy and performance of classification models, the quality of data must be guaranteed. There are various factors that affect the quality of these data, such as measurement accuracy, missing value, and sampling biases. This study focuses on the robustness of the classification model for mislabeled data. The reason we are interested in mislabeled data is that it significantly degrades the performance of the model more than other common measurement errors. In this study, the efficiency of various classification methods for data with mislabels is reviewed. Fisher's linear discriminant analysis model (LDA), quadratic discriminant analysis model (QDA), and logistic regression model were considered in the statistical model-based method, and SVM (support vector machine) and boosting are compared in the algorithm-based method. As a result of simulation study, it is found that the statistical model based approaches show better results in terms of robustness in the data assuming various situations of mislabels.

Improving the non-destructive maturity classification model for durian fruit using near-infrared spectroscopy

The maturity state of durian fruit is a key indicator of quality before trading. This research aims to improve the near-infrared (NIR) model for classifying the maturity stage of durian fruit using a completely non-destructive measurement. Both NIR spectrometers were investigated: the short wavelength NIR (SWNIR) ranging from 450 to 1000 nm and long wavelength NIR (LWNIR) ranging from 860 to 1750 nm. The samples collected for experimentation consisted of four stages: immaturity, prematurity, maturity, and ripe. Each fruit was scanned at the rind position on the main fertile lobe (header, middle, and tail) and stem. The classification models were developed using three supervised machine learning algorithms: linear discriminant analysis (LDA), support vector machine (SVM), and K-Nearest neighbours (KNN). The analysis results revealed that the use of durian rind spectra only obtained between 83.15% and 88.04% accuracy for the LWNIR spectrometer, while the SWNIR spectrometer provided 64.73 to 93.77% accuracy. The performance of model increases when developing with combination between rind and stem spectra. The LDA model developed using a combination of rind and stem spectra provided the greatest efficiency, exhibiting 97.28% and 100% accuracy for LWNIR and SWNIR spectrometers, respectively. The LDA model is therefore recommended for obtaining spectra from smoothing moving average (MA) + baseline of rind position and when used in combination with the MA + standard normal variance (SNV) of stem spectra. The NIR spectroscopy indicated high potential for non-destructive estimation of the durian maturity stage. This process could be used for quality control in the durian export industry to solve the problem of unripe durian being mixed with ripe fruit.

A linear discriminant analysis model of imbalanced associative learning in the mushroom body compartment

To adapt to their environments, animals learn associations between sensory stimuli and unconditioned stimuli. In invertebrates, olfactory associative learning primarily occurs in the mushroom body, which is segregated into separate compartments. Within each compartment, Kenyon cells (KCs) encoding sparse odor representations project onto mushroom body output neurons (MBONs) whose outputs guide behavior. Associated with each compartment is a dopamine neuron (DAN) that modulates plasticity of the KC-MBON synapses within the compartment. Interestingly, DAN-induced plasticity of the KC-MBON synapse is imbalanced in the sense that it only weakens the synapse and is temporally sparse. We propose a normative mechanistic model of the MBON as a linear discriminant analysis (LDA) classifier that predicts the presence of an unconditioned stimulus (class identity) given a KC odor representation (feature vector). Starting from a principled LDA objective function and under the assumption of temporally sparse DAN activity, we derive an online algorithm which maps onto the mushroom body compartment. Our model accounts for the imbalanced learning at the KC-MBON synapse and makes testable predictions that provide clear contrasts with existing models.

Feasibility of a Low-Power, Low-Resolution Laser-Induced Breakdown Spectroscopy Instrument for Analysis of Nickel Alloys: Quantification of the Major Alloying Elements and Classification.

A simple cost-effective laser-induced breakdown spectroscopy (LIBS) instrument was used for quantification of major elements in several nickel alloys and also sorting them. A compact low-power diode-pumped solid-state laser and a miniature low-resolution spectrometer were assembled for the LIBS instrument. Material properties of the nickel alloys depend mainly on the composition of the major elements, Ni, Cr, and Fe, ranging from a few to ∼60 wt%. The emission peaks at 547.7 nm, 520.4 nm, and 438.1 nm for Ni, Cr, and Fe, respectively, were chosen for this analysis. The analytical performance was found to be enough for the quantification of Ni, Cr, and Fe in the nickel alloys. Limits of detection and accuracy were estimated to be a few weight percent (wt%) and measurement precisions were less than 10% in terms of relative standard deviation. The calibration performance of this intensity-based method was compared with that of the "ratio method" which is used in conventional optical emission spectroscopy analyses. The comparison indicates that the intensity-based method is more appropriate with the low-performance LIBS instrument that detects emission peaks of only a few major elements. Also, multivariate modeling of the six different nickel alloy samples based on the emission peak intensities of Ni, Cr, and Fe was performed using k-nearest neighbors (KNN) and linear discriminant analysis (LDA). The KNN and ordinary LDA models showed 95.0% and 98.3% classification correctness for the separate test data set, respectively. To improve classification performance further, the two-step LDA model was trained. In this approach, the two closest sample classes responsible for the decrease in the classification correctness were separately modeled in the second step to exploit their difference effectively. The two-step LDA model showed 100% correctness in classifying the test objects. Our results indicate that such a low-performance LIBS instrument can be effectively utilized for quantitative analysis of the major elements in the nickel alloys and their rapid identification or sorting in combination with an appropriate multivariate modeling algorithm.

Abstract WP212: Association Of Thrombin Generation With Leukocyte Inflammatory Profile In Patients With Acute Ischemic Stroke

Thrombosis is central to the pathogenesis of acute ischemic stroke, with higher thrombin generation being associated with increased stroke risk. The immune system may contribute to thrombin generation in stroke and thus may offer novel strategies for stroke prevention. This study addresses the research question regarding the relationship of thrombin generation to leukocyte gene expression in patients with acute ischemic stroke. We isolated RNA from whole blood and examined the relationship to thrombin generation capacity in patients with acute ischemic stroke. Due to its effects on thrombin generation, patients on anticoagulants were excluded from the study. The relationship of gene expression with peak thrombin was evaluated by analysis of covariance across peak thrombin quartiles adjusted for sex and age. In 97 patients with acute ischemic stroke, peak thrombin was variable, ranging from 252.0 to 752.4 nM. Increased thrombin generation was associated with differences in thromboinflammatory leukocyte gene expression, including a decrease in ADAM metallopeptidase with thrombospondin type 1 motif 13 and an increase in nuclear factor B (NF-kB)-activating protein, protein disulfde isomerase family A member 5, and tissue factor pathway inhibitor 2. Pathways associated with peak thrombin included interleukin 6 signaling, thrombin signaling, and NF-kB signaling. A linear discriminant analysis model summarizing the immune activation associated with peak thrombin in a first cohort of stroke could distinguish patients with low peak thrombin from high peak thrombin in a second cohort of 112 patients with acute ischemic stroke.The identified genes and pathways support a role of the immune system contributing to thrombus formation in patients with stroke. These may have relevance to antithrombotic strategies for stroke prevention.

Metabolic Reprogramming in Colon Cancer Cells Persistently Infected with Newcastle Disease Virus.

Newcastle disease virus (NDV) is an oncolytic agent against various types of mammalian cancers. As with all cancer therapies, the development of cancer resistance, both innate and acquired, is becoming a challenge. In this study, we investigated persistently NDV-infected Caco-2 colon cancer cells, designated as virus-resistant (VR) Caco-2 cells, which were then able to resist NDV-mediated oncolysis. We applied single-cell Raman spectroscopy, combined with deuterium isotope probing (Raman-DIP) techniques, to investigate the metabolic adaptations and dynamics in VR Caco-2 cells. A linear discriminant analysis (LDA) model demonstrated excellent performance in differentiating VR Caco-2 from Caco-2 cells at single-cell level. By comparing the metabolic profiles in a time-resolved manner, the de novo synthesis of proteins and lipids was found upregulated, along with decreased DNA synthesis in VR Caco-2. The results suggest that VR Caco-2 cells might reprogram their metabolism and divert energy from proliferation to protein synthesis and lipidic modulation. The ability to identify and characterise single resistant cells among a population of cancer cells would help develop a deeper understanding of the resistance mechanisms and better tactics for developing effective cancer treatment.

Leaf-based species classification of hybrid cherry tomato plants by using hyperspectral imaging

Approaches based on near infrared hyperspectral imaging (NIR-HSI) technology combined with machine learning have been developed to classify the leaves of hybrid cherry tomatoes and then identify the species of hybrid cherry tomato plants. The near infrared (NIR) hyperspectral images of 400 cherry tomato leaves (100 per species) were collected in the wavelength range of 900–1700 nm. Machine learning algorithms such as linear discriminant analysis (LDA), random forest (RF), and support vector machine (SVM) were employed to construct leaf classification models with the hyperspectral data preprocessed by Savitzky-Golay (SG) smoothing filter, first derivative (first Der) and standard normal variate (SNV). Principle of Component Analysis (PCA) was also used to reduce the data dimension and extract spectral features. It is revealed that the LDA model reaches the highest classification accuracy among the three machine learning algorithms and SNV can lead to higher improvement in model accuracy than other preprocessing methods of SG smoothing and first Der. Analysis based on PCA spectral feature extraction demonstrates that differences occur in internal material content in the leaves of cherry tomato plants with different species, which renders the models being able to distinguish between the species. Another important work was performed to reveal the different effects of the mesophyll and vein regions (VR) on the accuracy of the leaf classification model. It is demonstrated that the classification accuracy is improved by a value of 0.033 or 0.042 when mesophyll substitutes vein or whole leaf as regions of interest (ROI) to extract reflectance spectra for modeling. As a result, the accuracy of the training and test set respectively reached a high value of 0.998 and 0.973 for the LDA classification model combined with the SNV preprocessing method. The results propose that the use of mesophyll region (MR) as ROI can improve the performance of the leaf classification model, which provides a new strategy for efficient and non-destructive classification of different hybrid cherry tomato plants.

Detection of Low-Level Adulteration of Hungarian Honey Using near Infrared Spectroscopy

Honey adulteration is a worldwide problem; however, its detection is a challenge for researchers and authorities. There are numerous ways of honey counterfeiting; amongst them, direct adulteration is one of the most common methods. Correlative techniques, such as near-infrared spectroscopy (NIRS), are useful tools in the detection of honey adulteration; however, this method has not been applied to Hungarian honeys. The aim of this research was to investigate the performance of NIRS for the detection of sugar syrup addition to Hungarian honeys at lower concentration levels (&lt;10% w/w). Acacia, rape, forest, sunflower, and linden honeys were mixed with high-fructose-content sugar syrup, rice syrup, or self-made glucose fructose syrup in 3%, 5%, and 10% w/w. NIRS analysis was performed in the spectral range of 950–1650 nm. Principal component analysis was coupled with linear discriminant analysis and partial least square regression models were built for the classification and prediction of adulteration levels, respectively. Our results showed that the performance of NIRS highly depends on both type of syrup and honey. PCA-LDA models provided the 100% correct classification of control in the case of all the models, while PLSR results could predict the added sugar syrup content in the case of rice and F40 syrup models, obtaining &gt;2.2 RPDCV value.

PLDA in i-vector based underwater acoustic signals classification

ABSTRACT Intelligent recognition of underwater acoustic targets is crucial for exploiting marine resources. In this paper, we propose a deep learning method for underwater acoustic signal recognition consisting of three steps: Firstly, we extract the underwater acoustic signal's identification vector (i-vector) using a Gaussian mixture model. Secondly, we construct a probabilistic linear discriminant analysis (PLDA) model that divides the i-vector into two segments, Finally, we used a multi-long short-term memory neural network (MLSTMNN) to learn the features, outputting the underwater acoustic signal labels through the softmax layer. Experiments using the ShipsEar dataset can obtain a recognition rate of 84.8%; compared to the traditional Mel frequency Cepstral Coefficients (MFCC) and convolutional neural networks, the recognition rate greatly improved.

Discriminant Analysis of Water Inrush Sources in the Weibei Coalfield, Shaanxi Province, China

Water inrush disasters in mining areas are one of the most serious geological disasters in coal mining. The purpose of this study is to study the establishment of a water chemical database and water inrush source discrimination model in the Weibei coalfield to provide the basis for regional hydrogeological conditions for future mining under pressure in the Weibei area, as well as a basis for the rapid identification of water inrush sources in the Weibei coalfield. In this paper, a conventional hydrochemical and trace element discrimination model for mine water inrush was established, and the hydrochemical characteristic files of the entire mining area were integrated. Based on 10 indicators, three hydrochemical discrimination models of rock stratum aquifers were established. Through the Mahalanobis distance test, it was found that the six selected variables, K+ + Na+, Mg2+, NH4+, Cl−, SO42−, and pH, have significant discrimination ability and good effect and can effectively distinguish the three main water inrush aquifers in the Weibei mining area. Then, the clustering stepwise discriminant analysis method was used to select 24 water samples and 14 trace element indicators from the conventional water chemistry test results. Based on principal component analysis, a principal component analysis discriminant model of trace elements was established for the four main aquifers. The accuracy and misjudgment rate of the Bayes multi-class linear discriminant using conventional ions as explanatory variables were 64.3% and 35.7%, respectively, showing a poor discriminant effect. On this basis, seven characteristic trace elements were analyzed according to Bayes multi-class linear discriminant analysis, the mutual influence and restriction relationship regarding the migration of these seven trace elements in the groundwater system of the mining area was determined, and the modified Bayes multi-class linear discriminant analysis model of trace elements for the water inrush source was established, which was more accurate than the conventional ion Bayes multi-class linear discriminant analysis model. The accuracy rate reached 92.9%. This research is of great significance for mine water-source identification and water-inrush prevention guidance.

Classification Performance of Machine Learning Methods for Identifying Resistance, Resilience, and Susceptibility to Haemonchus contortus Infections in Sheep.

This study investigated the feasibility of using easy-to-measure phenotypic traits to predict sheep resistant, resilient, and susceptible to gastrointestinal nematodes, compared the classification performance of multinomial logistic regression (MLR), linear discriminant analysis (LDA), random forest (RF), and artificial neural network (ANN) methods, and evaluated the applicability of the best classification model on each farm. The database comprised 3654 records of 1250 Santa Inês sheep from 6 farms. The animals were classified into resistant (2605 records), resilient (939 records), and susceptible (110 records) according to fecal egg count and packed cell volume. A random oversampling method was performed to balance the dataset. The classification methods were fitted using the information of age class, the month of record, farm, sex, Famacha© degree, body weight, and body condition score as predictors, and the resistance, resilience, and susceptibility to gastrointestinal nematodes as the target classes to be predicted considering data from all farms randomly. An additional leave-one-farm-out cross-validation technique was used to assess prediction quality across farms. The MLR and LDA models presented good performances in predicting susceptible and resistant animals. The results suggest that the use of readily available records and easily measurable traits may provide useful information for supporting management decisions at the farm level.

Deep learning models for building window-openings detection in heating season

The increasing use of monitoring systems such as Building Management System (BMS) or connected devices bring the opportunity to better evaluate, model or control both occupants' comfort and energy consumed by an operated building thanks to the consequent amount of data provided (e.g., air temperature, CO2 concentration, electricity consumption). Occupants' behavior and more specifically window-openings affect both occupants’ thermal comfort and building energy consumption and are therefore key components to consider. This paper presents a comparison of machine learning models applied on window-openings detection during the heating season such as: Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), Random Forest Classifier (RFC) and two Recurrent Neural Network (RNN), namely, Long Short Term Memory (LSTM) and Gated Recurrent Unit (GRU). While some applications of Artificial Intelligence (AI) methods applied on window-openings detection exist in the literature, this study proposes a detailed comparison of the main methods and focuses on the impact of feature engineering process considering four different data transformations based on field expertise and more than 800 different combinations built on six indoor and outdoor measurements. Results show that some of the proposed transformations and combinations positively impact all models performances. The best performances on window-openings detection are attained by using indoor temperature and CO2 concentration on RNN models with an average F1-score of 0.78 while LDA, SVM and RFC models tend to provide satisfying but lower performance around 0.70–72. In addition, by using the right transformation, significant results can be achieved by detecting up to 84–88% of window-opening times with the sole use of indoor air temperature measurements.

Serum and Urine Biomarkers Related to Kidney Fibrosis Predict Kidney Outcome in Czech Patients with IgA Nephropathy.

We evaluated biomarkers related to kidney fibrosis for the outcome of patients with IgA nephropathy (IgAN). Clinical parameters (estimated glomerular filtration rate, hypertension, proteinuria) and histological findings were assessed in 134 patients with IgAN at the time of diagnosis and followed up prospectively (mean follow-up time, 56.5 months). We measured biomarkers of collagen and laminin turnover in serum and urine collected at the time of kidney biopsy using a novel enzyme-linked immunosorbent assay. Linear discriminant analysis and logistic regression models were used to predict the patient's kidney outcome. Five serum and urine biomarkers of laminin and collagen turnover (sLG1M, sPRO-C3, sPRO-C6, uPRO-C6/Cr, uC3M/Cr) could significantly differentiae IgAN patients with a worse prognosis. Clinical parameters (glomerular filtration rate (GFR), proteinuria) distinguished patients at risk of IgAN progression with a specificity of 87.3% and a sensitivity of 45.2% (area under the curve-AUC 0.751). The addition of the biomarkers significantly increased the prognostic ability with a specificity of 85.1% and a sensitivity of 73.3% (AUC 0.905). We have identified three serum (sLG1M, sPRO-C3, sPRO-C6) and two urinary markers (uPRO-C6/Cr, u-C3M /Cr) that significantly improve the prognostic ability of markers of kidney function to identify an IgAN patient's risk of progressing to ESKD.

Production and perception of prevelar merger: Two-dimensional comparisons using Pillai scores and confusion matrices

Vowel merger production is quantified with gradient acoustic measures, while phonemic perception methods are often coarser, complicating comparisons within mergers in progress. This study implements a perception experiment in two-dimensional formant space (F1 × F2), allowing unified plotting, quantification, and statistics with production data. Production and perception are compared within 20 speakers for a two-part prevelar merger in progress in Pacific Northwest English, where mid-front /ɛ, e/ approximate or merge before voiced velar /ɡ/ (leg–vague merger), and low-front prevelar /æɡ/ raises toward them (bag-raising). Distributions are visualized with kernel density plots and overlap quantified with Pillai scores and confusion matrices from linear discriminant analysis models. Results suggest that leg–vague merger is perceived as more complete than it is produced (in both the sample and community), while bag-raising is highly variable in production but rejected in perception. Relationships between production and perception varied by age, with raising and merger progressing across two generations in production but not perception, followed by younger adults perceiving leg–vague merger but not producing it and varying in (minimal) raising perception while varying in bag-raising in production. Thus, prevelar raising/merger may be progressing among some social groups but reversing in others.

Appearance segmentation and documentation applied to cultural heritage surfaces

This paper describes the development and application of a novel supervised segmentation technique used for conservation documentation based on visible appearance changes of Cultural Heritage (CH) metal surfaces. The technique is based on employing a linear discriminant analysis model to classify Reflectance Transformation Imaging (RTI) reconstruction coefficients. The Hemispherical Harmonics (HSH) reconstruction coefficients for each pixel are first calculated and then normalized. This normalization increases the robustness and invariance of the application making it possible to apply it for documenting different surfaces and at different time intervals. In this paper, we presented three case studies related to corrosion assessment of CH objects through detection of corrosion and monitoring the degree of silver tarnishing. For each case study, a supervised data set is constructed, teaching the algorithm to recognize as distinct a specified appearance characteristic (such as corrosion, metal etc.) by comparing it to the reconstruction coefficients of each pixel. The segmented information is visualized by a simplified colormap. The calculated results are afterwards verified by visible inspection from conservation-restoration experts. The method can segment surfaces with changes in micro-geometry, but it reaches its limitation on surfaces with minimal topography and high specularity.