Uncertainty estimation and misclassification probability for classification models based on discriminant analysis and support vector machines

Distinguishing brain tumors by Label-free confocal micro-Raman spectroscopy

BACKGROUND

Brain tumors have serious adverse effects on public health and social economy. Accurate detection of brain tumor types is critical for effective and proactive treatment, and thus improve the survival of patients.

METHODS

Four types of brain tumor tissue sections were detected by Raman spectroscopy. Principal component analysis (PCA) has been used to reduce the dimensionality of the Raman spectra data. Linear discriminant analysis (LDA) and quadratic discriminant analysis (QDA) methods were utilized to discriminate different types of brain tumors.

RESULTS

Raman spectra were collected from 40 brain tumors. Variations in intensity and shift were observed in the Raman spectra positioned at 721, 854, 1004, 1032, 1128, 1248, 1449 cm-1 for different brain tumor tissues. The PCA results indicated that glioma, pituitary adenoma, and meningioma are difficult to differentiate from each other, whereas acoustic neuroma is clearly distinguished from the other three tumors. Multivariate analysis including QDA and LDA methods showed the classification accuracy rate of the QDA model was 99.47 %, better than the rate of LDA model was 95.07 %.

CONCLUSIONS

Raman spectroscopy could be used to extract valuable fingerprint-type molecular and chemical information of biological samples. The demonstrated technique has the potential to be developed to a rapid, label-free, and intelligent approach to distinguish brain tumor types with high accuracy.

Point-of-Care Disease Screening in Primary Care Using Saliva: A Biospectroscopy Approach for Lung Cancer and Prostate Cancer

Saliva is a largely unexplored liquid biopsy that can be readily obtained noninvasively. Not dissimilar to blood plasma or serum, it contains a vast array of bioconstituents that may be associated with the absence or presence of a disease condition. Given its ease of access, the use of saliva is potentially ideal in a point-of-care screening or diagnostic test. Herein, we developed a swab "dip" test in saliva obtained from consenting patients participating in a lung cancer-screening programme being undertaken in north-west England. A total of 998 saliva samples (31 designated as lung-cancer positive and 17 as prostate-cancer positive) were taken in the order in which they entered the clinic (i.e., there was no selection of participants) during the course of this prospective screening programme. Samples (sterile Copan blue rayon swabs dipped in saliva) were analysed using attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectroscopy. In addition to unsupervised classification on resultant infrared (IR) spectra using principal component analysis (PCA), a range of feature selection/extraction algorithms were tested. Following preprocessing, the data were split between training (70% of samples, 22 lung-cancer positive versus 664 other) and test (30% of samples, 9 lung-cancer positive versus 284 other) sets. The training set was used for model construction and the test set was used for validation. The best model was the PCA-quadratic discriminant analysis (QDA) algorithm. This PCA-QDA model was built using 8 PCs (90.4% of explained variance) and resulted in 93% accuracy for training and 91% for testing, with clinical sensitivity at 100% and specificity at 91%. Additionally, for prostate cancer patients amongst the male cohort (n = 585), following preprocessing, the data were split between training (70% of samples, 12 prostate-cancer positive versus 399 other) and test (30% of samples, 5 prostate-cancer positive versus 171 other) sets. A PCA-QDA model, again the best model, was built using 5 PCs (84.2% of explained variance) and resulted in 97% accuracy for training and 93% for testing, with clinical sensitivity at 100% and specificity at 92%. These results point to a powerful new approach towards the capability to screen large cohorts of individuals in primary care settings for underlying malignant disease.

Attenuated Total Reflection Fourier-Transform Infrared Spectral Discrimination in Human Tissue of Oesophageal Transformation to Adenocarcinoma

This study presents ATR-FTIR (attenuated total reflectance Fourier-transform infrared) spectral analysis of ex vivo oesophageal tissue including all classifications to oesophageal adenocarcinoma (OAC). The article adds further validation to previous human tissue studies identifying the potential for ATR-FTIR spectroscopy in differentiating among all classes of oesophageal transformation to OAC. Tissue spectral analysis used principal component analysis quadratic discriminant analysis (PCA-QDA), successive projection algorithm quadratic discriminant analysis (SPA-QDA), and genetic algorithm quadratic discriminant analysis (GA-QDA) algorithms for variable selection and classification. The variables selected by SPA-QDA and GA-QDA discriminated tissue samples from Barrett's oesophagus (BO) to OAC with 100% accuracy on the basis of unique spectral "fingerprints" of their biochemical composition. Accuracy test results including sensitivity and specificity were determined. The best results were obtained with PCA-QDA, where tissues ranging from normal to OAC were correctly classified with 90.9% overall accuracy (71.4-100% sensitivity and 89.5-100% specificity), including the discrimination between normal and inflammatory tissue, which failed in SPA-QDA and GA-QDA. All the models revealed excellent results for distinguishing among BO, low-grade dysplasia (LGD), high-grade dysplasia (HGD), and OAC tissues (100% sensitivities and specificities). This study highlights the need for further work identifying potential biochemical markers using ATR-FTIR in tissue that could be utilised as an adjunct to histopathological diagnosis for early detection of neoplastic changes in susceptible epithelium.

Intelligent Evaluation of Stone Cell Content of Korla Fragrant Pears by Vis/NIR Reflection Spectroscopy

Stone cells are a distinctive characteristic of pears and their formation negatively affects the quality of the fruit. To evaluate the stone cell content (SCC) of Korla fragrant pears, we developed a Vis/NIR spectroscopy system that allowed for the adjustment of the illuminating angle. The successive projective algorithm (SPA) and the Monte Carlo uninformative variable elimination (MCUVE) based on the sampling algorithm were used to select characteristic wavelengths. The particle swarm optimization (PSO) algorithm was used to optimize the combination of penalty factor C and kernel function parameter g. Support vector regression (SVR) was used to construct the evaluation model of the SCC. The SCC of the calibration set ranged from 0.240% to 0.657% and that of the validation set ranged from 0.315% to 0.652%. The SPA and MCUVE were used to optimize 57 and 83 characteristic wavelengths, respectively. The combinations of C and g were (6.2561, 0.2643) and (2.5133, 0.1128), respectively, when different characteristic wavelengths were used as inputs of SVR, indicating that the first combination had good generalization ability. The correlation coefficients of the SPA-SVR model after pre-processing the standardized normal variate (SNV) for both sets were 0.966 and 0.951, respectively. These results show that the SNV-SPA-SVR model satisfied the requirements of intelligent evaluation of SCC in Korla fragrant pears.

Recent Advances in Classification of Brain Tumor from MR Images - State of the Art Review from 2017 to 2021

BACKGROUND

The task of identifying a tumor in the brain is a complex problem that requires sophisticated skills and inference mechanisms to accurately locate the tumor region. The complex nature of the brain tissue makes the problem of locating, segmenting, and ultimately classifying Magnetic Resonance (MR) images a complex problem. The aim of this review paper is to consolidate the details of the most relevant and recent approaches proposed in this domain for the binary and multi-class classification of brain tumors using brain MR images.

OBJECTIVE

In this review paper, a detailed summary of the latest techniques used for brain MR image feature extraction and classification is presented. A lot of research papers have been published recently with various techniques proposed for identifying an efficient method for the correct recognition and diagnosis of brain MR images. The review paper allows researchers in the field to familiarize themselves with the latest developments and be able to propose novel techniques that have not yet been explored in this research domain. In addition, the review paper will facilitate researchers, who are new to machine learning algorithms for brain tumor recognition, to understand the basics of the field and pave the way for them to be able to contribute to this vital field of medical research.

RESULTS

In this paper, the review is performed for all recently proposed methods for both feature extraction and classification. It also identifies the combination of feature extraction methods and classification methods that when combined would be the most efficient technique for the recognition and diagnosis of brain tumor from MR images. In addition, the paper presents the performance metrics particularly the recognition accuracy, of selected research published between 2017- 2021.

Scores selection via Fisher's discriminant power in PCA-LDA to improve the classification of food data

This paper proposes an adaptation of the Fisher's discriminability criterion (named here as discriminant power, DP) for choosing principal components (obtained from Principal Component Analysis, PCA), which will be used to construct supervised Linear Discriminant Analysis (LDA) models for solving classification problems of food data. The proposed PCA-DP-LDA algorithm was then applied to (i) simulated data, (ii) classify soybean oils with respect to expiration date, and (iii) identify cachaça adulteration with wood extracts that simulated aging. For comparison, PCA-DP-LDA was evaluated against conventional PCA-LDA (based on explained variance) and Partial Least Squares-Discriminant Analysis (PLS-DA). Among them, PCA-DP-LDA achieved the most parsimonious and interpretable results, with similar or better classification performance. Therefore, the new algorithm can be considered a good alternative to the already well-established discriminant methods, being potentially applied where the discriminability of the principal components may not follow the same behavior of the explained variance.

A comparative analysis of different biofluids towards ovarian cancer diagnosis using Raman microspectroscopy

Biofluids, such as blood plasma or serum, are currently being evaluated for cancer detection using vibrational spectroscopy. These fluids contain information of key biomolecules, such as proteins, lipids, carbohydrates and nucleic acids, that comprise spectrochemical patterns to differentiate samples. Raman is a water-free and practically non-destructive vibrational spectroscopy technique, capable of recording spectrochemical fingerprints of biofluids with minimum or no sample preparation. Herein, we compare the performance of these two common biofluids (blood plasma and serum) together with ascitic fluid, towards ovarian cancer detection using Raman microspectroscopy. Samples from thirty-eight patients were analysed (n = 18 ovarian cancer patients, n = 20 benign controls) through different spectral pre-processing and discriminant analysis techniques. Ascitic fluid provided the best class separation in both unsupervised and supervised discrimination approaches, where classification accuracies, sensitivities and specificities above 80% were obtained, in comparison to 60–73% with plasma or serum. Ascitic fluid appears to be rich in collagen information responsible for distinguishing ovarian cancer samples, where collagen-signalling bands at 1004 cm⁻¹ (phenylalanine), 1334 cm⁻¹ (CH₃CH₂ wagging vibration), 1448 cm⁻¹ (CH₂ deformation) and 1657 cm⁻¹ (Amide I) exhibited high statistical significance for class differentiation (P < 0.001). The efficacy of vibrational spectroscopy, in particular Raman spectroscopy, combined with ascitic fluid analysis, suggests a potential diagnostic method for ovarian cancer.

Spectrochemical differentiation in gestational diabetes mellitus based on attenuated total reflection Fourier-transform infrared (ATR-FTIR) spectroscopy and multivariate analysis

Gestational diabetes mellitus (GDM) is a hyperglycaemic imbalance first recognized during pregnancy, and affects up to 22% of pregnancies worldwide, bringing negative maternal–fetal consequences in the short- and long-term. In order to better characterize GDM in pregnant women, 100 blood plasma samples (50 GDM and 50 healthy pregnant control group) were submitted Attenuated Total Reflection Fourier-transform infrared (ATR-FTIR) spectroscopy, using chemometric approaches, including feature selection algorithms associated with discriminant analysis, such as Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA) and Support Vector Machines (SVM), analyzed in the biofingerprint region between 1800 and 900 cm⁻¹ followed by Savitzky–Golay smoothing, baseline correction and normalization to Amide-I band (~ 1650 cm⁻¹). An initial exploratory analysis of the data by Principal Component Analysis (PCA) showed a separation tendency between the two groups, which were then classified by supervised algorithms. Overall, the results obtained by Genetic Algorithm Linear Discriminant Analysis (GA-LDA) were the most satisfactory, with an accuracy, sensitivity and specificity of 100%. The spectral features responsible for group differentiation were attributed mainly to the lipid/protein regions (1462–1747 cm⁻¹). These findings demonstrate, for the first time, the potential of ATR-FTIR spectroscopy combined with multivariate analysis as a screening tool for fast and low-cost GDM detection.

Identification of resistance in Escherichia coli and Klebsiella pneumoniae using excitation-emission matrix fluorescence spectroscopy and multivariate analysis

Klebsiella pneumoniae and Escherichia coli are part of the Enterobacteriaceae family, being common sources of community and hospital infections and having high antimicrobial resistance. This resistance profile has become the main problem of public health infections. Determining whether a bacterium has resistance is critical to the correct treatment of the patient. Currently the method for determination of bacterial resistance used in laboratory routine is the antibiogram, whose time to obtain the results can vary from 1 to 3 days. An alternative method to perform this determination faster is excitation-emission matrix (EEM) fluorescence spectroscopy combined with multivariate classification methods. In this paper, Linear Discriminant Analysis (LDA), Quadratic Discriminant Analysis (QDA) and Support Vector Machines (SVM), coupled with dimensionality reduction and variable selection algorithms: Principal Component Analysis (PCA), Genetic Algorithm (GA), and the Successive Projections Algorithm (SPA) were used. The most satisfactory models achieved sensitivity and specificity rates of 100% for all classes, both for E. coli and for K. pneumoniae. This finding demonstrates that the proposed methodology has promising potential in routine analyzes, streamlining the results and increasing the chances of treatment efficiency.

Spectrochemical analysis of liquid biopsy harnessed to multivariate analysis towards breast cancer screening

Mortality due to breast cancer could be reduced via screening programs where preliminary clinical tests employed in an asymptomatic well-population with the objective of identifying cancer biomarkers could allow earlier referral of women with altered results for deeper clinical analysis and treatment. The introduction of well-population screening using new and less-invasive technologies as a strategy for earlier detection of breast cancer is thus highly desirable. Herein, spectrochemical analyses harnessed to multivariate classification techniques are used as a bio-analytical tool for a Breast Cancer Screening Program using liquid biopsy in the form of blood plasma samples collected from 476 patients recruited over a 2-year period. This methodology is based on acquiring and analysing the spectrochemical fingerprint of plasma samples by attenuated total reflection Fourier-transform infrared spectroscopy; derived spectra reflect intrinsic biochemical composition, generating information on nucleic acids, carbohydrates, lipids and proteins. Excellent results in terms of sensitivity (94%) and specificity (91%) were obtained using this method in comparison with traditional mammography (88-93% and 85-94%, respectively). Additional advantages such as better disease prognosis thus allowing a more effective treatment, lower associated morbidity, fewer false-positive and false-negative results, lower-cost, and higher analytical frequency make this method attractive for translation to the clinical setting.

Tutorial: multivariate classification for vibrational spectroscopy in biological samples

Vibrational spectroscopy techniques, such as Fourier-transform infrared (FTIR) and Raman spectroscopy, have been successful methods for studying the interaction of light with biological materials and facilitating novel cell biology analysis. Spectrochemical analysis is very attractive in disease screening and diagnosis, microbiological studies and forensic and environmental investigations because of its low cost, minimal sample preparation, non-destructive nature and substantially accurate results. However, there is now an urgent need for multivariate classification protocols allowing one to analyze biologically derived spectrochemical data to obtain accurate and reliable results. Multivariate classification comprises discriminant analysis and class-modeling techniques where multiple spectral variables are analyzed in conjunction to distinguish and assign unknown samples to pre-defined groups. The requirement for such protocols is demonstrated by the fact that applications of deep-learning algorithms of complex datasets are being increasingly recognized as critical for extracting important information and visualizing it in a readily interpretable form. Hereby, we have provided a tutorial for multivariate classification analysis of vibrational spectroscopy data (FTIR, Raman and near-IR) highlighting a series of critical steps, such as preprocessing, data selection, feature extraction, classification and model validation. This is an essential aspect toward the construction of a practical spectrochemical analysis model for biological analysis in real-world applications, where fast, accurate and reliable classification models are fundamental.

Machine Learning Applications for Mass Spectrometry-Based Metabolomics

The metabolome of an organism depends on environmental factors and intracellular regulation and provides information about the physiological conditions. Metabolomics helps to understand disease progression in clinical settings or estimate metabolite overproduction for metabolic engineering. The most popular analytical metabolomics platform is mass spectrometry (MS). However, MS metabolome data analysis is complicated, since metabolites interact nonlinearly, and the data structures themselves are complex. Machine learning methods have become immensely popular for statistical analysis due to the inherent nonlinear data representation and the ability to process large and heterogeneous data rapidly. In this review, we address recent developments in using machine learning for processing MS spectra and show how machine learning generates new biological insights. In particular, supervised machine learning has great potential in metabolomics research because of the ability to supply quantitative predictions. We review here commonly used tools, such as random forest, support vector machines, artificial neural networks, and genetic algorithms. During processing steps, the supervised machine learning methods help peak picking, normalization, and missing data imputation. For knowledge-driven analysis, machine learning contributes to biomarker detection, classification and regression, biochemical pathway identification, and carbon flux determination. Of important relevance is the combination of different omics data to identify the contributions of the various regulatory levels. Our overview of the recent publications also highlights that data quality determines analysis quality, but also adds to the challenge of choosing the right model for the data. Machine learning methods applied to MS-based metabolomics ease data analysis and can support clinical decisions, guide metabolic engineering, and stimulate fundamental biological discoveries.

Distributed Systematic Grid-Connected Inverter Using IGBT Junction Temperature Predictive Control Method: An Optimization Approach

Distributed systematic grid-connected inverter practice needs to improve insulated gate bipolar transistor (IGBT) stability to ensure the safe operation. This study is to ensure the safety and reliability operation of the IGBT module in symmetry to meet the reliable and stable distributed systematic grid-connected inverter practice and the junction temperature is a parameter to assess its operating state. It is difficult to accurately acquire the IGBT junction temperature to be solved by a single method of combining the test and the modeling. The saturation voltage drop or collector current and module junction temperature data under different power cycles are measured by the power cycle test and the single pulse test. The improved chicken swarm optimization increases the chickens diversity and self-learning ability. The prediction model of the improved chicken swarm optimization-support vector machine is proposed to forecast the module junction temperature. The result showed to compare with the particle swarm optimization-support vector machine model and chicken swarm optimization-support vector machine model and showed the coincidence degree between the proposed model prediction value and the true value is higher. The mean absolute error ratio indicates the proposed model has a smaller error and a better prediction performance. The proposed model has a positive impact on improving the distributed systematic grid-connected inverter industrial development and promotes the new energy usage.

A quantitative uncertainty metric controls error in neural network-driven chemical discovery

Machine learning (ML) models, such as artificial neural networks, have emerged as a complement to high-throughput screening, enabling characterization of new compounds in seconds instead of hours. The promise of ML models to enable large-scale chemical space exploration can only be realized if it is straightforward to identify when molecules and materials are outside the model's domain of applicability. Established uncertainty metrics for neural network models are either costly to obtain (e.g., ensemble models) or rely on feature engineering (e.g., feature space distances), and each has limitations in estimating prediction errors for chemical space exploration. We introduce the distance to available data in the latent space of a neural network ML model as a low-cost, quantitative uncertainty metric that works for both inorganic and organic chemistry. The calibrated performance of this approach exceeds widely used uncertainty metrics and is readily applied to models of increasing complexity at no additional cost. Tightening latent distance cutoffs systematically drives down predicted model errors below training errors, thus enabling predictive error control in chemical discovery or identification of useful data points for active learning.