PCA-LDA Method Research Articles

High-Precision Intelligent Cancer Diagnosis Method: 2D Raman Figures Combined with Deep Learning.

Raman spectroscopy, as a label-free detection technology, has been widely used in tumor diagnosis. However, most tumor diagnosis procedures utilize multivariate statistical analysis methods for classification, which poses a major bottleneck toward achieving high accuracy. Here, we propose a concept called the two-dimensional (2D) Raman figure combined with convolutional neural network (CNN) to improve the accuracy. Two-dimensional Raman figures can be obtained from four transformation methods: spectral recurrence plot (SRP), spectral Gramian angular field (SGAF), spectral short-time Fourier transform (SSTFT), and spectral Markov transition field (SMTF). Two-dimensional CNN models all yield more than 95% accuracy, which is higher than the PCA-LDA method and the Raman-spectrum-CNN method, indicating that 2D Raman figure inputs combined with CNN may be one reason for gaining excellent performances. Among 2D-CNN models, the main difference is the conversion, where SRP is based on the structure of wavenumber series with the best performances (98.9% accuracy, 99.5% sensitivity, 98.3% specificity), followed by SGAF on the wavenumber series, SSTFT on wavenumber and intensity information, and SMTF on wavenumber position information. The inclusion of external information in the conversion may be another reason for improvement in the accuracy. The excellent capability shows huge potential for tumor diagnosis via 2D Raman figures and may be applied in other spectroscopy analytical fields.

Serum log-transformed Raman spectroscopy combined with multivariate analysis for the detection of echinococcosis

In this study, the Raman spectra of serum samples from patients with echinococcosis and healthy controls were recorded by a portable spectrometer under 532 nm excitation. Due to the resonance Raman effect, the two Raman peaks (at 1154 and 1515 cm-1) belonging to β carotene were significantly enhanced, resulting in relatively weak signal values for other peaks. In order to amplify the weak Raman peaks and improve the capability of Raman spectroscopy for the differentiation of healthy and echinococcosis infected serum samples, a logarithmic transformation preprocessing method was used. Furthermore, principal component analysis (PCA), combined with linear discriminant analysis (LDA), were used to distinguish echinococcosis patients from healthy volunteers. The results show that the diagnostic models formed by the log-transformed spectral data sets perform better in the balance of sensitivity and specificity. The accuracy, sensitivity, and specificity of the diagnostic model based on the spectral data set of optimal scale transformation were 98.00%, 98.10%, and 98.00%, respectively, which were slightly superior to the untransformed spectral data set. Our findings suggest that serum log-transformed Raman spectroscopy combined with the PCA-LDA method has great potential for improving the detection of echinococcosis.

Evaluation of the Capability of Horizontal ATR-FTMIR and UV-Visible Spectroscopy in the Discrimination of Virgin Olive Oils from the Moroccan Region of Beni Mellal-Khenifra

One of the most important challenges in the authentication of olive oil is the determination of the geographical origin of virgin olive oil. In this work, we evaluated the capacity of two spectroscopic techniques, UV-Visible and ATR-FTMIR, coupled with chemometric tools to determine the geographical origin of olive oils. These analytical approaches have been applied to samples that have been collected during the period of olive oil production, in the Moroccan region of Beni Mellal-Khenifra. To develop a rapid analysis tool capable of authenticating the geographical origin of virgin olive oils from five geographical areas of the Moroccan region of Beni Mellal-Khenifra, UV-Visible and ATR-FTMIR spectral data were processed by chemometric algorithms. PCA was applied on the spectral data set to represent the data in a very small space, and then discrimination methods were applied on the principal components synthesized by the PCA. The application of the PCA-LDA method on the spectral data of UV-Visible and ATR-FTMIR shows a good ability to classify olive oils according to their geographical origin with a percentage of correct classification that represents 90.24% and 85.87%, respectively, and the processing of the spectral data of UV-Visible and ATR-FTMIR by PCA-SVM allows differentiating correctly between five olive oils with a correct classification rate of 100% and 97.56, respectively. This study demonstrated the feasibility of UV-Visible and ATR-FTMIR fingerprinting (routine technique) for the geographical classification of olive oils in the Moroccan region of Beni Mellal-Khenifra. Such developed methods can be proposed as alternative and complementary methods to authenticate the geographical origin of virgin olive oil.

Feature Reduction Method for Cognition and Classification of IoT Devices Based on Artificial Intelligence

In recent years, the scale and application scenarios of the Internet of Things (IoT) have been expanding. Since traditional algorithms are unable to meet wireless networks computing capability requirement in the IoT, more and more research institutions and scholars have turned their eyes to artificial intelligence (AI) methods. Because the IoT device uses wireless networks to communicate in most scenarios, this paper systematically studies the method of feature dimension reduction of wireless communication signals. In this paper, we will take the power amplifier radio frequency (RF) fingerprinting as an example. Focusing on reducing the high dimensionality of RF fingerprint features and the uncorrelated or redundant features in the features space, the RF fingerprint feature dimension reduction method is mainly studied. Based on the principal component analysis (PCA), linear discriminant analysis (LDA), and auto encoder (AE) research, this paper studies the PCA-LDA method and uses the distance ratio criterion to evaluate the separability of features. The simulation results show that the classification accuracy of PCA-LDA is superior to PCA, LDA, and AE in most SNR, and the characteristics of PCA-LDA is more separable.

Raman spectroscopic detection of high-grade cervical cytology: Using morphologically normal appearing cells

This study aims to detect high grade squamous intraepithelial cells (HSIL) by investigating HSIL associated biochemical changes in morphologically normal appearing intermediate and superficial cells using Raman spectroscopy. Raman spectra (n = 755) were measured from intermediate and superficial cells from negative cytology ThinPrep specimens (n = 18) and from morphologically normal appearing intermediate and superficial cells from HSIL cytology ThinPrep specimens (n = 17). The Raman data was subjected to multivariate algorithms including the standard principal component analysis (PCA)-linear discriminant analysis (LDA) and partial least squares discriminant analysis (PLS-DA) together with random subsets cross-validation for discriminating negative cytology from HSIL. The PCA-LDA method yielded sensitivities of 74.9%, 72.8%, and 75.6% and specificities of 89.9%, 81.9%, and 84.5%, for HSIL diagnosis based on the dataset obtained from intermediate, superficial and mixed intermediate/superficial cells, respectively. The PLS-DA method provided improved sensitivities of 95.5%, 95.2% and 96.1% and specificities of 92.7%, 94.7% and 93.5% compared to the PCA-LDA method. The results demonstrate that the biochemical signatures of morphologically normal appearing cells can be used to discriminate between negative and HSIL cytology. In addition, it was found that mixed intermediate and superficial cells could be used for HSIL diagnosis as the biochemical differences between negative and HSIL cytology were greater than the biochemical differences between intermediate and superficial cell types.

Label-free and stable serum analysis based on Ag-NPs/PSi surface-enhanced Raman scattering for noninvasive lung cancer detection.

Surface-enhanced Raman scattering (SERS) has a broad application prospect in the field of tumor detection owing to its ultrahigh detective sensitivity. However, SERS analysis of serum remain a challenge in terms of repeatability and stability due to the maldistribution of the silver nanoparticles (Ag-NPs)-serum. With the aim to make up for this shortcoming, we report a new method for obtaining stable serum Raman signals utilizing the ordered arrays of pyramidal silicon (PSi) and Ag-NPs. We prove the practicability of this method by detecting the samples of serum from 50 lung cancer patients and 50 normal healthy people. Principal component analysis (PCA) of the serum SERS spectra shows that the spectral data of the two sample groups can form obvious and completely separated clusters. The receiver operating characteristic curve provides the sensitivity (100%) and specificity (90%) from the PCA-LDA method. This research indicates that a stable and label-free analysis technique of serum SERS based on Ag-NPs/PSi and PCA-LDA is promising for noninvasive lung cancer diagnoses.

Evaluation of liver fibrosis using Raman spectroscopy and infrared thermography: A pilot study

Liver fibrosis is a pathological process that can escalate to cirrhosis and then liver failure, a major public health concern that affect hundreds of millions of people in both developed and developing countries. Detection of liver fibrosis during its earlier stages is a matter of great importance which may allow prevention of development of cirrhosis in patients with chronic liver disease. In this work, Raman spectroscopy and thermography were evaluated to detect early pathological signs of liver fibrosis in rats in which liver fibrosis was induced using carbon tetrachloride. Results show that Raman spectra of healthy and fibrotic livers significantly differ among each other and can be classified by principal component analysis and discriminant analysis. The PCA-LDA method has a sensitivity of 100%, specificity 85% and diagnostic accuracy of 93.5%. Thermography also revealed characteristic temperature patterns for fibrotic livers compared to healthy livers. Current data suggest that Raman spectroscopy and thermography could be used to detect fibrosis in ex vivo liver samples.

Multiple feature descriptors based model for individual identification in group photos

Detection of person in the crowd is most exigent facial recognition aspect. Faces in a group can comprise the criticality of face localization, partial occlusion, and facial overlapping and real time background scenes. In this paper, single and multiple persons recognition is provided for group pictures. This offered method at first localized the facial area of the individuals. These identified faces are divided in full or the partial view clusters by observing the oval map. While performing the recognition, separate algorithms will be applied to both facial groups. Full face recognition is here implied using LBP, Gabor and structured feature fusion based distance analysis method. These methods will be applied individually and jointly to identify the facial image. Partial face recognition is here implied using discriminative structural analysis. This structural formation is here obtained using structure points and base curve identification. Maximum structural points and curve ratio map is considered as the identified image. Group images are captured at home, office, classroom and roadside with multiple situations. The experimentation interpretation is measured disjointedly for full view and partial view images obtained from complex scenes and scenarios. The evaluation is here applied to perform the facial region identification and the group photo recognition. The facial region extraction is applied on eight datasets. These datasets are IMM, Caltech, CMU, Bau, FDDB, LFW, Pointing and Self Collected Dataset. The facial extraction is applied on complex background images with single and multiple faces. The comparative evaluation shows that the localization stage of proposed model has reduced the FAR and FRR. The complete recognition model is applied on real time family dataset, random web celebrity dataset associated with LFW and FDDB datasets. The evaluation is provided for recognition of front and partial face images in different scenarios. The comparative results of facial recognition are evaluated against PCA, ICA and PCA-LDA methods. The results conclude that the proposed methods have improved the accuracy for both full and partial faces.

Characterization and rapid detection of adulterations in sesame oil using FT-MIR and PCA-LDA

The objective of this research work was to use Attenuated total reflectance-Fourier transform mid-infrared (ATR-FTMIR) spectroscopy coupled with chemometrics for the detection of the adulteration of sesame oil. Adulteration with sunflower oil, soybean oil or colza oil is one of the most difficult to detect due to the similar composition of them and sesame oil. Adulterations of sesame oil with different percentages of sunflower oil, soybean oil and colza oil were measured using ATR-FTMIR spectroscopy. The spectral data were subjected to a preliminary derivative elaboration based on the Savitzky–Golay algorithm to reduce the noise and extract a largest number of analytical information from spectra. Linear discriminant analysis (LDA) was adopted as classification method, and Principle component analysis (PCA) was employed to compress the original data set into a reduced new set of variables before LDA. The detection results indicated that the discriminant model built by PCA-LDA method could identify sesame oil adulterations in the 0–40% weight ratio range of edible oils, with an accuracy value of 94.64%. This work shows that PCA and LDA are useful chemometric tools for the multivariate characterization and discrimination of sesame oil adulteration with seed oils.

Sparse matrix transform-based linear discriminant analysis for hyperspectral image classification

Due to the high dimensionality of hyperspectral image (HSI), dimension reduction or feature extraction is usually needed before the HSI classification. Traditional linear discriminant analysis (LDA) method for feature extraction usually encounters difficulty because the available training samples in HSI classification are limited, which causes the singularity of data scatter matrix. In this paper, we propose a sparse matrix transform-based LDA (SMT-LDA) algorithm for the HSI classification. By using SMT, the total scatter matrix used in LDA can be constrained to have an eigen-decomposition where the eigenvectors can be sparsely parametrized by a limited number of Givens rotations. In this way, the estimated scatter matrix is always positive definite and well conditioned even in the case of limited training samples. The proposed SMT-LDA method is compared with regularized LDA and PCA-LDA methods on two benchmark hyperspectral data sets. Experimental results indicate that the performance of the proposed method is overall superior to these methods, especially for small-sample-size classification.

Obscene Video Recognition Using Fuzzy SVM and New Sets of Features

In this paper, a novel approach for identifying normal and obscene videos is proposed. In order to classify different episodes of a video independently and discard the need to process all frames, first, key frames are extracted and skin regions are detected for groups of video frames starting with key frames. In the second step, three different features including 1- structural features based on single frame information, 2- features based on spatiotemporal volume and 3-motion-based features, are extracted for each episode of video. The PCA-LDA method is then applied to reduce the size of structural features and select more distinctive features. For the final step, we use fuzzy or a Weighted Support Vector Machine (WSVM) classifier to identify video episodes. We also employ a multilayer Kohonen network as an initial clustering algorithm to increase the ability to discriminate between the extracted features into two classes of videos. Features based on motion and periodicity characteristics increase the efficiency of the proposed algorithm in videos with bad illumination and skin colour variation. The proposed method is evaluated using 1100 videos in different environmental and illumination conditions. The experimental results show a correct recognition rate of 94.2% for the proposed algorithm.

Characterization and differentiation of normal and abnormal spermatozoa via micro-Raman spectroscopy

Growth in the percentage of spermatozoa with aberrant sperm head morphologies has been correlated with a rise in male infertility. In our study, micro-Raman spectroscopy using a new substrate was utilized to characterize and differentiate the morphologically normal and abnormal human sperm cells based on their different biochemical components by showing their different specific Raman spectra. A detailed classification based on the PCA–LDA method was performed showing a diagnostic sensitivity of 76% and specificity of 91%, with 80% classification accuracy. Our results suggest that micro-Raman spectroscopy can serve as a reliable and non-invasive tool for detection and differentiation of human spermatozoa.