Multiple Classification Approach Research Articles

Traffic Sign Classification using Deep Learning Comparative Study

In recent years, the advancement of deep learning technologies has significantly impacted various domains and the field of transportation is no exception. As the demand for robust and accurate traffic sign classification systems continues to rise, this study presents an in-depth exploration and comparison of various techniques employed in the field. Focusing on state-of-the-art methodologies, we assess the effectiveness and performance of multiple classification approaches for traffic sign recognition. The review presents an extensive survey of the literature, encompassing traditional computer vision methods, machine learning algorithms, and the latest advancements in deep learning. The comparative analysis aims to identify the strengths and limitations of each technique, considering factors such as computational efficiency and accuracy. Additionally, the paper implements four models—CNN, ResNet50, VGG19, and EfficientNetB7—for traffic sign classification on the GTSRB dataset, the accuracy results are reported as 99.25%, 99.28%, 98.95%, and 98.24% respectively.

Validity of Machine Learning in Detecting Complicated Appendicitis in a Resource-Limited Setting: Findings from Vietnam.

Complicated appendicitis, a potentially life-threatening condition, is common. However, the diagnosis of this condition is mainly based on physician's experiences and advanced diagnostic equipment. This study built and validated machine learning models to facilitate the detection of complicated appendicitis. A retrospective cohort study was conducted based on medical charts of all patients undergoing a laparoscopic appendectomy at a city hospital during 2016-2020. The synthetic minority over-sampling technique (SMOTE) was used to adjust for the imbalance. Multiple classification approaches were used to train and validate models including support vector machine (SVM), decision tree (DT), K-nearest neighbor (KNN), logistic regression (LR), artificial neural network (ANN), and gradient boosting (GB). Among 1,950 patients included in the data analysis, there were 483 patients identified as having complicated appendicitis (24.8%). Based on data without SMOTE adjustment for imbalance, the accuracy levels and AUCs were high in all models using different parameters, ranging from 0.687 to 0.815. After adjusting for imbalance data using SMOTE, AUC and accuracy levels in the models using imbalance adjusted data were higher. Of these, the GB had all AUC and accuracy values of approximately 0.8 or more in both adjusted and unadjusted data. Machine learning approaches including SVM, DT, logistic, KNN, ANN, and GB have a high level of validity in classifying patients with complicated appendicitis and patients without complicated appendicitis. Among these, GB had the highest level of validity and should be used or further validated. Our study indicates the beneficial potentials of machine learning techniques in a clinical setting in general and in the diagnosis of complicated appendicitis in particular.

Drug identification by electroanalysis with multiple classification approaches

Selective approaches in analytical chemistry are essential in drug analysis. Spectroscopic methods are often performed to identify drugs, and electroanalysis is mostly used in quantitative assays. However, chemometrics are tools that can be applied to analytical methods in order to enhance selectivity. Thus, the aim of this work was to propose a novel drug identification approach with electroanalysis with several classification algorithms and to compare those results with Raman spectroscopy. The commercial tablets and analytical standards of diclofenac, amoxicillin, hydrochlorothiazide, bromazepam, piroxicam, sulfadiazine, albendazole, cyclobenzaprine and ibuprofen were studied. The classification approaches of linear discrimination analysis (LDA), partial least squares discriminant analysis (PLSDA), support vector machine and k-nearest neighbours (KNN) were employed. Differential pulse voltammetry and Raman spectroscopy were used. The models AS, CT and AS-CT were proposed by the division of the dataset in three parts, being only the analytical standards (AS), only the commercial tablets (CT) and both analytical standards and commercial tablets (AS-CT). For voltammetry and Raman LDA and PLSDA yielded the best accuracy in model AS. For the commercial tablets (model CT), voltammetry showed better performance than Raman. In model AS-CT the KNN classifications displayed better accuracy for both voltammetry and Raman. Voltammetry showed a great performance in the identification of analytical standards and commercial tablet drugs, higher versatility and easier employment due to the nonrequirement of variable selection in comparison to Raman and showed a better performance than Raman for the identification of commercial tablets.

Plant Traits Response to Grazing Exclusion by Fencing Assessed via Multiple Classification Approach: A Case from a Subalpine Meadow

Using multiple classification approach to examine plant traits response to grazing and fencing (as without access to grazing) is rare. Here we used multiple classification approach to examine plant diversity, productivity and species traits response to grazing and fencing over a three-year period on the eastern part of the Qing-Hai Tibetan Plateau. The results showed that most common species response to the fence was poorly noticeable. The fencing meadows compared to those under long-term free grazing are characterized by significantly higher total species richness, but species richness declined with sampling years gradually, regardless of grazing or fencing. The correlation showed that species richness was negatively associated with mean annual temperature significantly, suggesting that abiotic factors (e.g. annual temperature) could also play important roles in driving species richness in this subalpine meadow. Total aboveground biomass was not associated with mean annual temperature and rainfall. The fencing meadow demonstrated higher community aboveground biomass relative to the grazing ones, especially the abundance of legume and graminoids increased, while the proportions of sedge decreased, suggested that grazing disturbance favours the increase of reproductive success of sedge (e.g. Kobresia humilis) in this subalpine meadow. Growth form-based analyses combined with canopy height categories should be recommended to reveal general rules and mechanisms relating to grazing.

Abstract 3579: A novel algorithm to identify TCR-binding somatic mutations from human cancers

Abstract A large number of pre-clinical and clinical studies have strongly implicated that combining cancer-derived neo-antigen vaccines with checkpoint control inhibitors will enhance priming and expansion of tumor-specific naïve and memory T-cells resulting in superior efficacy and durability of response. The neo-antigens, derived from somatic mutations are prime candidates for cancer vaccines, not subjected to host’s central and peripheral tolerance. Identifying potential T-cell engaging neo-epitopes among a large number of somatic mutations is like finding a specific needle in a stack of needles. Currently, the available T-cell neo-epitope prioritization pipelines rely primarily on two attributes - the class-I HLA-binding affinity of the mutant peptide compared to the wild-type counterpart, and the level of expression of the mutated gene in tumor cells. The higher the binding differential, and higher the expression level of the mutant allele, the greater is the likelihood for the peptide to be presented on antigen-presenting cells. These approaches however fall short of predicting whether the HLA-bound peptide will engage T-cells by binding to T-cell receptors (TCRs). We have developed a novel algorithm to predict the binding of HLA-peptide complexes to TCRs by analyzing the physico-chemical composition of the amino acids and their positional biases in the 9-mers from crystal structures of HLA-peptide-TCR complex. We applied machine learning approach to build a classification model that can predict whether a given 9-mer peptide is a TCR-binder or not by identifying whether an amino acid at a given position carries key features that will facilitate interaction with the TCR. We applied this approach to positive and negative TCR interactions selected from Immune Epitope data base (IEDB). We tested multiple classification approaches and found that Random Forest and ClassificationviaRegression methods provided the best performance. We achieved more than 99% accuracy at 10-fold cross validation on both training and unseen test datasets. We further validated our model using positive and negative peptides curated from published papers reporting clinical trial results of checkpoint control inhibitors. The performance of the two classification models was evaluated on two different TCR-binding assays - dextramer binding and IFN-γ release. The ClassificationviaRegression method showed a higher positive predictive value for the dextramer-binding assay, whereas the Random Forest method showed a higher positive predictive value for the IFN-γ ELISPOT assay, suggesting subtle differences between the two classification methods. The inclusion of the TCR binding step to our T-cell neo-epitope prioritization pipeline increased accuracy of prediction, reduced false positives and selected potential neo-epitopes to a manageable number for testing in cell-based assays. Citation Format: Priyanka Shah, Kiran V. Paul, Malini Manoharan, Amitabha Chaudhuri, Ravi Gupta. A novel algorithm to identify TCR-binding somatic mutations from human cancers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 3579. doi:10.1158/1538-7445.AM2017-3579

Multiple RF classifier for the hippocampus segmentation: Method and validation on EADC-ADNI Harmonized Hippocampal Protocol

The hippocampus has a key role in a number of neurodegenerative diseases, such as Alzheimer's Disease. Here we present a novel method for the automated segmentation of the hippocampus from structural magnetic resonance images (MRI), based on a combination of multiple classifiers. The method is validated on a cohort of 50 T1 MRI scans, comprehending healthy control, mild cognitive impairment, and Alzheimer's Disease subjects. The preliminary release of the EADC-ADNI Harmonized Protocol training labels is used as gold standard. The fully automated pipeline consists of a registration using an affine transformation, the extraction of a local bounding box, and the classification of each voxel in two classes (background and hippocampus). The classification is performed slice-by-slice along each of the three orthogonal directions of the 3D-MRI using a Random Forest (RF) classifier, followed by a fusion of the three full segmentations. Dice coefficients obtained by multiple RF (0.87 ± 0.03) are larger than those obtained by a single monolithic RF applied to the entire bounding box, and are comparable to state-of-the-art. A test on an external cohort of 50 T1 MRI scans shows that the presented method is robust and reliable. Additionally, a comparison of local changes in the morphology of the hippocampi between the three subject groups is performed. Our work showed that a multiple classification approach can be implemented for the segmentation for the measurement of volume and shape changes of the hippocampus with diagnostic purposes.

Statistical inference for remote sensing-based estimates of net deforestation

Statistical inference requires expression of an estimate in probabilistic terms, usually in the form of a confidence interval. An approach to constructing confidence intervals for remote sensing-based estimates of net deforestation is illustrated. The approach is based on post-classification methods using two independent forest/non-forest classifications because sufficient numbers of observations of forest/non-forest change were not available for direct classification. Further, the approach uses a model-assisted estimator with information from a traditional error matrix for the forest/non-forest classifications to compensate for bias as the result of classification errors and to estimate variances. Classifications were obtained using a logistic regression model, forest inventory data, and two dates of Landsat imagery, although the approach to inference can be used with multiple classification approaches. For the study area in northeastern Minnesota, USA, overall pixel-level accuracies for the year 2002 and 2007 forest/non-forest classifications were 0.85–0.88, and estimates of proportion net deforestation for the 2002–2007 interval were less in absolute value than 0.015. However, standard errors for the remote sensing-based estimates of net deforestation were on the order of 0.02–0.04, meaning that the estimates were not statistically significantly different from zero. Particular attention is directed to the potentially severe sample size and classification accuracy requirements necessary for estimates of net deforestation to be detected as statistically significantly different from zero.

What Makes a Pattern? Matching Decoding Methods to Data in Multivariate Pattern Analysis

Research in neuroscience faces the challenge of integrating information across different spatial scales of brain function. A promising technique for harnessing information at a range of spatial scales is multivariate pattern analysis (MVPA) of functional magnetic resonance imaging (fMRI) data. While the prevalence of MVPA has increased dramatically in recent years, its typical implementations for classification of mental states utilize only a subset of the information encoded in local fMRI signals. We review published studies employing multivariate pattern classification since the technique’s introduction, which reveal an extensive focus on the improved detection power that linear classifiers provide over traditional analysis techniques. We demonstrate using simulations and a searchlight approach, however, that non-linear classifiers are capable of extracting distinct information about interactions within a local region. We conclude that for spatially localized analyses, such as searchlight and region of interest, multiple classification approaches should be compared in order to match fMRI analyses to the properties of local circuits.

A shift from lithostratigraphic to allostratigraphic classification of Quaternary glacial deposits

The bedrock of the northern halves of North America and Europe is covered by Quaternary glacial deposits, forming a surficial overburden that is relatively thin, nonlithified, litho logically variable on a small scale (in terms of grain-size, mineralogy, texture, fabric, structure, and color), and often has a well-preserved depositional topography. These geologically unique characteristics and the fact that the glacial overburden was long considered to be of only restricted economic value have caused it to be treated differently in geological research from the older, regularly lithified strata. Due to the striking geomorphology of these glacial deposits, their investigation has also been incorporated into physical geography research. Thus, the segregation of the Quaternary research community into different schools of geology and geography has created multiple classification approaches and has caused the formal stratigraphic classifications successfully applied in pre-Quater nary geology to be applied less regularly to Quaternary glacial strata. This has led to inefficient use of Quaternary geological data for scientific and socio-economic purposes. The few currently existing national Quaternary stratigraphic frameworks are based on lithostratigraphy. These are poorly suited for describing deposits in glaciated shield areas in particular; we propose a classification for such areas based on the combined use of allostratigraphic and lithostratigraphic data, with alloformations as the fundamental units and lithostratigraphic units filling out the framework where appropriate. This classification would provide a hierarchical framework for gla ciogenic deposits that could potentially support stratigraphic information systems, databases, and digital spatial models more effectively than the traditional lithostratigraphic frameworks.

Identifying spatial relationships in neural processing using a multiple classification approach

The application of statistical classification methods to in vivo functional neuroimaging data makes it possible to explore spatial patterns in task-related changes in neural processing. Cluster analysis is one group of descriptive statistical procedures that can assist in identifying classes of brain regions that exhibit similar task-related functionality. In practice, a limitation of cluster analysis is that the performances of clustering algorithms rely on unknown characteristics of the data, making it difficult to determine which procedure best suits a particular analysis. We present a multiple classification approach that incorporates numerous algorithms, evaluates the associated classifications, and either selects a plausible partition relative to the others considered or pools the results from the numerous methods. The multiple classification approach utilizes a new performance criterion, called the relative information (RI) measure, to evaluate the quality of the candidate partitions and as the basis for producing a composite classification image. Employing multiple classifications, rather than a single algorithm, our methodology increases the chance of detecting the functional relationships within the data and, therefore, produces more reliable results. We apply our methodology to a PET study to explore spatial relationships in measured brain function associated with increasing blood alcohol concentration levels, and we perform a simulation study to evaluate the performance of RI.