Combined Feature Set Research Articles

Integration of CNN Models and Machine Learning Methods in Credit Score Classification: 2D Image Transformation and Feature Extraction

Abstract The problem of accurately classifying credit scores is critical for financial institutions to assess individual creditworthiness and effectively manage credit risk. Traditional methods often face limitations when processing large datasets, resulting in lower accuracy and longer processing time. To address this issue, this paper proposes a novel approach to credit score classification by integrating convolutional neural networks (CNN) with machine learning methods. First, a 1D dataset of sequential text data is transformed into 2D greyscale images to use 2D CNN models for feature extraction and classification. Six CNN architectures—DenseNet201, GoogLeNet, MobileNetV2, ResNet18, ShuffleNet, and SqueezeNet—are implemented, and the features in the last layer (1000 features) of each CNN are classified using the softmax method. To further improve the performance, the two best CNN models were selected, and a new fully connected layer (NewFC) was added. A class-based feature set [3 × 31,695] representing three credit score types (good, poor, and standard) was extracted from each model and merged into a feature set [6 × 31,695]. This combined feature set was then reclassified using KNN, LDA, Naive Bayes, and SVM algorithms. The performance of both CNN and machine learning methods was evaluated using accuracy, precision, sensitivity, specificity, and F-score metrics. To optimize classification performance and reduce computational cost, the RelieF algorithm was used to select the best 5 out of 6 features. Compared to using all 6 features, significant improvements in accuracy and efficiency were observed, demonstrating the effectiveness of the proposed method in credit score classification.

Vegetation Classification in a Mountain–Plain Transition Zone in the Sichuan Basin, China

Developing an effective vegetation classification method for mountain–plain transition zones is critical for understanding ecological patterns, evaluating ecosystem services, and guiding conservation efforts. Existing methods perform well in mountainous and plain areas but lack verification in mountain–plain transition zones. This study utilized terrain data and Sentinel-1 and Sentinel-2 imagery to extract topographic, spectral, texture, and SAR features as well as the vegetation index. By combining feature sets and applying feature elimination algorithms, the classification performance of one-dimensional convolutional neural networks (1D-CNNs), Random Forest (RF), and Multilayer Perceptron (MLP) was evaluated to determine the optimal feature combinations and methods. The results show the following: (1) multi-feature combinations, especially spectral and topographic features, significantly improved classification accuracy; (2) Recursive Feature Elimination based on Random Forest (RF-RFE) outperformed ReliefF in feature selection, identifying more representative features; (3) all three algorithms performed well, with consistent spatial results. The MLP algorithm achieved the best overall accuracy (OA: 81.65%, Kappa: 77.75%), demonstrating robustness and lower dependence on feature quantity. This study presents an efficient and robust vegetation classification workflow, verifies its applicability in mountain–plain transition zones, and provides valuable insights for small-region vegetation classification under similar topographic conditions globally.

A novel non-invasive EEG-SSVEP diagnostic tool for color vision deficiency in individuals with locked-in syndrome.

Color vision deficiency (CVD), a common visual impairment, affects individuals' ability to differentiate between various colors due to malfunctioning or absent color photoreceptors in the retina. Currently available diagnostic tests require a behavioral response, rendering them unsuitable for individuals with limited physical and communication abilities, such as those with locked-in syndrome. This study introduces a novel, non-invasive method that employs brain signals, specifically Steady-State Visually Evoked Potentials (SSVEPs), along with Ishihara plates to diagnose CVD. This method aims to provide an alternative diagnostic tool that addresses the limitations of current tests. Electroencephalography (EEG) recordings were obtained from 16 subjects, including 5 with CVD (specifically Deuteranomaly), using channels O1, O2, Pz, and Cz. The subjects were exposed to visual stimuli at frequencies of 15 Hz and 18 Hz to assess the proposed method. The subjects focused on specific visual stimuli in response to questions related to the Ishihara plates. Their responses were analyzed to determine the presence of CVD. Feature extraction was performed using Power Spectral Density (PSD), Canonical Correlation Analysis (CCA), and a combined PSD + CCA, followed by classification to categorize subjects into two classes: normal vision and CVD. The results indicate that the proposed method effectively diagnoses CVD in individuals with limited communication abilities. The classification accuracy of SSVEP exceeded 75% across the three classifiers: Decision Tree (DT), K-Nearest Neighbors (KNN), and Support Vector Machine (SVM). The SVM classifier demonstrated higher accuracy compared to the other classifiers, exceeding 90%. These observations suggest that the SVM classifier, utilizing the combined feature set of PSD + CCA, may be the most effective in this classification task. These findings demonstrate that the proposed method is an accurate and reliable diagnostic tool for CVD, particularly for individuals unable to communicate.

Detection of Aspergilloma Disease Using Feature-Selection-Based Vision Transformers.

Background: Aspergilloma disease is a fungal mass found in organs such as the sinuses and lungs, caused by the fungus Aspergillus. This disease occurs due to the accumulation of mucus, inflamed cells, and altered blood elements. Various surgical methods are used in clinical settings for the treatment of aspergilloma disease. Expert opinion is crucial for the diagnosis of the disease. Recent advancements in next-generation technologies have made them crucial for disease detection. Deep-learning models, which benefit from continuous technological advancements, are already integrated into current early diagnosis systems. Methods: This study is distinguished by the use of vision transformers (ViTs) rather than traditional deep-learning models. The data used in this study were obtained from patients treated at the Department of Thoracic Surgery at Fırat University. The dataset consists of two class types: aspergilloma disease images and non-aspergilloma disease images. The proposed approach consists of pre-processing, model training, feature extraction, efficient feature selection, feature fusion, and classification processes. In the pre-processing step, unnecessary regions of the images were cropped and data augmentation techniques were applied for model training. Three types of ViT models (vit_base_patch16, vit_large_patch16, and vit_base_resnet50) were used for model training. The feature sets obtained from training the models were merged, and the combined feature set was processed using feature selection methods (Chi2, mRMR, and Relief). Efficient features selected by these methods (Chi2 and mRMR, Chi2 and Relief, and mRMR and Relief) were combined in certain proportions to obtain more effective feature sets. Machine-learning methods were used in the classification process. Results: The most successful result in the detection of aspergilloma disease was achieved using Support Vector Machines (SVMs). The SVM method achieved a 99.70% overall accuracy with the cross-validation technique in classification. Conclusions: These results highlight the benefits of the suggested method for identifying aspergilloma.

Multimodal AI Combining Clinical and Imaging Inputs Improves Prostate Cancer Detection.

Deep learning (DL) studies for the detection of clinically significant prostate cancer (csPCa) on magnetic resonance imaging (MRI) often overlook potentially relevant clinical parameters such as prostate-specific antigen, prostate volume, and age. This study explored the integration of clinical parameters and MRI-based DL to enhance diagnostic accuracy for csPCa on MRI. We retrospectively analyzed 932 biparametric prostate MRI examinations performed for suspected csPCa (ISUP ≥2) at 2 institutions. Each MRI scan was automatically analyzed by a previously developed DL model to detect and segment csPCa lesions. Three sets of features were extracted: DL lesion suspicion levels, clinical parameters (prostate-specific antigen, prostate volume, age), and MRI-based lesion volumes for all DL-detected lesions. Six multimodal artificial intelligence (AI) classifiers were trained for each combination of feature sets, employing both early (feature-level) and late (decision-level) information fusion methods. The diagnostic performance of each model was tested internally on 20% of center 1 data and externally on center 2 data (n = 529). Receiver operating characteristic comparisons determined the optimal feature combination and information fusion method and assessed the benefit of multimodal versus unimodal analysis. The optimal model performance was compared with a radiologist using PI-RADS. Internally, the multimodal AI integrating DL suspicion levels with clinical features via early fusion achieved the highest performance. Externally, it surpassed baselines using clinical parameters (0.77 vs 0.67 area under the curve [AUC], P < 0.001) and DL suspicion levels alone (AUC: 0.77 vs 0.70, P = 0.006). Early fusion outperformed late fusion in external data (0.77 vs 0.73 AUC, P = 0.005). No significant performance gaps were observed between multimodal AI and radiologist assessments (internal: 0.87 vs 0.88 AUC; external: 0.77 vs 0.75 AUC, both P > 0.05). Multimodal AI (combining DL suspicion levels and clinical parameters) outperforms clinical and MRI-only AI for csPCa detection. Early information fusion enhanced AI robustness in our multicenter setting. Incorporating lesion volumes did not enhance diagnostic efficacy.

ScHiClassifier: a deep learning framework for cell type prediction by fusing multiple feature sets from single-cell Hi-C data.

Single-cell high-throughput chromosome conformation capture (Hi-C) technology enables capturing chromosomal spatial structure information at the cellular level. However, to effectively investigate changes in chromosomal structure across different cell types, there is a requisite for methods that can identify cell types utilizing single-cell Hi-C data. Current frameworks for cell type prediction based on single-cell Hi-C data are limited, often struggling with features interpretability and biological significance, and lacking convincing and robust classification performance validation. In this study, we propose four new feature sets based on the contact matrix with clear interpretability and biological significance. Furthermore, we develop a novel deep learning framework named scHiClassifier based on multi-head self-attention encoder, 1D convolution and feature fusion, which integrates information from these four feature sets to predict cell types accurately. Through comprehensive comparison experiments with benchmark frameworks on six datasets, we demonstrate the superior classification performance and the universality of the scHiClassifier framework. We further assess the robustness of scHiClassifier through data perturbation experiments and data dropout experiments. Moreover, we demonstrate that using all feature sets in the scHiClassifier framework yields optimal performance, supported by comparisons of different feature set combinations. The effectiveness and the superiority of the multiple feature set extraction are proven by comparison with four unsupervised dimensionality reduction methods. Additionally, we analyze the importance of different feature sets and chromosomes using the "SHapley Additive exPlanations" method. Furthermore, the accuracy and reliability of the scHiClassifier framework in cell classification for single-cell Hi-C data are supported through enrichment analysis. The source code of scHiClassifier is freely available at https://github.com/HaoWuLab-Bioinformatics/scHiClassifier.

Detection of Thymoma Disease Using mRMR Feature Selection and Transformer Models

Background: Thymoma is a tumor that originates in the thymus gland, a part of the human body located behind the breastbone. It is a malignant disease that is rare in children but more common in adults and usually does not spread outside the thymus. The exact cause of thymic disease is not known, but it is thought to be more common in people infected with the EBV virus at an early age. Various surgical methods are used in clinical settings to treat thymoma. Expert opinion is very important in the diagnosis of the disease. Recently, next-generation technologies have become increasingly important in disease detection. Today’s early detection systems already use transformer models that are open to technological advances. Methods: What makes this study different is the use of transformer models instead of traditional deep learning models. The data used in this study were obtained from patients undergoing treatment at Fırat University, Department of Thoracic Surgery. The dataset consisted of two types of classes: thymoma disease images and non-thymoma disease images. The proposed approach consists of preprocessing, model training, feature extraction, feature set fusion between models, efficient feature selection, and classification. In the preprocessing step, unnecessary regions of the images were cropped, and the region of interest (ROI) technique was applied. Four types of transformer models (Deit3, Maxvit, Swin, and ViT) were used for model training. As a result of the training of the models, the feature sets obtained from the best three models were merged between the models (Deit3 and Swin, Deit3 and ViT, Deit3 and ViT, Swin and ViT, and Deit3 and Swin and ViT). The combined feature set of the model (Deit3 and ViT) that gave the best performance with fewer features was analyzed using the mRMR feature selection method. The SVM method was used in the classification process. Results: With the mRMR feature selection method, 100% overall accuracy was achieved with feature sets containing fewer features. The cross-validation technique was used to verify the overall accuracy of the proposed approach and 99.22% overall accuracy was achieved in the analysis with this technique. Conclusions: These findings emphasize the added value of the proposed approach in the detection of thymoma.

Enhancing Predictive Accuracy for Recurrence-Free Survival in Head and Neck Tumor: A Comparative Study of Weighted Fusion Radiomic Analysis.

Despite advancements in oncology, predicting recurrence-free survival (RFS) in head and neck (H&N) cancer remains challenging due to the heterogeneity of tumor biology and treatment responses. This study aims to address the research gap in the prognostic efficacy of traditional clinical predictors versus advanced radiomics features and to explore the potential of weighted fusion techniques for enhancing RFS prediction. We utilized clinical data, radiomic features from CT and PET scans, and various weighted fusion algorithms to stratify patients into low- and high-risk groups for RFS. The predictive performance of each model was evaluated using Kaplan-Meier survival analysis, and the significance of differences in RFS rates was assessed using confidence interval (CI) tests. The weighted fusion model with a 90% emphasis on PET features significantly outperformed individual modalities, yielding the highest C-index. Additionally, the incorporation of contextual information by varying peritumoral radii did not substantially improve prediction accuracy. While the clinical model and the radiomics model, individually, did not achieve statistical significance in survival differentiation, the combined feature set showed improved performance. The integration of radiomic features with clinical data through weighted fusion algorithms enhances the predictive accuracy of RFS outcomes in head and neck cancer. Our findings suggest that the utilization of multi-modal data helps in developing more reliable predictive models and underscore the potential of PET imaging in refining prognostic assessments. This study propels the discussion forward, indicating a pivotal step toward the adoption of precision medicine in cancer care.

Predicting the Arousal and Valence Values of Emotional States Using Learned, Predesigned, and Deep Visual Features.

The cognitive state of a person can be categorized using the circumplex model of emotional states, a continuous model of two dimensions: arousal and valence. The purpose of this research is to select a machine learning model(s) to be integrated into a virtual reality (VR) system that runs cognitive remediation exercises for people with mental health disorders. As such, the prediction of emotional states is essential to customize treatments for those individuals. We exploit the Remote Collaborative and Affective Interactions (RECOLA) database to predict arousal and valence values using machine learning techniques. RECOLA includes audio, video, and physiological recordings of interactions between human participants. To allow learners to focus on the most relevant data, features are extracted from raw data. Such features can be predesigned, learned, or extracted implicitly using deep learners. Our previous work on video recordings focused on predesigned and learned visual features. In this paper, we extend our work onto deep visual features. Our deep visual features are extracted using the MobileNet-v2 convolutional neural network (CNN) that we previously trained on RECOLA's video frames of full/half faces. As the final purpose of our work is to integrate our solution into a practical VR application using head-mounted displays, we experimented with half faces as a proof of concept. The extracted deep features were then used to predict arousal and valence values via optimizable ensemble regression. We also fused the extracted visual features with the predesigned visual features and predicted arousal and valence values using the combined feature set. In an attempt to enhance our prediction performance, we further fused the predictions of the optimizable ensemble model with the predictions of the MobileNet-v2 model. After decision fusion, we achieved a root mean squared error (RMSE) of 0.1140, a Pearson's correlation coefficient (PCC) of 0.8000, and a concordance correlation coefficient (CCC) of 0.7868 on arousal predictions. We achieved an RMSE of 0.0790, a PCC of 0.7904, and a CCC of 0.7645 on valence predictions.

Combined Feature Set with Logistic Regression Model to Detect Credit Card Frauds in Real Time Applications

Online payment methods are gaining popularity and are widely used, both in-store and online. Because to the Internet and smart mobile devices, conducting such transactions is quick, simple, and stress-free. However, online payment fraud is common due to the open nature of the internet, which allows criminals to use techniques such as eavesdropping, phishing, infiltration, denial-of-service, database theft, and man-in-the-middle assault. Online payment fraud is on the rise, and it is a big contributor to global economic losses. Financial services, healthcare, insurance, and other industries have long been plagued by fraud. Online fraud has developed in tandem with the use of digital payment systems such as credit/debit cards, PhonePe, Gpay, and Paytm. Furthermore, fraudsters and criminals are adept at evasion strategies, allowing them to steal more. Developing a secure system for client authentication and fraud protection is tough since there is always a workaround. This means that fraud detection systems play an important role in preventing financial crimes. Over time, victims of internet transaction fraud have incurred tremendous financial losses. The growth of cutting-edge technologies and global connection has led to a surge in online fraud. To reduce these expenses, it is critical to develop effective fraud detection systems. Machine learning and statistical tools make detecting dishonest money deals much easier. The scarcity of data, the sensitive nature of the data, and the uneven class distributions make it challenging to implement efficient fraud detection models. Given the delicate nature of the information, it is difficult to draw conclusions and construct more accurate models. This study offers a Linked Feature Set with Combined Feature Set with Logistic Regression (CFS-LoR) Model for accurate detection of online payment frauds. In comparison to extant models, the proposed model exhibits a highly accurate detection capability.

Interpretable Machine Learning Prediction of Voltage and Specific Capacity for Electrode Materials

AbstractThe application of machine learning (ML) is becoming widespread and playing an important role in the property prediction and design of battery. Average voltage and specific capacity are two important indicators of electrode materials, which determine suitability of electrode materials in batteries. In this study, 4351 data are collected from Material Project and 324 features are extracted from Material Project or calculated by Matminer. Four ML models are discussed for prediction of average voltage and specific capacity. In order to ensure the prediction accuracy and efficiency of ML models, sequential backward selection (SBS) method is introduced to select optimal feature set, which can reduce redundancy features. The combination of Deep Neural Network model and optimal feature set selected by the SBS method achieves accurate prediction and outperforms Crystal Graph Convolutional Neural Network and Graph‐attention Graph Neural Network. The interpretability analysis provides insight into the relationship between features and two target properties. To cope with the poor prediction performance for metal‐ion battery with scarce data, transfer learning is adopted and an excellent improvement is achieved in the prediction in Na‐ion, Mg‐ion and Ca‐ion electrode material. It can be concluded that ML is an effective approach to battery property prediction and design.

Predicting peripheral neuropathy following neoadjuvant therapy in patients with breast cancer.

e12639 Background: Administration of neoadjuvant chemotherapy (NACT) or radiation is widely used to improve outcomes in patients with breast cancer. This strategy has demonstrated remarkable efficacy in reducing tumor burden and increasing the likelihood of breast-conserving surgery. However, neoadjuvant therapy can induce peripheral neuropathy, a condition characterized by nerve damage or dysfunction in the extremities that often manifests as pain, tingling, or weakness in the hands and feet. The exact mechanism remains unclear, although likely involves the cumulative toxic effects of therapy on peripheral nerves. Treatment-induced neuropathy poses a substantial unaddressed challenge that can significantly impact the patient’s quality of life. Therefore, predicting and ameliorating neuropathy are crucial aspects of comprehensive cancer care. Methods: We used an AI spatial biology platform to create and assess 3D depictions of a cohort of breast cancer patients based on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) data. Clinical, pathological, and radiological data were collected on 74 patients from Yale New Haven Hospital. Due to missing regimen data, the final dataset included 71 patients, with mean age 51.1 (14.5) years. An AI-based segmentation algorithm was successfully employed on 63 patients, of whom 13 had neuropathy following NACT. Vascular and microvascular (MV/V) kinetic parameters were computed for each, and radiomic features were extracted. Machine learning models were developed for clinical features, MV/V features, and the combined set of features. All models were tested with leave-one-out cross-validation. Results: The clinical model did not forecast neuropathy in any patient, indicating poor predictive ability. The MV/V model performed the best, with accuracy, precision, recall, and AUROC of 0.84, 1.0, 0.18, and 0.69, respectively. The combined model showed intermediate performance, with accuracy, precision, recall, and AUROC of 0.81, 0.5, 0.18, and 0.58, respectively. The patient population limits broad conclusions, but suggests features that are aligned with increased occurrence of neuropathy. In particular, vascular features that describe transfer of nutrients between tissues, as well as the density of blood vessels within and around the tumor, are prominent within both the MV/V and combined models. Conclusions: These results indicate that features associated with the degree of proximal and intratumoral vascularity may be predictive of peripheral neuropathy in breast cancer patients. Moreover, these features can be extracted from standard-of-care DCE-MRI data. Expansion of this vascular analysis could enable physicians to better plan and manage care for patients by tailoring treatment dose. [Table: see text]

Application of Statistical Features in Vibro-acoustic Signals to Detect Early Browning Disorder in Pears Compared with Food Chemistry Method

- Browning in pears is one of the most serious diseases in pear fruit, which is caused by Alternaria alternata. The browning process is accompanied by changes in the chemical properties of the fruit, which affect its taste and nutritional composition. Additionally, as a typical postharvest disease, internal browning in pears can cause fruit tissue decay during storage that can reduce the shelf stability of fruit, and bring serious losses to sellers. Because it is difficult to identify the browning pears by appearance, a non-destructive detection technology is highly desirable to correctly discriminate a pear at the early stage of browning for increasing the market value. Firstly, 11 and 7 statistical features were calculated from the time-domain and frequency-domain, respectively. Then, sensitive features in time-domain set, frequency-domain set and combined feature set were selected by the distance evaluation technology, respectively. These selected features were used to train classifier based on K-nearest neighbor algorithm under different K-values. With the selected combined features adopted, the constructed KNN classifier performed the best classification performance. It allowed a high overall accuracy of 91.8 % to classify the healthy and browning pears. Also, the F1 value of 92.6 % indicated that the classifier can be successfully generalized. Therefore, the classification model established in this study is effective for identifying the early browning disorder in pear fruit.

Abstract 3515: SynAI: An AI-driven cancer drug synergism prediction platform

Abstract Introduction: The SynAI solution is an adaptive AI-driven in silico drug synergism screening solution aiming to discover the potential therapeutic value of compounds at an early development stage. Given one of both compound inputs in SMILE sequences, SynAI can predict the potential Bliss score of compounds in any given cell line without the need for compound synthesis or structural analysis.The evaluation version of SynAI is accessible at https://synai.crownbio.com Methods: The AI core of SynAI was constructed using the MLP (multi-layer perceptron) network under the PyTorch machine learning framework. The AI core of SynAI was trained against NCI-Almanac and DrugCombDB datasets. In total, these datasets consist of over 12 million in vitro synergism tests across 150 cancer cell lines of different origins. On average, each cell line is tested against over 6000 two-compound combinations of FDA-approved cancer drugs. One MLP network was trained for each cell line. Essentially, these networks predict the Bliss score for any combination of SMILE-based feature sets known as molecular fingerprints. The networks were trained against the NCI Almanac set and verified against NCI and DrugCombDB using n-fold cross-validation to avoid model overfitting. During the training, a hyperparameter tuning (HT) study was performed for SynAI and other popular algorithms; allowing an objective comparison of SynAI performances. Results: Compared to existing synergism prediction platforms, SynAI yields similar performance in all categories (cf. Table 1) but provides more flexibility for data input using SMILE sequences directly. In addition, the AI core of SynAI can be constantly updated with new inputs from different cell lines and drug combinations. Table 1. Pearson cross correlation (PCC) between measured and predicted Bliss scores of NCI test set Conclusions: Its adaptive and dynamic nature allows SynAI to learn from new data feeds from future studies and has significant potential in reducing the time and cost of synergism screening. PCC Score Cell Lines Algorithm MCF7 OVCAR-8 SK-MEL-5 SynAI 0.68 ± 0.02 0.56 ± 0.07 0.86 ± 0.02 RF 0.64 ± 0.02 0.55 ± 0.03 0.89 ± 0.02 GBX 0.66 ± 0.02 0.48 ± 0.05 0.88 ± 0.02 RNN 0.54 ± 0.12 0.43 ± 0.06 0.83 ± 0.08 Citation Format: Kuan Yan, Runjun Jia, Sheng Guo. SynAI: An AI-driven cancer drug synergism prediction platform [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 3515.

Supervised learning for automatic emotion recognition in Parkinson’s disease through smartwatch signals

People with Parkinson’s Disease (PwPD) usually experience several neuropsychiatric signs such as anxiety, depression, and negative emotions that contribute to disability and worsening of quality of life. Notwithstanding, the assessment of these symptoms are largely underrated, subjective and difficult due to a large overlapping with other PD symptoms, like hypomimia and bradikinesia. The aim and novelty of the current work is to study and validate a method for automatic emotion recognition in PwPD during daily living through autonomic signals acquired by acceptable and low-cost consumer technology. The best shallow learning algorithm and the best minimal feature set are individuated. 11 PwPD and 8 subjects with no history of neurological injury or illness were enrolled in the study. Participants were asked to watch video clips purposely selected to arouse emotions, and annotate arousal and valence of emotions triggered by video clips, while their heart rate, skin conductance, and temperature were recorded by a smartwatch. Smartwatch data was used for features extraction, while participants’ reported arousal and valence were used as gold-standard to train machine learning algorithms for emotion classification (low/high arousal, positive/negative valence). Different feature sets and different algorithms (i.e. decision tree (DT), random forest (RF), support vector machine (SVM), and multilayer perceptron (MLP)) were evaluated to find the best solution for each group of participant. In each group of participants, it was possible to find a combination of feature set and algorithm to reach a classification accuracy greater than 90%. The Random Forest reached the best performance in both groups and for both valence and arousal. For each classification task (valence or arousal, PwPD or controls), the best model was selected and the minimal feature set was found by performing a recursive feature elimination based on the Shapley value. A lower accuracy of appraisal emerged for arousal compared to valence. Obtained results showed the feasibility of automatic emotion recognition in PwPDs through autonomic signals. Autonomic dysfunction in PwPDs may explain the lower arousal accuracy. The findings warrant confirmation from trials on larger samples and there are open issues to be deepened in future work.

SVM-Root: Identification of Root-Associated Proteins in Plants by Employing the Support Vector Machine with Sequence-Derived Features

Background: Root is a desirable trait for modern plant breeding programs, as the roots play a pivotal role in the growth and development of plants. Therefore, identification of the genes governing the root traits is an essential research component. With regard to the identification of root-associated genes/proteins, the existing wet-lab experiments are resource intensive and the gene expression studies are species-specific. Thus, we proposed a supervised learning-based computational method for the identification of root-associated proteins. Method: The problem was formulated as a binary classification, where the root-associated proteins and non-root-associated proteins constituted the two classes. Four different machine learning algorithms such as support vector machine (SVM), extreme gradient boosting, random forest, and adaptive boosting were employed for the classification of proteins of the two classes. Sequence-derived features such as AAC, DPC, CTD, PAAC, and ACF were used as input for the learning algorithms. Results: The SVM achieved higher accuracy with the 250 selected features of AAC+DPC+CTD than that of other possible combinations of feature sets and learning algorithms. Specifically, SVM with the selected features achieved overall accuracies of 0.74, 0.73, and 0.73 when evaluated with single 5-fold cross-validation (5F-CV), repeated 5F-CV, and independent test set, respectively. Conclusions: A web-enabled prediction tool SVM-Root (https://iasri-sg.icar.gov.in/svmroot/) has been developed for the computational prediction of the root-associated proteins. Being the first of its kind, the proposed model is believed to supplement the existing experimental methods and high throughput GWAS and transcriptome studies.

Deep causal speech enhancement and recognition using efficient long-short term memory Recurrent Neural Network.

Long short-term memory (LSTM) has been effectively used to represent sequential data in recent years. However, LSTM still struggles with capturing the long-term temporal dependencies. In this paper, we propose an hourglass-shaped LSTM that is able to capture long-term temporal correlations by reducing the feature resolutions without data loss. We have used skip connections in non-adjacent layers to avoid gradient decay. In addition, an attention process is incorporated into skip connections to emphasize the essential spectral features and spectral regions. The proposed LSTM model is applied to speech enhancement and recognition applications. The proposed LSTM model uses no future information, resulting in a causal system suitable for real-time processing. The combined spectral feature sets are used to train the LSTM model for improved performance. Using the proposed model, the ideal ratio mask (IRM) is estimated as a training objective. The experimental evaluations using short-time objective intelligibility (STOI) and perceptual evaluation of speech quality (PESQ) have demonstrated that the proposed model with robust feature representation obtained higher speech intelligibility and perceptual quality. With the TIMIT, LibriSpeech, and VoiceBank datasets, the proposed model improved STOI by 16.21%, 16.41%, and 18.33% over noisy speech, whereas PESQ is improved by 31.1%, 32.9%, and 32%. In seen and unseen noisy situations, the proposed model outperformed existing deep neural networks (DNNs), including baseline LSTM, feedforward neural network (FDNN), convolutional neural network (CNN), and generative adversarial network (GAN). With the Kaldi toolkit for automated speech recognition (ASR), the proposed model significantly reduced the word error rates (WERs) and reached an average WER of 15.13% in noisy backgrounds.

A Model Free Gait Recognition using Random Forest Method

This paper presents an approach for model free gait recognition using supervised learning. The method involves utilizing gait entropy images (GEnIs) as features to represent gait information. These GEnI are generated by calculating entropy for individual pixels within silhouettes throughout a full gait cycle. To enhance the robustness of gait recognition, we introduce the inclusion of view angle information, which corresponds to the perspective from which the gait sequence is captured. This additional feature is combined with the GEnI in order to create a comprehensive representation. In scenarios such as real-time CCTV footage, the view angles are frequently unknown. To address this, we propose a two phase algorithm. In the first phase, we predict the view angle of a gait sequence. This prediction is achieved by training a random forest classifier. Once the predicted angle feature is obtained, it is integrated with the GEnI feature. Subsequently, in the second phase, another random forest classifier is trained for gait recognition based on this combined feature set. The capabilty of the proposed approach is assessed through experiments performed on public dataset CASIA-B. Experimental outcomes and comparison with the state-of-the-art provide substantial evidence supporting the effectiveness of our approach. In a cross-walking scenario, we attain accuracies of 95.91% for regular walking, 62.21% for walking while carrying a bag, and 27.64% for walking while wearing a coat.

Automated Cough Sound Analysis for Detecting Childhood Pneumonia.

Pneumonia is one of the leading causes of death in children. Prompt diagnosis and treatment can help prevent these deaths, particularly in resource poor regions where deaths due to pneumonia are highest. Clinical symptom-based screening of childhood pneumonia yields excessive false positives, highlighting the necessity for additional rapid diagnostic tests. Cough is a prevalent symptom of acute respiratory illnesses and the sound of a cough can indicate the underlying pathological changes resulting from respiratory infections. In this study, we propose a fully automated approach to evaluate cough sounds to distinguish pneumonia from other acute respiratory diseases in children. The proposed method involves cough sound denoising, cough sound segmentation, and cough sound classification. The denoising algorithm utilizes multi-conditional spectral mapping with a multilayer perceptron network while the segmentation algorithm detects cough sounds directly from the denoised audio waveform. From the segmented cough signal, we extract various handcrafted features and feature embeddings from a pretrained deep learning network. A multilayer perceptron is trained on the combined feature set for detecting pneumonia. The method we propose is evaluated using a dataset comprising cough sounds from 173 children diagnosed with either pneumonia or other acute respiratory diseases. On average, the denoising algorithm improved the signal-to-noise ratio by 44%. Furthermore, a sensitivity and specificity of 91% and 86%, respectively, is achieved in cough segmentation and 82% and 71%, respectively, in detecting childhood pneumonia using cough sounds alone. This demonstrates its potential as a rapid diagnostic tool, such as using smartphone technology.

Cardiac MRI of characteristic motion findings in right bundle branch block.

While there have been many descriptions of characteristic motion findings in left bundle branch block (LBBB), there are few published descriptions of such findings in right bundle branch block (RBBB).The purpose of this study was to assess the frequency of particular regional motion findings in cardiac magnetic resonance imaging (CMR) studies of patients with RBBB, compared with normal subjects. We focused on three distinctive motion patterns that can be seen in RBBB during early systole: delayed apex-ward motion of the RV base, "reverse septal flash", and "basal bulge". The presence and relative magnitude of these findings were independently scored by four experienced observers, in 3-chamber and 4-chamber CMR cines, for both normal subjects and patients with RBBB. These motion patterns were found to be strongly associated with the presence of RBBB. While only moderately sensitive, they were quite specific for RBBB, when present. In particular, with ROC analysis, a combined feature set of the findings in the 4-chamber view had an area under the curve of 0.81.This previously undescribed set of RBBB-associated early-systolic regional motion features (delayed apex-ward motion of the RV base, "reverse septal flash", and "basal bulge") is strongly suggestive of RBBB when present, particularly in the 4-chamber view. Although here evaluated with CMR, it is also likely to be associated with RBBB when seen with other cardiac imaging modalities.