Learning from Imbalanced Data

Identifying distinctive brain regions related to consumer choice behaviors on branded foods using activation likelihood estimation and machine learning

Introduction

Brand equity plays a crucial role in a brand's commercial success; however, research on the brain regions associated with brand equity has had mixed results. This study aimed to investigate key brain regions associated with the decision-making of branded and unbranded foods using quantitative neuroimaging meta-analysis and machine learning.

Methods

Quantitative neuroimaging meta-analysis was performed using the activation likelihood method. Activation of the ventral medial prefrontal cortex (VMPFC) overlapped between branded and unbranded foods. The lingual and parahippocampal gyri (PHG) were activated in the case of branded foods, whereas no brain regions were characteristically activated in response to unbranded foods. We proposed a novel predictive method based on the reported foci data, referencing the multi-voxel pattern analysis (MVPA) results. This approach is referred to as the multi-coordinate pattern analysis (MCPA). We conducted the MCPA, adopting the sparse partial least squares discriminant analysis (sPLS-DA) to detect unique brain regions associated with branded and unbranded foods based on coordinate data. The sPLS-DA is an extended PLS method that enables the processing of categorical data as outcome variables.

Results

We found that the lingual gyrus is a distinct brain region in branded foods. Thus, the VMPFC might be a core brain region in food categories in consumer behavior, regardless of whether they are branded foods. Moreover, the connection between the PHG and lingual gyrus might be a unique neural mechanism in branded foods.

Discussion

As this mechanism engages in imaging the feature-self based on emotionally subjective contextual associative memories, brand managers should create future-oriented relevancies between brands and consumers to build valuable brands.

SNO-DCA: A model for predicting S-nitrosylation sites based on densely connected convolutional networks and attention mechanism

Protein S-nitrosylation is a reversible oxidative reduction post-translational modification that is widely present in the biological community. S-nitrosylation can regulate protein function and is closely associated with a variety of diseases, thus identifying S-nitrosylation sites are crucial for revealing the function of proteins and related drug discovery. Traditional experimental methods are time-consuming and expensive; therefore, it is necessary to explore more efficient computational methods. Deep learning algorithms perform well in the field of bioinformatics sites prediction, and many studies show that they outperform existing machine learning algorithms. In this work, we proposed a deep learning algorithm-based predictor SNO-DCA for distinguishing between S-nitrosylated and non-S-nitrosylated sequences. First, one-hot encoding of protein sequences was performed. Second, the dense convolutional blocks were used to capture feature information, and an attention module was added to weigh different features to improve the prediction ability of the model. The 10-fold cross-validation and independent testing experimental results show that our SNO-DCA model outperforms existing S-nitrosylation sites prediction models under imbalanced data. In this paper, a web server prediction website: https://sno.cangmang.xyz/SNO-DCA/was established to provide an online prediction service for users. SNO-DCA can be available at https://github.com/peanono/SNO-DCA.

Interpretable machine learning models for predicting in-hospital and 30 days adverse events in acute coronary syndrome patients in Kuwait

The relationships between acute coronary syndromes (ACS) adverse events and the associated risk factors are typically complicated and nonlinear, which poses significant challenges to clinicians' attempts at risk stratification. Here, we aim to explore the implementation of modern risk stratification tools to untangle how these complex factors shape the risk of adverse events in patients with ACS. We used an interpretable multi-algorithm machine learning (ML) approach and clinical features to fit predictive models to 1,976 patients with ACS in Kuwait. We demonstrated that random forest (RF) and extreme gradient boosting (XGB) algorithms, remarkably outperform traditional logistic regression model (AUCs = 0.84 & 0.79 for RF and XGB, respectively). Our in-hospital adverse events model identified left ventricular ejection fraction as the most important predictor with the highest interaction strength with other factors. However, using the 30-days adverse events model, we found that performing an urgent coronary artery bypass graft was the most important predictor, with creatinine levels having the strongest overall interaction with other related factors. Our ML models not only untangled the non-linear relationships that shape the clinical epidemiology of ACS adverse events but also elucidated their risk in individual patients based on their unique features.

Neural Systems Underlying the Implementation of Working Memory Removal Operations

Recently, multi-voxel pattern analysis has verified that information can be removed from working memory (WM) via three distinct operations replacement, suppression, or clearing compared to information being maintained ( Kim et al., 2020). While univariate analyses and classifier importance maps in Kim et al. (2020) identified brain regions that contribute to these operations, they did not elucidate whether these regions represent the operations similarly or uniquely. Using Leiden-community-detection on a sample of 55 humans (17 male), we identified four brain networks, each of which has a unique configuration of multi-voxel activity patterns by which it represents these WM operations. The visual network (VN) shows similar multi-voxel patterns for maintain and replace, which are highly dissimilar from suppress and clear, suggesting this network differentiates whether an item is held in WM or not. The somatomotor network (SMN) shows a distinct multi-voxel pattern for clear relative to the other operations, indicating the uniqueness of this operation. The default mode network (DMN) has distinct patterns for suppress and clear, but these two operations are more similar to each other than to maintain and replace, a pattern intermediate to that of the VN and SMN. The frontoparietal control network (FPCN) displays distinct multi-voxel patterns for each of the four operations, suggesting that this network likely plays an important role in implementing these WM operations. These results indicate that the operations involved in removing information from WM can be performed in parallel by distinct brain networks, each of which has a particular configuration by which they represent these operations.

CDMO: Chaotic Dwarf Mongoose Optimization Algorithm for feature selection

In this paper, a modified version of Dwarf Mongoose Optimization Algorithm (DMO) for feature selection is proposed. DMO is a novel technique of the swarm intelligence algorithms which mimic the foraging behavior of the Dwarf Mongoose. The developed method, named Chaotic DMO (CDMO), is considered a wrapper-based model which selects optimal features that give higher classification accuracy. To speed up the convergence and increase the effectiveness of DMO, ten chaotic maps were used to modify the key elements of Dwarf Mongoose movement during the optimization process. To evaluate the efficiency of the CDMO, ten different UCI datasets are used and compared against the original DMO and other well-known Meta-heuristic techniques, namely Ant Colony optimization (ACO), Whale optimization algorithm (WOA), Artificial rabbit optimization (ARO), Harris hawk optimization (HHO), Equilibrium optimizer (EO), Ring theory based harmony search (RTHS), Random switching serial gray-whale optimizer (RSGW), Salp swarm algorithm based on particle swarm optimization (SSAPSO), Binary genetic algorithm (BGA), Adaptive switching gray-whale optimizer (ASGW) and Particle Swarm optimization (PSO). The experimental results show that the CDMO gives higher performance than the other methods used in feature selection. High value of accuracy (91.9-100%), sensitivity (77.6-100%), precision (91.8-96.08%), specificity (91.6-100%) and F-Score (90-100%) for all ten UCI datasets are obtained. In addition, the proposed method is further assessed against CEC'2022 benchmarks functions.

Relabeling for Indoor Localization Using Stationary Beacons in Nursing Care Facilities

In this study, we propose an augmentation method for machine learning based on relabeling data in caregiving and nursing staff indoor localization with Bluetooth Low Energy (BLE) technology. Indoor localization is used to monitor staff-to-patient assistance in caregiving and to gain insights into workload management. However, improving accuracy is challenging when there is a limited amount of data available for training. In this paper, we propose a data augmentation method to reuse the Received Signal Strength (RSS) from different beacons by relabeling to the locations with less samples, resolving data imbalance. Standard deviation and Kullback-Leibler divergence between minority and majority classes are used to measure signal pattern to find matching beacons to relabel. By matching beacons between classes, two variations of relabeling are implemented, specifically full and partial matching. The performance is evaluated using the real-world dataset we collected for five days in a nursing care facility installed with 25 BLE beacons. A Random Forest model is utilized for location recognition, and performance is compared using the weighted F1-score to account for class imbalance. By increasing the beacon data with our proposed relabeling method for data augmentation, we achieve a higher minority class F1-score compared to augmentation with Random Sampling, Synthetic Minority Oversampling Technique (SMOTE) and Adaptive Synthetic Sampling (ADASYN). Our proposed method utilizes collected beacon data by leveraging majority class samples. Full matching demonstrated a 6 to 8% improvement from the original baseline overall weighted F1-score.

Unlabeled data selection for active learning in image classification

Active Learning has emerged as a viable solution for addressing the challenge of labeling extensive amounts of data in data-intensive applications such as computer vision and neural machine translation. The main objective of Active Learning is to automatically identify a subset of unlabeled data samples for annotation. This identification process is based on an acquisition function that assesses the value of each sample for model training. In the context of computer vision, image classification is a crucial task that typically requires a substantial training dataset. This research paper introduces innovative selection methods within the Active Learning framework, aiming to identify informative images from unlabeled datasets while minimizing the number of required training data. The proposed methods, namely Similari-ty-based Selection, Prediction Probability-based Selection, and Competence-based Active Learning, have been extensively evaluated through experiments conducted on popular datasets like Cifar10 and Cifar100. The experimental results demonstrate that the proposed methods outperform random selection and conventional selection techniques. The superior performance of the novel selection methods underscores their effectiveness in enhancing the Active Learning process for image classification tasks.

Protein-protein and protein-nucleic acid binding site prediction via interpretable hierarchical geometric deep learning

Identification of protein-protein and protein-nucleic acid binding sites provides insights into biological processes related to protein functions and technical guidance for disease diagnosis and drug design. However, accurate predictions by computational approaches remain highly challenging due to the limited knowledge of residue binding patterns. The binding pattern of a residue should be characterized by the spatial distribution of its neighboring residues combined with their physicochemical information interaction, which yet cannot be achieved by previous methods. Here, we design GraphRBF, a hierarchical geometric deep learning model to learn residue binding patterns from big data. To achieve it, GraphRBF describes physicochemical information interactions by designing an enhanced graph neural network and characterizes residue spatial distributions by introducing a prioritized radial basis function neural network. After training and testing, GraphRBF shows great improvements over existing state-of-the-art methods and strong interpretability of its learned representations. Applying GraphRBF to the SARS-CoV-2 omicron spike protein, it successfully identifies known epitopes of the protein. Moreover, it predicts multiple potential binding regions for new nanobodies or even new drugs with strong evidence. A user-friendly online server for GraphRBF is freely available at http://liulab.top/GraphRBF/server.

Prognostic Value of 18F-FDG PET Radiomics Features at Baseline in PET-Guided Consolidation Strategy in Diffuse Large B-Cell Lymphoma: A Machine-Learning Analysis from the GAINED Study

The results of the GA in Newly Diagnosed Diffuse Large B-Cell Lymphoma (GAINED) study demonstrated the success of an 18F-FDG PET-driven approach to allow early identification-for intensification therapy-of diffuse large B-cell lymphoma patients with a high risk of relapse. Besides, some works have reported the prognostic value of baseline PET radiomics features (RFs). This work investigated the added value of such biomarkers on survival of patients involved in the GAINED protocol. Methods: Conventional PET features and RFs were computed from 18F-FDG PET at baseline and extracted using different volume definitions (patient level, largest lesion, and hottest lesion). Clinical features and the consolidation treatment information were also considered in the model. Two machine-learning pipelines were trained with 80% of patients and tested on the remaining 20%. The training was repeated 100 times to highlight the test set variability. For the 2-y progression-free survival (PFS) outcome, the pipeline included a data augmentation and an elastic net logistic regression model. Results for different feature groups were compared using the mean area under the curve (AUC). For the survival outcome, the pipeline included a Cox univariate model to select the features. Then, the model included a split between high- and low-risk patients using the median of a regression score based on the coefficients of a penalized Cox multivariate approach. The log-rank test P values over the 100 loops were compared with a Wilcoxon signed-ranked test. Results: In total, 545 patients were included for the 2-y PFS classification and 561 for survival analysis. Clinical features alone, consolidation features alone, conventional PET features, and RFs extracted at patient level achieved an AUC of, respectively, 0.65 ± 0.07, 0.64 ± 0.06, 0.60 ± 0.07, and 0.62 ± 0.07 (0.62 ± 0.07 for the largest lesion and 0.54 ± 0.07 for the hottest). Combining clinical features with the consolidation features led to the best AUC (0.72 ± 0.06). Adding conventional PET features or RFs did not improve the results. For survival, the log-rank P values of the model involving clinical and consolidation features together were significantly smaller than all combined-feature groups (P < 0.007). Conclusion: The results showed that a concatenation of multimodal features coupled with a simple machine-learning model does not seem to improve the results in terms of 2-y PFS classification and PFS prediction for patient treated according to the GAINED protocol.

Seq-InSite: sequence supersedes structure for protein interaction site prediction

MOTIVATION

Proteins accomplish cellular functions by interacting with each other, which makes the prediction of interaction sites a fundamental problem. As experimental methods are expensive and time consuming, computational prediction of the interaction sites has been studied extensively. Structure-based programs are the most accurate, while the sequence-based ones are much more widely applicable, as the sequences available outnumber the structures by two orders of magnitude. Ideally, we would like a tool that has the quality of the former and the applicability of the latter.

RESULTS

We provide here the first solution that achieves these two goals. Our new sequence-based program, Seq-InSite, greatly surpasses the performance of sequence-based models, matching the quality of state-of-the-art structure-based predictors, thus effectively superseding the need for models requiring structure. The predictive power of Seq-InSite is illustrated using an analysis of evolutionary conservation for four protein sequences.

AVAILABILITY AND IMPLEMENTATION

Seq-InSite is freely available as a web server at http://seq-insite.csd.uwo.ca/ and as free source code, including trained models and all datasets used for training and testing, at https://github.com/lucian-ilie/Seq-InSite.

Odor dilution sorting as a clinical test of olfactory function: normative values and reliability data

Clinical assessment of an individual's sense of smell has gained prominence, but its resource-intensive nature necessitates the exploration of self-administered methods. In this study, a cohort of 68 patients with olfactory loss and 55 controls were assessed using a recently introduced olfactory test. This test involves sorting 2 odorants (eugenol and phenylethyl alcohol) in 5 dilutions according to odor intensity, with an average application time of 3.5 min. The sorting task score, calculated as the mean of Kendall's Tau between the assigned and true dilution orders and normalized to [0,1], identified a cutoff for anosmia at a score ≤ 0.7. This cutoff, which marks the 90th percentile of scores obtained with randomly ordered dilutions, had a balanced accuracy of 89% (78% to 97%) for detecting anosmia, comparable to traditional odor threshold assessments. Retest evaluations suggested a score difference of ±0.15 as a cutoff for clinically significant changes in olfactory function. In conclusion, the olfactory sorting test represents a simple, self-administered approach to the detection of anosmia or preserved olfactory function. With balanced accuracy similar to existing brief olfactory tests, this method offers a practical and user-friendly alternative for screening anosmia, addressing the need for resource-efficient assessments in clinical settings.

Machine-learning-guided Directed Evolution for AAV Capsid Engineering

Target gene delivery is crucial to gene therapy. Adeno-associated virus (AAV) has emerged as a primary gene therapy vector due to its broad host range, long-term expression, and low pathogenicity. However, AAV vectors have some limitations, such as immunogenicity and insufficient targeting. Designing or modifying capsids is a potential method of improving the efficacy of gene delivery, but hindered by weak biological basis of AAV, complexity of the capsids, and limitations of current screening methods. Artificial intelligence (AI), especially machine learning (ML), has great potential to accelerate and improve the optimization of capsid properties as well as decrease their development time and manufacturing costs. This review introduces the traditional methods of designing AAV capsids and the general steps of building a sequence-function ML model, highlights the applications of ML in the development workflow, and summarizes its advantages and challenges.

Predicting Impact Outcomes and Maximum Spreading of Drop Impact on Heated Nanostructures Using Machine Learning

Accurate prediction of droplet behavior upon impact on a heated nanostructured surface is vital for various industrial applications. In this study, we leverage multiple data-driven machine learning (ML) techniques to model the impact outcome and droplet spreading, employing existing experimental data. Our approach incorporates a comprehensive range of critical control parameters, such as the impact velocity (V), surface temperature (Ts), nanopillars' packing fraction (ϕ), and surface roughness (r). We obtain optimal results when utilizing the artificial neural network classification (ANNC) to construct a phase diagram that encompasses all of the experimental impact behaviors. Additionally, we utilize the support vector regression (SVR) method to model the maximum spreading factor (βmax) as a function of the Weber number (We), defined as the ratio of droplet kinetic to surface energy, and Ts for each surface combination. Consistent with previous experimental observations, our results illustrate that nanostructures not only introduce distinct impact behaviors, such as central jetting, but also influence the boundaries among the deposition, rebound, and splashing regimes within the phase diagram. An increase in ϕ at a constant r promotes deposition and spreading events, while increasing r at a constant ϕ results in enhanced heat transfer to promote the Leidenfrost effect for the rebound regime and a greater disturbance of the liquid lamella to trigger splashing. The SVR prediction reveals the existence of a We-number threshold governed by the nanostructure parameters. Beyond this threshold, the maximum spreading factor (βmax) of a spreading droplet becomes independent of the surface temperature (Ts) as We increases, suggesting that fluid properties are likely the dominating factors influencing the spreading dynamics in the extreme We range.

In Silico Screening of Natural Flavonoids against 3-Chymotrypsin-like Protease of SARS-CoV-2 Using Machine Learning and Molecular Modeling

The "Long-COVID syndrome" has posed significant challenges due to a lack of validated therapeutic options. We developed a novel multi-step virtual screening strategy to reliably identify inhibitors against 3-chymotrypsin-like protease of SARS-CoV-2 from abundant flavonoids, which represents a promising source of antiviral and immune-boosting nutrients. We identified 57 interacting residues as contributors to the protein-ligand binding pocket. Their energy interaction profiles constituted the input features for Machine Learning (ML) models. The consensus of 25 classifiers trained using various ML algorithms attained 93.9% accuracy and a 6.4% false-positive-rate. The consensus of 10 regression models for binding energy prediction also achieved a low root-mean-square error of 1.18 kcal/mol. We screened out 120 flavonoid hits first and retained 50 drug-like hits after predefined ADMET filtering to ensure bioavailability and safety profiles. Furthermore, molecular dynamics simulations prioritized nine bioactive flavonoids as promising anti-SARS-CoV-2 agents exhibiting both high structural stability (root-mean-square deviation < 5 Å for 218 ns) and low MM/PBSA binding free energy (<-6 kcal/mol). Among them, KB-2 (PubChem-CID, 14630497) and 9-O-Methylglyceofuran (PubChem-CID, 44257401) displayed excellent binding affinity and desirable pharmacokinetic capabilities. These compounds have great potential to serve as oral nutraceuticals with therapeutic and prophylactic properties as care strategies for patients with long-COVID syndrome.

A human-interpretable machine learning pipeline based on ultrasound to support leiomyosarcoma diagnosis

The preoperative evaluation of myometrial tumors is essential to avoid delayed treatment and to establish the appropriate surgical approach. Specifically, the differential diagnosis of leiomyosarcoma (LMS) is particularly challenging due to the overlapping of clinical, laboratory and ultrasound features between fibroids and LMS. In this work, we present a human-interpretable machine learning (ML) pipeline to support the preoperative differential diagnosis of LMS from leiomyomas, based on both clinical data and gynecological ultrasound assessment of 68 patients (8 with LMS diagnosis). The pipeline provides the following novel contributions: (i) end-users have been involved both in the definition of the ML tasks and in the evaluation of the overall approach; (ii) clinical specialists get a full understanding of both the decision-making mechanisms of the ML algorithms and the impact of the features on each automatic decision. Moreover, the proposed pipeline addresses some of the problems concerning both the imbalance of the two classes by analyzing and selecting the best combination of the synthetic oversampling strategy of the minority class and the classification algorithm among different choices, and the explainability of the features at global and local levels. The results show very high performance of the best strategy (AUC = 0.99, F1 = 0.87) and the strong and stable impact of two ultrasound-based features (i.e., tumor borders and consistency of the lesions). Furthermore, the SHAP algorithm was exploited to quantify the impact of the features at the local level and a specific module was developed to provide a template-based natural language (NL) translation of the explanations for enhancing their interpretability and fostering the use of ML in the clinical setting.

A fusion of VGG-16 and ViT models for improving bone tumor classification in computed tomography

Background and Objective

Bone tumors present significant challenges in orthopedic medicine due to variations in clinical treatment approaches for different tumor types, which includes benign, malignant, and intermediate cases. Convolutional Neural Networks (CNNs) have emerged as prominent models for tumor classification. However, their limited perception ability hinders the acquisition of global structural information, potentially affecting classification accuracy. To address this limitation, we propose an optimized deep learning algorithm for precise classification of diverse bone tumors.

Materials and Methods

Our dataset comprises 786 computed tomography (CT) images of bone tumors, featuring sections from two distinct bone species, namely the tibia and femur. Sourced from The Second Affiliated Hospital of Fujian Medical University, the dataset was meticulously preprocessed with noise reduction techniques. We introduce a novel fusion model, VGG16-ViT, leveraging the advantages of the VGG-16 network and the Vision Transformer (ViT) model. Specifically, we select 27 features from the third layer of VGG-16 and input them into the Vision Transformer encoder for comprehensive training. Furthermore, we evaluate the impact of secondary migration using CT images from Xiangya Hospital for validation.

Results

The proposed fusion model demonstrates notable improvements in classification performance. It effectively reduces the training time while achieving an impressive classification accuracy rate of 97.6%, marking a significant enhancement of 8% in sensitivity and specificity optimization. Furthermore, the investigation into secondary migration's effects on experimental outcomes across the three models reveals its potential to enhance system performance.

Conclusion

Our novel VGG-16 and Vision Transformer joint network exhibits robust classification performance on bone tumor datasets. The integration of these models enables precise and efficient classification, accommodating the diverse characteristics of different bone tumor types. This advancement holds great significance for the early detection and prognosis of bone tumor patients in the future.

Orthodontic craniofacial pattern diagnosis: cephalometric geometry and machine learning

Efficient and reliable diagnosis of craniofacial patterns is critical to orthodontic treatment. Although machine learning (ML) is time-saving and high-precision, prior knowledge should validate its reliability. This study proposed a craniofacial ML diagnostic workflow base on a cephalometric geometric model through clinical verification. A cephalometric geometric model was established to determine the landmark location by analyzing 408 X-ray lateral cephalograms. Through geometric information and feature engineering, nine supervised ML algorithms were conducted for sagittal and vertical skeleton patterns. After dimension reduction, plane decision boundary and landmark contribution contours were depicted to demonstrate the diagnostic consistency and the consistency with clinical norms. As a result, multi-layer perceptron achieved 97.56% accuracy for sagittal, while linear support vector machine reached 90.24% for the vertical. Sagittal diagnoses showed average superiority (91.60 ± 5.43)% over the vertical (82.25 ± 6.37)%, where discriminative algorithms exhibited more steady performance (93.20 ± 3.29)% than the generative (85.98 ± 9.48)%. Further, the Kruskal-Wallis H test was carried out to explore statistical differences in diagnoses. Though sagittal patterns had no statistical difference in diagnostic accuracy, the vertical showed significance. All aspects of the tests indicated that the proposed craniofacial ML workflow was highly consistent with clinical norms and could supplement practical diagnosis.

Fair patient model: Mitigating bias in the patient representation learned from the electronic health records

OBJECTIVE

To pre-train fair and unbiased patient representations from Electronic Health Records (EHRs) using a novel weighted loss function that reduces bias and improves fairness in deep representation learning models.

METHODS

We defined a new loss function, called weighted loss function, in the deep representation learning model to balance the importance of different groups of patients and features. We applied the proposed model, called Fair Patient Model (FPM), to a sample of 34,739 patients from the MIMIC-III dataset and learned patient representations for four clinical outcome prediction tasks.

RESULTS

FPM outperformed the baseline models in terms of three fairness metrics: demographic parity, equality of opportunity difference, and equalized odds ratio. FPM also achieved comparable predictive performance with the baselines, with an average accuracy of 0.7912. Feature analysis revealed that FPM captured more information from clinical features than the baselines.

CONCLUSION

FPM is a novel method to pre-train fair and unbiased patient representations from the EHR data using a weighted loss function. The learned representations can be used for various downstream tasks in healthcare and can be extended to other domains where fairness is important.

Climate change affects the distribution of diversity across marine food webs

Many studies predict shifts in species distributions and community size composition in response to climate change, yet few have demonstrated how these changes will be distributed across marine food webs. We use Bayesian Additive Regression Trees to model how climate change will affect the habitat suitability of marine fish species across a range of body sizes and belonging to different feeding guilds, each with different habitat and feeding requirements in the northeast Atlantic shelf seas. Contrasting effects of climate change are predicted for feeding guilds, with spatially extensive decreases in the species richness of consumers lower in the food web (planktivores) but increases for those higher up (piscivores). Changing spatial patterns in predator-prey mass ratios and fish species size composition are also predicted for feeding guilds and across the fish assemblage. In combination, these changes could influence nutrient uptake and transformation, transfer efficiency and food web stability, and thus profoundly alter ecosystem structure and functioning.

Machine Learning Insights: Predicting Hepatic Encephalopathy After TIPS Placement

PURPOSE

To develop and assess machine learning (ML) models' ability to predict post-procedural hepatic encephalopathy (HE) following transjugular intrahepatic portosystemic shunt (TIPS) placement.

MATERIALS AND METHODS

In this retrospective study, 327 patients who underwent TIPS for hepatic cirrhosis between 2005 and 2019 were analyzed. Thirty features (8 clinical, 10 laboratory, 12 procedural) were collected, and HE development regardless of severity was recorded one month follow-up. Univariate statistical analysis was performed with numeric and categoric data, as appropriate. Feature selection is used with a sequential feature selection model with fivefold cross-validation (CV). Three ML models were developed using support vector machine (SVM), logistic regression (LR) and CatBoost, algorithms. Performances were evaluated with nested fivefold-CV technique.

RESULTS

Post-procedural HE was observed in 105 (32%) patients. Patients with variceal bleeding (p = 0.008) and high post-porto-systemic pressure gradient (p = 0.004) had a significantly increased likelihood of developing HE. Also, patients having only one indication of bleeding or ascites were significantly unlikely to develop HE as well as Budd-Chiari disease (p = 0.03). The feature selection algorithm selected 7 features. Accuracy ratios for the SVM, LR and CatBoost, models were 74%, 75%, and 73%, with area under the curve (AUC) values of 0.82, 0.83, and 0.83, respectively.

CONCLUSION

ML models can aid identifying patients at risk of developing HE after TIPS placement, providing an additional tool for patient selection and management.

Synthetic Tabular Data Based on Generative Adversarial Networks in Health Care: Generation and Validation Using the Divide-and-Conquer Strategy

BACKGROUND

Synthetic data generation (SDG) based on generative adversarial networks (GANs) is used in health care, but research on preserving data with logical relationships with synthetic tabular data (STD) remains challenging. Filtering methods for SDG can lead to the loss of important information.

OBJECTIVE

This study proposed a divide-and-conquer (DC) method to generate STD based on the GAN algorithm, while preserving data with logical relationships.

METHODS

The proposed method was evaluated on data from the Korea Association for Lung Cancer Registry (KALC-R) and 2 benchmark data sets (breast cancer and diabetes). The DC-based SDG strategy comprises 3 steps: (1) We used 2 different partitioning methods (the class-specific criterion distinguished between survival and death groups, while the Cramer V criterion identified the highest correlation between columns in the original data); (2) the entire data set was divided into a number of subsets, which were then used as input for the conditional tabular generative adversarial network and the copula generative adversarial network to generate synthetic data; and (3) the generated synthetic data were consolidated into a single entity. For validation, we compared DC-based SDG and conditional sampling (CS)-based SDG through the performances of machine learning models. In addition, we generated imbalanced and balanced synthetic data for each of the 3 data sets and compared their performance using 4 classifiers: decision tree (DT), random forest (RF), Extreme Gradient Boosting (XGBoost), and light gradient-boosting machine (LGBM) models.

RESULTS

The synthetic data of the 3 diseases (non-small cell lung cancer [NSCLC], breast cancer, and diabetes) generated by our proposed model outperformed the 4 classifiers (DT, RF, XGBoost, and LGBM). The CS- versus DC-based model performances were compared using the mean area under the curve (SD) values: 74.87 (SD 0.77) versus 63.87 (SD 2.02) for NSCLC, 73.31 (SD 1.11) versus 67.96 (SD 2.15) for breast cancer, and 61.57 (SD 0.09) versus 60.08 (SD 0.17) for diabetes (DT); 85.61 (SD 0.29) versus 79.01 (SD 1.20) for NSCLC, 78.05 (SD 1.59) versus 73.48 (SD 4.73) for breast cancer, and 59.98 (SD 0.24) versus 58.55 (SD 0.17) for diabetes (RF); 85.20 (SD 0.82) versus 76.42 (SD 0.93) for NSCLC, 77.86 (SD 2.27) versus 68.32 (SD 2.37) for breast cancer, and 60.18 (SD 0.20) versus 58.98 (SD 0.29) for diabetes (XGBoost); and 85.14 (SD 0.77) versus 77.62 (SD 1.85) for NSCLC, 78.16 (SD 1.52) versus 70.02 (SD 2.17) for breast cancer, and 61.75 (SD 0.13) versus 61.12 (SD 0.23) for diabetes (LGBM). In addition, we found that balanced synthetic data performed better.

CONCLUSIONS

This study is the first attempt to generate and validate STD based on a DC approach and shows improved performance using STD. The necessity for balanced SDG was also demonstrated.

Machine Learning Approaches Identify Chemical Features for Stage-Specific Antimalarial Compounds

Efficacy data from diverse chemical libraries, screened against the various stages of the malaria parasite Plasmodium falciparum, including asexual blood stage (ABS) parasites and transmissible gametocytes, serve as a valuable reservoir of information on the chemical space of compounds that are either active (or not) against the parasite. We postulated that this data can be mined to define chemical features associated with the sole ABS activity and/or those that provide additional life cycle activity profiles like gametocytocidal activity. Additionally, this information could provide chemical features associated with inactive compounds, which could eliminate any future unnecessary screening of similar chemical analogs. Therefore, we aimed to use machine learning to identify the chemical space associated with stage-specific antimalarial activity. We collected data from various chemical libraries that were screened against the asexual (126 374 compounds) and sexual (gametocyte) stages of the parasite (93 941 compounds), calculated the compounds' molecular fingerprints, and trained machine learning models to recognize stage-specific active and inactive compounds. We were able to build several models that predict compound activity against ABS and dual activity against ABS and gametocytes, with Support Vector Machines (SVM) showing superior abilities with high recall (90 and 66%) and low false-positive predictions (15 and 1%). This allowed the identification of chemical features enriched in active and inactive populations, an important outcome that could be mined for essential chemical features to streamline hit-to-lead optimization strategies of antimalarial candidates. The predictive capabilities of the models held true in diverse chemical spaces, indicating that the ML models are therefore robust and can serve as a prioritization tool to drive and guide phenotypic screening and medicinal chemistry programs.

Predicting Thalassemia Using Feature Selection Techniques: A Comparative Analysis

Thalassemia represents one of the most common genetic disorders worldwide, characterized by defects in hemoglobin synthesis. The affected individuals suffer from malfunctioning of one or more of the four globin genes, leading to chronic hemolytic anemia, an imbalance in the hemoglobin chain ratio, iron overload, and ineffective erythropoiesis. Despite the challenges posed by this condition, recent years have witnessed significant advancements in diagnosis, therapy, and transfusion support, significantly improving the prognosis for thalassemia patients. This research empirically evaluates the efficacy of models constructed using classification methods and explores the effectiveness of relevant features that are derived using various machine-learning techniques. Five feature selection approaches, namely Chi-Square (χ2), Exploratory Factor Score (EFS), tree-based Recursive Feature Elimination (RFE), gradient-based RFE, and Linear Regression Coefficient, were employed to determine the optimal feature set. Nine classifiers, namely K-Nearest Neighbors (KNN), Decision Trees (DT), Gradient Boosting Classifier (GBC), Linear Regression (LR), AdaBoost, Extreme Gradient Boosting (XGB), Random Forest (RF), Light Gradient Boosting Machine (LGBM), and Support Vector Machine (SVM), were utilized to evaluate the performance. The χ2 method achieved accuracy, registering 91.56% precision, 91.04% recall, and 92.65% f-score when aligned with the LR classifier. Moreover, the results underscore that amalgamating over-sampling with Synthetic Minority Over-sampling Technique (SMOTE), RFE, and 10-fold cross-validation markedly elevates the detection accuracy for αT patients. Notably, the Gradient Boosting Classifier (GBC) achieves 93.46% accuracy, 93.89% recall, and 92.72% F1 score.

Classification Algorithms for Liver Epidemic Identification

Situated in the upper right region of the abdomen, beneath the diaphragm and above the stomach, lies the liver. It is a crucial organ essential for the proper functioning of the body.  The principal tasks are to eliminate generated waste produced by our organs, and digestive food and preserve vitamins and energy materials. It performs many important functions in the body, it regulates the balance of hormones in the body filtering and removing bacteria, viruses, and other harmful substances from the blood. In certain dire circumstances, the outcome can unfortunately result in fatality. There exist numerous classifications of liver diseases, based on their causes or distinguishing characteristics. Some common categories of liver disease include Viral hepatitis, Autoimmune liver disease, Metabolic liver disease, Alcohol-related liver disease, Non-alcoholic fatty liver disease, Genetic liver disease, Drug-induced liver injury, Biliary tract disorders. Machine learning algorithms can help identify patterns and risk factors that may be difficult for humans to detect. With this clinicians can enable early diagnosis of diseases, leading to better treatment outcomes and improved patient care. In this research work, different types of machine learning methods are implemented and compared in terms of performance metrics to identify whether a person effected or not. The algorithms used here for predicting liver patients are Random Forest classifier, K-nearest neighbor, XGBoost, Decision tree, Logistic Regression, support vector machine, Extra Trees Classifier. The experimental results showed that the accuracy of various machine learning models-Random Forest classifier-67.4%, K-nearest neighbor-54.8%, XGBoost-72%, Decision tree-65.1%, Logistic Regression-68.0%, support vector machine-65.1%, Extra Trees Classifier-70.2% after applying Synthetic Minority Over-sampling technique.

Explainable machine learning for the prediction and assessment of complex drought impacts

Drought is a common and costly natural disaster with broad social, economic, and environmental impacts. Machine learning (ML) has been widely applied in scientific research because of its outstanding performance on predictive tasks. However, for practical applications like disaster monitoring and assessment, the cost of the models failure, especially false negative predictions, might significantly affect society. Stakeholders are not satisfied with or do not "trust" the predictions from a so-called black box. The explainability of ML models becomes progressively crucial in studying drought and its impacts. In this work, we propose an explainable ML pipeline using the XGBoost model and SHAP model based on a comprehensive database of drought impacts in the U.S. The XGBoost models significantly outperformed the baseline models in predicting the occurrence of multi-dimensional drought impacts derived from the text-based Drought Impact Reporter, attaining an average F2 score of 0.883 at the national level and 0.942 at the state level. The interpretation of the models at the state scale indicates that the Standardized Precipitation Index (SPI) and Standardized Temperature Index (STI) contribute significantly to predicting multi-dimensional drought impacts. The time scalar, importance, and relationships of the SPI and STI vary depending on the types of drought impacts and locations. The patterns between the SPI variables and drought impacts indicated by the SHAP values reveal an expected relationship in which negative SPI values positively contribute to complex drought impacts. The explainability based on the SPI variables improves the trustworthiness of the XGBoost models. Overall, this study reveals promising results in accurately predicting complex drought impacts and rendering the relationships between the impacts and indicators more interpretable. This study also reveals the potential of utilizing explainable ML for the general social good to help stakeholders better understand the multi-dimensional drought impacts at the regional level and motivate appropriate responses.

An explainable artificial intelligence framework for risk prediction of COPD in smokers

BACKGROUND

Since the inconspicuous nature of early signs associated with Chronic Obstructive Pulmonary Disease (COPD), individuals often remain unidentified, leading to suboptimal opportunities for timely prevention and treatment. The purpose of this study was to create an explainable artificial intelligence framework combining data preprocessing methods, machine learning methods, and model interpretability methods to identify people at high risk of COPD in the smoking population and to provide a reasonable interpretation of model predictions.

METHODS

The data comprised questionnaire information, physical examination data and results of pulmonary function tests before and after bronchodilatation. First, the factorial analysis for mixed data (FAMD), Boruta and NRSBoundary-SMOTE resampling methods were used to solve the missing data, high dimensionality and category imbalance problems. Then, seven classification models (CatBoost, NGBoost, XGBoost, LightGBM, random forest, SVM and logistic regression) were applied to model the risk level, and the best machine learning (ML) model's decisions were explained using the Shapley additive explanations (SHAP) method and partial dependence plot (PDP).

RESULTS

In the smoking population, age and 14 other variables were significant factors for predicting COPD. The CatBoost, random forest, and logistic regression models performed reasonably well in unbalanced datasets. CatBoost with NRSBoundary-SMOTE had the best classification performance in balanced datasets when composite indicators (the AUC, F1-score, and G-mean) were used as model comparison criteria. Age, COPD Assessment Test (CAT) score, gross annual income, body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), anhelation, respiratory disease, central obesity, use of polluting fuel for household heating, region, use of polluting fuel for household cooking, and wheezing were important factors for predicting COPD in the smoking population.

CONCLUSION

This study combined feature screening methods, unbalanced data processing methods, and advanced machine learning methods to enable early identification of COPD risk groups in the smoking population. COPD risk factors in the smoking population were identified using SHAP and PDP, with the goal of providing theoretical support for targeted screening strategies and smoking population self-management strategies.

Artificial intelligence evaluation of focused assessment with sonography in trauma

BACKGROUND

The focused assessment with sonography in trauma (FAST) is a widely used imaging modality to identify the location of life-threatening hemorrhage in a hemodynamically unstable trauma patient. This study evaluates the role of artificial intelligence in interpretation of the FAST examination abdominal views, as it pertains to adequacy of the view and accuracy of fluid survey positivity.

METHODS

Focused assessment with sonography for trauma examination images from 2015 to 2022, from trauma activations, were acquired from a quaternary care level 1 trauma center with more than 3,500 adult trauma evaluations, annually. Images pertaining to the right upper quadrant and left upper quadrant views were obtained and read by a surgeon or radiologist. Positivity was defined as fluid present in the hepatorenal or splenorenal fossa, while adequacy was defined by the presence of both the liver and kidney or the spleen and kidney for the right upper quadrant or left upper quadrant views, respectively. Four convolutional neural network architecture models (DenseNet121, InceptionV3, ResNet50, Vgg11bn) were evaluated.

RESULTS

A total of 6,608 images, representing 109 cases were included for analysis within the "adequate" and "positive" data sets. The models relayed 88.7% accuracy, 83.3% sensitivity, and 93.6% specificity for the adequate test cohort, while the positive cohort conferred 98.0% accuracy, 89.6% sensitivity, and 100.0% specificity against similar models. Augmentation improved the accuracy and sensitivity of the positive models to 95.1% accurate and 94.0% sensitive. DenseNet121 demonstrated the best accuracy across tasks.

CONCLUSION

Artificial intelligence can detect positivity and adequacy of FAST examinations with 94% and 97% accuracy, aiding in the standardization of care delivery with minimal expert clinician input. Artificial intelligence is a feasible modality to improve patient care imaging interpretation accuracy and should be pursued as a point-of-care clinical decision-making tool.

LEVEL OF EVIDENCE

Diagnostic Test/Criteria; Level III.

Drop Loss for Person Attribute Recognition With Imbalanced Noisy-Labeled Samples

Person attribute recognition (PAR) aims to simultaneously predict multiple attributes of a person. Existing deep learning-based PAR methods have achieved impressive performance. Unfortunately, these methods usually ignore the fact that different attributes have an imbalance in the number of noisy-labeled samples in the PAR training datasets, thus leading to suboptimal performance. To address the above problem of imbalanced noisy-labeled samples, we propose a novel and effective loss called drop loss for PAR. In the drop loss, the attributes are treated differently in an easy-to-hard way. In particular, the noisy-labeled candidates, which are identified according to their gradient norms, are dropped with a higher drop rate for the harder attribute. Such a manner adaptively alleviates the adverse effect of imbalanced noisy-labeled samples on model learning. To illustrate the effectiveness of the proposed loss, we train a simple ResNet-50 model based on the drop loss and term it DropNet. Experimental results on two representative PAR tasks (including facial attribute recognition and pedestrian attribute recognition) demonstrate that the proposed DropNet achieves comparable or better performance in terms of both balanced accuracy and classification accuracy over several state-of-the-art PAR methods.

Target-oriented deep learning-based image registration with individualized test-time adaptation

BACKGROUND

A classic approach in medical image registration is to formulate an optimization problem based on the image pair of interest, and seek a deformation vector field (DVF) to minimize the corresponding objective, often iteratively. It has a clear focus on the targeted pair, but is typically slow. In contrast, more recent deep-learning-based registration offers a much faster alternative and can benefit from data-driven regularization. However, learning is a process to "fit" the training cohort, whose image or motion characteristics or both may differ from the pair of images to be tested, which is the ultimate goal of registration. Therefore, generalization gap poses a high risk with direct inference alone.

PURPOSE

In this study, we propose an individualized adaptation to improve test sample targeting, to achieve a synergy of efficiency and performance in registration.

METHODS

Using a previously developed network with an integrated motion representation prior module as the implementation backbone, we propose to adapt the trained registration network further for image pairs at test time to optimize the individualized performance. The adaptation method was tested against various characteristics shifts caused by cross-protocol, cross-platform, and cross-modality, with test evaluation performed on lung CBCT, cardiac MRI, and lung MRI, respectively.

RESULTS

Landmark-based registration errors and motion-compensated image enhancement results demonstrated significantly improved test registration performance from our method, compared to tuned classic B-spline registration and network solutions without adaptation.

CONCLUSIONS

We have developed a method to synergistically combine the effectiveness of pre-trained deep network and the target-centric perspective of optimization-based registration to improve performance on individual test data.

Overcoming the Challenges in the Development and Implementation of Artificial Intelligence in Radiology: A Comprehensive Review of Solutions Beyond Supervised Learning

Artificial intelligence (AI) in radiology is a rapidly developing field with several prospective clinical studies demonstrating its benefits in clinical practice. In 2022, the Korean Society of Radiology held a forum to discuss the challenges and drawbacks in AI development and implementation. Various barriers hinder the successful application and widespread adoption of AI in radiology, such as limited annotated data, data privacy and security, data heterogeneity, imbalanced data, model interpretability, overfitting, and integration with clinical workflows. In this review, some of the various possible solutions to these challenges are presented and discussed; these include training with longitudinal and multimodal datasets, dense training with multitask learning and multimodal learning, self-supervised contrastive learning, various image modifications and syntheses using generative models, explainable AI, causal learning, federated learning with large data models, and digital twins.

PerSEveML: A Web-Based Tool to Identify Persistent Biomarker Structure for Rare Events Using Integrative Machine Learning Approach

Omics datasets often pose a computational challenge due to their high dimensionality, large size, and non-linear structures. Analyzing these datasets becomes especially daunting in the presence of rare events. Machine learning (ML) methods have gained traction for analyzing rare events, yet there remains a limited exploration of bioinformatics tools that integrate ML techniques to comprehend the underlying biology. Expanding upon our previously developed computational framework of an integrative machine learning approach1, we introduce PerSEveML, an interactive web-based that uses crowd-sourced intelligence to predict rare events and determine feature selection structures. PerSEveML provides a comprehensive overview of the integrative approach through evaluation metrics that help users understand the contribution of individual ML methods to the prediction process. Additionally, PerSEveML calculates entropy and rank scores, which visually organize input features into a persistent structure of selected, unselected, and fluctuating categories that help researchers uncover meaningful hypotheses regarding the underlying biology. We have evaluated PerSEveML on three diverse biologically complex data sets with extremely rare events from small to large scale and have demonstrated its ability to generate valid hypotheses. PerSEveML is available at https://biostats-shinyr.kumc.edu/PerSEveML/ and https://github.com/sreejatadutta/PerSEveML.

Augmentation strategies for an imbalanced learning problem on a novel COVID-19 severity dataset

Since the beginning of the COVID-19 pandemic, many different machine learning models have been developed to detect and verify COVID-19 pneumonia based on chest X-ray images. Although promising, binary models have only limited implications for medical treatment, whereas the prediction of disease severity suggests more suitable and specific treatment options. In this study, we publish severity scores for the 2358 COVID-19 positive images in the COVIDx8B dataset, creating one of the largest collections of publicly available COVID-19 severity data. Furthermore, we train and evaluate deep learning models on the newly created dataset to provide a first benchmark for the severity classification task. One of the main challenges of this dataset is the skewed class distribution, resulting in undesirable model performance for the most severe cases. We therefore propose and examine different augmentation strategies, specifically targeting majority and minority classes. Our augmentation strategies show significant improvements in precision and recall values for the rare and most severe cases. While the models might not yet fulfill medical requirements, they serve as an appropriate starting point for further research with the proposed dataset to optimize clinical resource allocation and treatment.

Using Proteomics Data to Identify Personalized Treatments in Multiple Myeloma: A Machine Learning Approach

This paper describes a machine learning (ML) decision support system to provide a list of chemotherapeutics that individual multiple myeloma (MM) patients are sensitive/resistant to, based on their proteomic profile. The methodology used in this study involved understanding the parameter space and selecting the dominant features (proteomics data), identifying patterns of proteomic profiles and their association to the recommended treatments, and defining the decision support system of personalized treatment as a classification problem. During the data analysis, we compared several ML algorithms, such as linear regression, Random Forest, and support vector machines, to classify patients as sensitive/resistant to therapeutics. A further analysis examined data-balancing techniques that emerged due to the small cohort size. The results suggest that utilizing proteomics data is a promising approach for identifying effective treatment options for patients with MM (reaching on average an accuracy of 81%). Although this pilot study was limited by the small patient cohort (39 patients), which restricted the training and validation of the explored ML solutions to identify complex associations between proteins, it holds great promise for developing personalized anti-MM treatments using ML approaches.

2S-BUSGAN: A Novel Generative Adversarial Network for Realistic Breast Ultrasound Image with Corresponding Tumor Contour Based on Small Datasets

Deep learning (DL) models in breast ultrasound (BUS) image analysis face challenges with data imbalance and limited atypical tumor samples. Generative Adversarial Networks (GAN) address these challenges by providing efficient data augmentation for small datasets. However, current GAN approaches fail to capture the structural features of BUS and generated images lack structural legitimacy and are unrealistic. Furthermore, generated images require manual annotation for different downstream tasks before they can be used. Therefore, we propose a two-stage GAN framework, 2s-BUSGAN, for generating annotated BUS images. It consists of the Mask Generation Stage (MGS) and the Image Generation Stage (IGS), generating benign and malignant BUS images using corresponding tumor contours. Moreover, we employ a Feature-Matching Loss (FML) to enhance the quality of generated images and utilize a Differential Augmentation Module (DAM) to improve GAN performance on small datasets. We conduct experiments on two datasets, BUSI and Collected. Moreover, results indicate that the quality of generated images is improved compared with traditional GAN methods. Additionally, our generated images underwent evaluation by ultrasound experts, demonstrating the possibility of deceiving doctors. A comparative evaluation showed that our method also outperforms traditional GAN methods when applied to training segmentation and classification models. Our method achieved a classification accuracy of 69% and 85.7% on two datasets, respectively, which is about 3% and 2% higher than that of the traditional augmentation model. The segmentation model trained using the 2s-BUSGAN augmented datasets achieved DICE scores of 75% and 73% on the two datasets, respectively, which were higher than the traditional augmentation methods. Our research tackles imbalanced and limited BUS image data challenges. Our 2s-BUSGAN augmentation method holds potential for enhancing deep learning model performance in the field.

Data Augmentation Techniques for Machine Learning Applied to Optical Spectroscopy Datasets in Agrifood Applications: A Comprehensive Review

Machine learning (ML) and deep learning (DL) have achieved great success in different tasks. These include computer vision, image segmentation, natural language processing, predicting classification, evaluating time series, and predicting values based on a series of variables. As artificial intelligence progresses, new techniques are being applied to areas like optical spectroscopy and its uses in specific fields, such as the agrifood industry. The performance of ML and DL techniques generally improves with the amount of data available. However, it is not always possible to obtain all the necessary data for creating a robust dataset. In the particular case of agrifood applications, dataset collection is generally constrained to specific periods. Weather conditions can also reduce the possibility to cover the entire range of classifications with the consequent generation of imbalanced datasets. To address this issue, data augmentation (DA) techniques are employed to expand the dataset by adding slightly modified copies of existing data. This leads to a dataset that includes values from laboratory tests, as well as a collection of synthetic data based on the real data. This review work will present the application of DA techniques to optical spectroscopy datasets obtained from real agrifood industry applications. The reviewed methods will describe the use of simple DA techniques, such as duplicating samples with slight changes, as well as the utilization of more complex algorithms based on deep learning generative adversarial networks (GANs), and semi-supervised generative adversarial networks (SGANs).

Prediction of CRISPR-Cas9 off-target activities with mismatches and indels based on hybrid neural network

The CRISPR/Cas9 system has significantly advanced the field of gene editing, yet its clinical application is constrained by the considerable challenge of off-target effects. Although numerous deep learning models for off-target prediction have been proposed, most struggle to effectively extract the nuanced features of guide RNA (gRNA) and DNA sequence pairs and to mitigate information loss during data transmission within the model. To address these limitations, we introduce a novel Hybrid Neural Network (HNN) model that employs a parallelized network structure to fully extract pertinent features from different positions and types of bases in the sequence to minimize information loss. Notably, this study marks the first application of word embedding techniques to extract information from sequence pairs that contain insertions and deletions (Indels). Comprehensive evaluation across diverse datasets indicates that our proposed model outperforms existing state-of-the-art prediction methods in off-target prediction. The datasets and source codes supporting this study can be found at https://github.com/Yang-k955/CRISPR-HW.

Multivariate linear mixture models for the prediction of febrile seizure risk and recurrence: a prospective case-control study

Our goal was to identify highly accurate empirical models for the prediction of the risk of febrile seizure (FS) and FS recurrence. In a prospective, three-arm, case-control study, we enrolled 162 children (age 25.8 ± 17.1 months old, 71 females). Participants formed one case group (patients with FS) and two control groups (febrile patients without seizures and healthy controls). The impact of blood iron status, peak body temperature, and participants' demographics on FS risk and recurrence was investigated with univariate and multivariate statistics. Serum iron concentration, iron saturation, and unsaturated iron-binding capacity differed between the three investigated groups (pFWE < 0.05). These serum analytes were key variables in the design of novel multivariate linear mixture models. The models classified FS risk with higher accuracy than univariate approaches. The designed bi-linear classifier achieved a sensitivity/specificity of 82%/89% and was closest to the gold-standard classifier. A multivariate model assessing FS recurrence provided a difference (pFWE < 0.05) with a separating sensitivity/specificity of 72%/69%. Iron deficiency, height percentile, and age were significant FS risk factors. In addition, height percentile and hemoglobin concentration were linked to FS recurrence. Novel multivariate models utilizing blood iron status and demographic variables predicted FS risk and recurrence among infants and young children with fever.

Urinary Metabolic Distinction of Niemann-Pick Class 1 Disease through the Use of Subgroup Discovery

In this investigation, we outline the applications of a data mining technique known as Subgroup Discovery (SD) to the analysis of a sample size-limited metabolomics-based dataset. The SD technique utilized a supervised learning strategy, which lies midway between classificational and descriptive criteria, in which given the descriptive property of a dataset (i.e., the response target variable of interest), the primary objective was to discover subgroups with behaviours that are distinguishable from those of the complete set (albeit with a differential statistical distribution). These approaches have, for the first time, been successfully employed for the analysis of aromatic metabolite patterns within an NMR-based urinary dataset collected from a small cohort of patients with the lysosomal storage disorder Niemann-Pick class 1 (NPC1) disease (n = 12) and utilized to distinguish these from a larger number of heterozygous (parental) control participants. These subgroup discovery strategies discovered two different NPC1 disease-specific metabolically sequential rules which permitted the reliable identification of NPC1 patients; the first of these involved 'normal' (intermediate) urinary concentrations of xanthurenate, 4-aminobenzoate, hippurate and quinaldate, and disease-downregulated levels of nicotinate and trigonelline, whereas the second comprised 'normal' 4-aminobenzoate, indoxyl sulphate, hippurate, 3-methylhistidine and quinaldate concentrations, and again downregulated nicotinate and trigonelline levels. Correspondingly, a series of five subgroup rules were generated for the heterozygous carrier control group, and 'biomarkers' featured in these included low histidine, 1-methylnicotinamide and 4-aminobenzoate concentrations, together with 'normal' levels of hippurate, hypoxanthine, quinolinate and hypoxanthine. These significant disease group-specific rules were consistent with imbalances in the combined tryptophan-nicotinamide, tryptophan, kynurenine and tyrosine metabolic pathways, along with dysregulations in those featuring histidine, 3-methylhistidine and 4-hydroxybenzoate. In principle, the novel subgroup discovery approach employed here should also be readily applicable to solving metabolomics-type problems of this nature which feature rare disease classification groupings with only limited patient participant and sample sizes available.

On Scene Injury Severity Prediction (OSISP) model for trauma developed using the Swedish Trauma Registry

BACKGROUND

Providing optimal care for trauma, the leading cause of death for young adults, remains a challenge e.g., due to field triage limitations in assessing a patient's condition and deciding on transport destination. Data-driven On Scene Injury Severity Prediction (OSISP) models for motor vehicle crashes have shown potential for providing real-time decision support. The objective of this study is therefore to evaluate if an Artificial Intelligence (AI) based clinical decision support system can identify severely injured trauma patients in the prehospital setting.

METHODS

The Swedish Trauma Registry was used to train and validate five models - Logistic Regression, Random Forest, XGBoost, Support Vector Machine and Artificial Neural Network - in a stratified 10-fold cross validation setting and hold-out analysis. The models performed binary classification of the New Injury Severity Score and were evaluated using accuracy metrics, area under the receiver operating characteristic curve (AUC) and Precision-Recall curve (AUCPR), and under- and overtriage rates.

RESULTS

There were 75,602 registrations between 2013-2020 and 47,357 (62.6%) remained after eligibility criteria were applied. Models were based on 21 predictors, including injury location. From the clinical outcome, about 40% of patients were undertriaged and 46% were overtriaged. Models demonstrated potential for improved triaging and yielded AUC between 0.80-0.89 and AUCPR between 0.43-0.62.

CONCLUSIONS

AI based OSISP models have potential to provide support during assessment of injury severity. The findings may be used for developing tools to complement field triage protocols, with potential to improve prehospital trauma care and thereby reduce morbidity and mortality for a large patient population.

Deep learning and clustering approaches for dental implant size classification based on periapical radiographs

This study investigated two artificial intelligence (AI) methods for automatically classifying dental implant diameter and length based on periapical radiographs. The first method, deep learning (DL), involved utilizing the pre-trained VGG16 model and adjusting the fine-tuning degree to analyze image data obtained from periapical radiographs. The second method, clustering analysis, was accomplished by analyzing the implant-specific feature vector derived from three key points coordinates of the dental implant using the k-means++ algorithm and adjusting the weight of the feature vector. DL and clustering model classified dental implant size into nine groups. The performance metrics of AI models were accuracy, sensitivity, specificity, F1-score, positive predictive value, negative predictive value, and area under the receiver operating characteristic curve (AUC-ROC). The final DL model yielded performances above 0.994, 0.950, 0.994, 0.974, 0.952, 0.994, and 0.975, respectively, and the final clustering model yielded performances above 0.983, 0.900, 0.988, 0.923, 0.909, 0.988, and 0.947, respectively. When comparing the AI model before tuning and the final AI model, statistically significant performance improvements were observed in six out of nine groups for DL models and four out of nine groups for clustering models based on AUC-ROC. Two AI models showed reliable classification performances. For clinical applications, AI models require validation on various multicenter data.

Opportunities and challenges of supervised machine learning for the classification of motor evoked potentials according to muscles

BACKGROUND

Even for an experienced neurophysiologist, it is challenging to look at a single graph of an unlabeled motor evoked potential (MEP) and identify the corresponding muscle. We demonstrate that supervised machine learning (ML) can successfully perform this task.

METHODS

Intraoperative MEP data from supratentorial surgery on 36 patients was included for the classification task with 4 muscles: Extensor digitorum (EXT), abductor pollicis brevis (APB), tibialis anterior (TA) and abductor hallucis (AH). Three different supervised ML classifiers (random forest (RF), k-nearest neighbors (kNN) and logistic regression (LogReg)) were trained and tested on either raw or compressed data. Patient data was classified considering either all 4 muscles simultaneously, 2 muscles within the same extremity (EXT versus APB), or 2 muscles from different extremities (EXT versus TA).

RESULTS

In all cases, RF classifiers performed best and kNN second best. The highest performances were achieved on raw data (4 muscles 83%, EXT versus APB 89%, EXT versus TA 97% accuracy).

CONCLUSIONS

Standard ML methods show surprisingly high performance on a classification task with intraoperative MEP signals. This study illustrates the power and challenges of standard ML algorithms when handling intraoperative signals and may lead to intraoperative safety improvements.

Toward practical machine-learning-based diagnosis for drug-naïve women with major depressive disorder using EEG channel reduction approach

BACKGROUND

A machine-learning-based computer-aided diagnosis (CAD) system can complement the traditional diagnostic error for major depressive disorder (MDD) using trait-like neurophysiological biomarkers. Previous studies have shown that the CAD system has the potential to differentiate between female MDD patients and healthy controls. The aim of this study was to develop a practically useful resting-state electroencephalography (EEG)-based CAD system to assist in the diagnosis of drug-naïve female MDD patients by considering both the drug and gender effects. In addition, the feasibility of the practical use of the resting-state EEG-based CAD system was evaluated using a channel reduction approach.

METHODS

Eyes-closed, resting-state EEG data were recorded from 49 drug-naïve female MDD patients and 49 sex-matched healthy controls. Six different EEG feature sets were extracted: power spectrum densities (PSDs), phase-locking values (PLVs), and network indices for both sensor- and source-level, and four different EEG channel montages (62, 30, 19, and 10-channels) were designed to investigate the channel reduction effects in terms of classification performance.

RESULTS

The classification performances for each feature set were evaluated using a support vector machine with leave-one-out cross-validation. The optimum classification performance was achieved when using sensor-level PLVs (accuracy: 83.67 % and area under curve: 0.92). Moreover, the classification performance was maintained until the number of EEG channels was reduced to 19 (over 80 % accuracy).

CONCLUSION

We demonstrated the promising potential of sensor-level PLVs as diagnostic features when developing a resting-state EEG-based CAD system for the diagnosis of drug-naïve female MDD patients and verified the feasibility of the practical use of the developed resting-state EEG-based CAD system using the channel reduction approach.

Topology-based radiomic features for prediction of parotid gland cancer malignancy grade in magnetic resonance images

PURPOSE

The malignancy grades of parotid gland cancer (PGC) have been assessed for a decision of treatment policies. Therefore, we have investigated the feasibility of topology-based radiomic features for the prediction of parotid gland cancer (PGC) malignancy grade in magnetic resonance (MR) images.

MATERIALS AND METHODS

Two-dimensional T1- and T2-weighted MR images of 39 patients with PGC were selected for this study. Imaging properties of PGC can be quantified using the topology, which could be useful for assessing the number of the k-dimensional holes or heterogeneity in PGC regions using invariants of the Betti numbers. Radiomic signatures were constructed from 41,472 features obtained after a harmonization using an elastic net model. PGC patients were stratified using a logistic classification into low/intermediate- and high-grade malignancy groups. The training data were increased by four times to avoid the overfitting problem using a synthetic minority oversampling technique. The proposed approach was assessed using a 4-fold cross-validation test.

RESULTS

The highest accuracy of the proposed approach was 0.975 for the validation cases, whereas that of the conventional approach was 0.694.

CONCLUSION

This study indicated that topology-based radiomic features could be feasible for the noninvasive prediction of the malignancy grade of PGCs.

Accurate prediction of calving in dairy cows by applying feature engineering and machine learning

Prediction of calving is key to dairy cow management. Current trends of increasing herd sizes globally can directly impact the time that farmers spend monitoring individual animals. Automated monitoring on behavioural and physiological changes prior to parturition can be used to develop machine learning solutions for calving prediction. In this study, we developed a machine learning algorithm for the prediction of calving in dairy cows. We demonstrated that temperature and activity index information retrieved from a commercial reticuloruminal bolus sensor can accurately predict calving from 1-day to 5-days in advance. The best prediction solution using data from 82 dairy cows, achieved up to 87.81 % in accuracy, 92.99 % in specificity, 75.84 % in sensitivity, 82.99 % in positive predictive value (PPV), 78.85 % in F-score, and 90.02 % in negative predictive value (NPV) on the test dataset when using information from 2-days in advance and all the subsets of feature characteristics (temperature + drinking + activity). The performance only decreased by 2.45 % points in accuracy, 0.74 % points in specificity, 6.41 % points in sensitivity, 2.45 % points in positive predictive value, 4.91 % points in F-score, and 2.44 % points in negative predictive value on the test dataset when using all feature characteristics and 5-days in advance information compared to using all features and information from 2-days in advance. Full evaluation of the performance of the prediction showed an improvement when using all the different subsets of feature characteristics together (temperature, activity, and drinking) compared to using temperature features only. When adding activity and drinking to the subset of temperature features, an average increase of 2.70, 1.52, 5.40, 4.39, 5.02, 2.13 % points in accuracy, specificity, sensitivity, PPV, F-score, and NPV, respectively, was obtained. Notably, evaluation of feature importance (i.e., relative weight of any given feature in relation to model prediction) showed that 3-5 (depending on the selected days in advance model) of the top ten features were derived from drinking behaviour, showing the relevance that this behaviour can have in the prediction of calving. This algorithm can provide a useful tool for automated calving prediction in dairy cows which has potential for improvement of health, welfare, and productivity in the dairy industry.

Chronic Degradation of Seasonally Dry Tropical Forests Increases the Incidence of Genotoxicity in Birds

Multiple studies have shown that exposure to pollutants can increase genotoxic damage in different taxa. However, to our knowledge, the effects of environmental stress have been explored little. In certain stressful ecosystems, such as seasonally dry tropical forests, the combined effects of anthropogenic activities and ongoing global changes can cause an increase in environmental stresses, in turn, may trigger physiological and genetic effects on biodiversity. The present aims to assess changes in the prevalence of genotoxic damage in birds within three states of forest degradation in the Tumbesian Region of Western Ecuador. We used blood samples from 50 bird species to determine the frequency of micronucleus and nuclear abnormalities in erythrocytes. Our results revealed a significant impact of forest degradation on the occurrence probability of micronucleus and nuclear abnormalities at the community level. Localities with higher levels of degradation exhibited higher levels of abnormalities. However, when analyzing the dominant species, we found contrasting responses. While Lepidocolaptes souleyetii showed a reduction in the proportion of nuclear abnormalities from the natural to shrub-dominated localities Troglodytes aedon and Polioptila plumbea showed an increase for semi-natural and shrub-dominated respectively. We concluded that the degradation process of these tropical forests increases the stress of bird community generating genotoxic damage. Bird responses seem to be species-specific, which could explain the differences in changes in bird composition reported in other studies.

Investigating the Ability of Radiomics Features for Diagnosis of the Active Plaque of Multiple Sclerosis Patients

Background

Multiple sclerosis (MS) is the most common non-traumatic disabling disease.

Objective

The aim of this study is to investigate the ability of radiomics features for diagnosing active plaques in patients with MS from T2 Fluid Attenuated Inversion Recovery (FLAIR) images.

Material and Methods

In this experimental study, images of 82 patients with 122 MS lesions were investigated. Boruta and Relief algorithms were used for feature selection on the train data set (70%). Four different classifier algorithms, including Multi-Layer Perceptron (MLP), Gradient Boosting (GB), Decision Tree (DT), and Extreme Gradient Boosting (XGB) were used as classifiers for modeling. Finally, Performance metrics were obtained on the test data set (30%) with 1000 bootstrap and 95% confidence intervals (95% CIs).

Results

A total of 107 radiomics features were extracted for each lesion, of which 7 and 8 features were selected by the Relief method and Boruta method, respectively. DT classifier had the best performance in the two feature selection algorithms. The best performance on the test data set was related to Boruta-DT with an average accuracy of 0.86, sensitivity of 1.00, specificity of 0.84, and Area Under the Curve (AUC) of 0.92 (95% CI: 0.92-0.92).

Conclusion

Radiomics features have the potential for diagnosing MS active plaque by T2 FLAIR image features. Additionally, choosing the feature selection and classifier algorithms plays an important role in the diagnosis of active plaque in MS patients. The radiomics-based predictive models predict active lesions accurately and non-invasively.

Evaluation of automated detection of head position on lateral cephalometric radiographs based on deep learning techniques

BACKGROUND

Lateral cephalometric radiograph (LCR) is crucial to diagnosis and treatment planning of maxillofacial diseases, but inappropriate head position, which reduces the accuracy of cephalometric measurements, can be challenging to detect for clinicians. This non-interventional retrospective study aims to develop two deep learning (DL) systems to efficiently, accurately, and instantly detect the head position on LCRs.

METHODS

LCRs from 13 centers were reviewed and a total of 3000 radiographs were collected and divided into 2400 cases (80.0 %) in the training set and 600 cases (20.0 %) in the validation set. Another 300 cases were selected independently as the test set. All the images were evaluated and landmarked by two board-certified orthodontists as references. The head position of the LCR was classified by the angle between the Frankfort Horizontal (FH) plane and the true horizontal (HOR) plane, and a value within - 3°- 3° was considered normal. The YOLOv3 model based on the traditional fixed-point method and the modified ResNet50 model featuring a non-linear mapping residual network were constructed and evaluated. Heatmap was generated to visualize the performances.

RESULTS

The modified ResNet50 model showed a superior classification accuracy of 96.0 %, higher than 93.5 % of the YOLOv3 model. The sensitivity&recall and specificity of the modified ResNet50 model were 0.959, 0.969, and those of the YOLOv3 model were 0.846, 0.916. The area under the curve (AUC) values of the modified ResNet50 and the YOLOv3 model were 0.985 ± 0.04 and 0.942 ± 0.042, respectively. Saliency maps demonstrated that the modified ResNet50 model considered the alignment of cervical vertebras, not just the periorbital and perinasal areas, as the YOLOv3 model did.

CONCLUSIONS

The modified ResNet50 model outperformed the YOLOv3 model in classifying head position on LCRs and showed promising potential in facilitating making accurate diagnoses and optimal treatment plans.

Short text classification with machine learning in the social sciences: The case of climate change on Twitter

To analyse large numbers of texts, social science researchers are increasingly confronting the challenge of text classification. When manual labeling is not possible and researchers have to find automatized ways to classify texts, computer science provides a useful toolbox of machine-learning methods whose performance remains understudied in the social sciences. In this article, we compare the performance of the most widely used text classifiers by applying them to a typical research scenario in social science research: a relatively small labeled dataset with infrequent occurrence of categories of interest, which is a part of a large unlabeled dataset. As an example case, we look at Twitter communication regarding climate change, a topic of increasing scholarly interest in interdisciplinary social science research. Using a novel dataset including 5,750 tweets from various international organizations regarding the highly ambiguous concept of climate change, we evaluate the performance of methods in automatically classifying tweets based on whether they are about climate change or not. In this context, we highlight two main findings. First, supervised machine-learning methods perform better than state-of-the-art lexicons, in particular as class balance increases. Second, traditional machine-learning methods, such as logistic regression and random forest, perform similarly to sophisticated deep-learning methods, whilst requiring much less training time and computational resources. The results have important implications for the analysis of short texts in social science research.

Computer vision meets microfluidics: a label-free method for high-throughput cell analysis

In this paper, we review the integration of microfluidic chips and computer vision, which has great potential to advance research in the life sciences and biology, particularly in the analysis of cell imaging data. Microfluidic chips enable the generation of large amounts of visual data at the single-cell level, while computer vision techniques can rapidly process and analyze these data to extract valuable information about cellular health and function. One of the key advantages of this integrative approach is that it allows for noninvasive and low-damage cellular characterization, which is important for studying delicate or fragile microbial cells. The use of microfluidic chips provides a highly controlled environment for cell growth and manipulation, minimizes experimental variability and improves the accuracy of data analysis. Computer vision can be used to recognize and analyze target species within heterogeneous microbial populations, which is important for understanding the physiological status of cells in complex biological systems. As hardware and artificial intelligence algorithms continue to improve, computer vision is expected to become an increasingly powerful tool for in situ cell analysis. The use of microelectromechanical devices in combination with microfluidic chips and computer vision could enable the development of label-free, automatic, low-cost, and fast cellular information recognition and the high-throughput analysis of cellular responses to different compounds, for broad applications in fields such as drug discovery, diagnostics, and personalized medicine.

Machine learning reveals genetic modifiers of the immune microenvironment of cancer

Heritability in the immune tumor microenvironment (iTME) has been widely observed yet remains largely uncharacterized. Here, we developed a machine learning approach to map iTME modifiers within loci from genome-wide association studies (GWASs) for breast cancer (BrCa) incidence. A random forest model was trained on a positive set of immune-oncology (I-O) targets, and then used to assign I-O target probability scores to 1,362 candidate genes in linkage disequilibrium with 155 BrCa GWAS loci. Cluster analysis of the most probable candidates revealed two subfamilies of genes related to effector functions and adaptive immune responses, suggesting that iTME modifiers impact multiple aspects of anticancer immunity. Two of the top ranking BrCa candidates, LSP1 and TLR1, were orthogonally validated as iTME modifiers using BrCa patient biopsies and comparative mapping studies, respectively. Collectively, these data demonstrate a robust and flexible framework for functionally fine-mapping GWAS risk loci to identify translatable therapeutic targets.