High True Positive Rate Research Articles

Leveraging machine learning models for anemia severity detection among pregnant women following ANC: Ethiopian context

BackgroundAnemia during pregnancy is a significant public health concern, particularly in resource-limited settings. Machine learning (ML) offers promising avenues for improved anemia detection and management. This study investigates the potential of ML models in predicting anemia severity among pregnant women attending Antenatal Care (ANC) visits in Ethiopia.MethodsData from the Ethiopian Demographic Health Survey, specialized hospitals, and public hospitals were utilized. The dataset included individuals diagnosed with severe (65.12%), moderate (15.63%), mild (16.65%) anemia, and non-anemic (2.61%) cases. Feature selection employed filter methods based on mutual information, and F-score was used to assess anemia severity prediction across four classes. Six ML models (MLP-NN, XGBoost, GNB, Decision Tree, Random Forest, and KNN) were evaluated using accuracy, precision, recall, and F1-score.ResultsThe Random Forest classifier achieved the best overall performance across all categories, with an accuracy of 97%, precision of 93%, recall of 93%, and F1-score of 93%. This indicates high true positive rates and low false positive rates. While other models like XGBoost, MLP-NN, and Decision Tree showed good performance, they weren’t quite as strong as Random Forest. Classifiers like KNN and GNB had lower overall accuracy and a tendency to misclassify some cases.ConclusionsThis study demonstrates the promising potential of Random Forest in predicting anemia severity among pregnant women in Ethiopia. The findings contribute to a more holistic understanding of anemia risk factors and pave the way for improved early detection and targeted interventions.

HawkEye: An end-host method to detect the Low-rate Denial-of-Service attack of cross-traffic over bottleneck links

The adaptive mechanisms of the Transmission Control Protocol (TCP) address network congestion and other unpredictable network conditions. They ensure the reliability of data transmission and the stability of the network. Unfortunately, the vulnerabilities in these adaptive mechanisms are targeted explicitly by low-rate denial-of-service (LDoS) attacks, which severely degrade network service quality. Only by modifying these protocols and addressing their vulnerabilities can one entirely prevent LDoS attacks. Although various improved TCP algorithms exist, they often fail to identify LDoS attacks accurately and, in some cases, may even reduce TCP performance. Furthermore, traditional LDoS attack detection methods rely on intermediate devices, which do not meet TCP’s end-to-end performance optimization needs. We introduce HawkEye, which moves the detection mechanism to the end hosts to address this issue. HawkEye uses an improved genetic algorithm to fine-tune the parameters of the LightGBM on the sending host, integrating multiple network traffic features to detect LDoS attacks. Experimental results show that our proposed method achieves the high accuracy, high true positive rate, and low false positive rate, successfully addressing the limitations of end-host detection of LDoS attacks and providing an innovative and effective solution for enhancing LDoS attack detection.

Digital Image Processing to Detect Adaptive Evolution.

In recent years, advances in image processing and machine learning have fueled a paradigm shift in detecting genomic regions under natural selection. Early machine learning techniques employed population-genetic summary statistics as features, which focus on specific genomic patterns expected by adaptive and neutral processes. Though such engineered features are important when training data are limited, the ease at which simulated data can now be generated has led to the recent development of approaches that take in image representations of haplotype alignments and automatically extract important features using convolutional neural networks. Digital image processing methods termed α-molecules are a class of techniques for multiscale representation of objects that can extract a diverse set of features from images. One such α-molecule method, termed wavelet decomposition, lends greater control over high-frequency components of images. Another α-molecule method, termed curvelet decomposition, is an extension of the wavelet concept that considers events occurring along curves within images. We show that application of these α-molecule techniques to extract features from image representations of haplotype alignments yield high true positive rate and accuracy to detect hard and soft selective sweep signatures from genomic data with both linear and nonlinear machine learning classifiers. Moreover, we find that such models are easy to visualize and interpret, with performance rivaling those of contemporary deep learning approaches for detecting sweeps.

Protein-Protein Interaction Networks Derived from Classical and Machine Learning-Based Natural Language Processing Tools.

The study of protein-protein interactions (PPIs) provides insight into various biological mechanisms, including the binding of antibodies to antigens, enzymes to inhibitors or promoters, and receptors to ligands. Recent studies of PPIs have led to significant biological breakthroughs. For example, the study of PPIs involved in the human:SARS-CoV-2 viral infection mechanism aided in the development of SARS-CoV-2 vaccines. Though several databases exist for the manual curation of PPI networks, text mining methods have been routinely demonstrated as useful alternatives for newly studied or understudied species, where databases are incomplete. Here, the relationship extraction performance of several open-source classical text processing, machine learning (ML)-based natural language processing (NLP), and large language model (LLM)-based NLP tools was compared. Overall, our results indicated that networks derived from classical methods tend to have high true positive rates at the expense of having overconnected networks, ML-based NLP methods have lower true positive rates but networks with the closest structures to the target network, and LLM-based NLP methods tend to exist between the two other approaches, with variable performances. The selection of a specific NLP approach should be tied to the needs of a study and text availability, as models varied in performance due to the amount of text provided.

A robust distance-based approach for detecting multidimensional outliers

ABSTRACT Identifying outliers in data analysis is a critical task, as outliers can significantly influence the results and conclusions drawn from a dataset. This study explores the use of the Mahalanobis distance metric for detecting outliers in multivariate data, focusing on a novel approach inspired by the work of M. Falk, [On mad and comedians, Ann. Inst. Stat. Math. 49 (1997), pp. 615–644]. The proposed method is rigorously tested through extensive simulation analysis, where it demonstrates high True Positive Rates (TPR) and low False Positive Rates (FPR) when compared to other existing outlier detection techniques. Through extensive simulation analysis, we empirically evaluate the affine equivariance and breakdown properties of our proposed distance measure and it is evident from the outputs that our robust distance measure demonstrates effective results with respect to the measures FPR and TPR. The proposed method was applied to seven different datasets, showing promising true positive rates (TPR) and false positive rates (FPR), and it outperformed several well-known outlier identification approaches. We can effectively use our proposed distance measure in fields demanding outlier detection.

Regulatory genome annotation of 33 insect species

Annotation of newly sequenced genomes frequently includes genes, but rarely covers important non-coding genomic features such as the cis-regulatory modules—e.g., enhancers and silencers—that regulate gene expression. Here, we begin to remedy this situation by developing a workflow for rapid initial annotation of insect regulatory sequences, and provide a searchable database resource with enhancer predictions for 33 genomes. Using our previously developed SCRMshaw computational enhancer prediction method, we predict over 2.8 million regulatory sequences along with the tissues where they are expected to be active, in a set of insect species ranging over 360 million years of evolution. Extensive analysis and validation of the data provides several lines of evidence suggesting that we achieve a high true-positive rate for enhancer prediction. One, we show that our predictions target specific loci, rather than random genomic locations. Two, we predict enhancers in orthologous loci across a diverged set of species to a significantly higher degree than random expectation would allow. Three, we demonstrate that our predictions are highly enriched for regions of accessible chromatin. Four, we achieve a validation rate in excess of 70% using in vivo reporter gene assays. As we continue to annotate both new tissues and new species, our regulatory annotation resource will provide a rich source of data for the research community and will have utility for both small-scale (single gene, single species) and large-scale (many genes, many species) studies of gene regulation. In particular, the ability to search for functionally related regulatory elements in orthologous loci should greatly facilitate studies of enhancer evolution even among distantly related species.

Regulatory genome annotation of 33 insect species.

Annotation of newly sequenced genomes frequently includes genes, but rarely covers important non-coding genomic features such as the cis-regulatory modules-e.g., enhancers and silencers-that regulate gene expression. Here, we begin to remedy this situation by developing a workflow for rapid initial annotation of insect regulatory sequences, and provide a searchable database resource with enhancer predictions for 33 genomes. Using our previously developed SCRMshaw computational enhancer prediction method, we predict over 2.8 million regulatory sequences along with the tissues where they are expected to be active, in a set of insect species ranging over 360 million years of evolution. Extensive analysis and validation of the data provides several lines of evidence suggesting that we achieve a high true-positive rate for enhancer prediction. One, we show that our predictions target specific loci, rather than random genomic locations. Two, we predict enhancers in orthologous loci across a diverged set of species to a significantly higher degree than random expectation would allow. Three, we demonstrate that our predictions are highly enriched for regions of accessible chromatin. Four, we achieve a validation rate in excess of 70% using in vivo reporter gene assays. As we continue to annotate both new tissues and new species, our regulatory annotation resource will provide a rich source of data for the research community and will have utility for both small-scale (single gene, single species) and large-scale (many genes, many species) studies of gene regulation. In particular, the ability to search for functionally related regulatory elements in orthologous loci should greatly facilitate studies of enhancer evolution even among distantly related species.

Assessment of Racial Bias within the Risk Analysis Index of Frailty

Objective: Our objective was to assess potential racial bias within the Risk Analysis Index (RAI). Background: Patient risk measures are rarely tested for racial bias. Measures of frailty, like the RAI, need to be evaluated for poor predictive performance among Black patients. Methods: Retrospective cohort study using April 2010–March 2019 Veterans Affairs Surgical Quality Improvement Program and 2010–2019 National Surgical Quality Improvement Program data. The performance of the RAI and several potential variants were compared between Black and White cases using various metrics to predict mortality (180-day for Veterans Affairs Surgical Quality Improvement Program, 30-day for National Surgical Quality Improvement Program). Results: Using the current, clinical threshold, the RAI performed as good or better among Black cases across various performance metrics versus White. When a higher threshold was used, Black cases had higher true positive rates but lower true negative rates, yielding 2.0% higher balanced accuracy. No RAI variant noticeably eliminated bias, improved parity across both true positives and true negatives, or improved overall model performance. Conclusions: The RAI tends to predict mortality among Black patients better than it predicts mortality among White patients. As existing bias-reducing techniques were not effective, further research into bias-reducing techniques is needed, especially for clinical risk predictions. We recommend using the RAI for both statistical analysis of surgical cohorts and quality improvement programs, such as the Surgical Pause.

Predicting rock art sites in the Pajeú watershed, Brazil

This paper presents an Archaeological Predictive Model (APM) to predict rock art archaeological sites in the Pajeú Watershed, a semiarid region in Pernambuco, Brazil. The model uses Machine Learning (ML) algorithms and re-sampling techniques to account for the unbalanced data set of rock art sites and test different inductive methods for predicting site location. The results show a satisfactory statistical evaluation, with high true positive rates with all ML algorithms and resampling techniques used, indicating a high potential for predicting rock art site locations. The predictive maps generated from the model output, show that certain features, such as aspect, elevation and the distance to different lithologies, are particularly important. The overall model's performance could be corroborated with a test in another semi-arid region, next to the Pajeú watershed, where areas with high favorability of finding rock art sites are predicted near to already known archaeological sites.

Curvature analysis of CI electrode arrays: a novel approach to categorize perimodiolar positions without anatomical landmarks.

Perimodiolar electrode arrays may be positioned regular, over-inserted or under-inserted into the cochlea depending on the cochlear size and shape. The study aimed to examine whether there are differences between these groups in the local curvature along the intracochlear array. Individual curvature variables were developed to categorize the groups and the relationship between the curvature and the angular insertion depth at the electrode tip was analyzed. The curvature along the intracochlear array was measured in the CBCT image of 85 perimodiolar electrodes of a single type. The mean curvature and the ratio of the mean curvature at contacts E14-16 to the mean curvature at E7-8 (bowing ratio) were calculated across the array, and its true positive rate (TPR) and false positive rate (FPR) were calculated to establish optimal threshold values to categorize the groups. 68.2% of the cases were categorized as regular positioned, 22.4% had an over-insertion and 9.4% had an under-insertion. The mean curvature was significantly weaker with under-insertion (< 342°) than with normal insertion depth (≥ 342°). With an over-insertion, the bowing ratio was < 1 and otherwise > 1. Both the mean curvature and bowing ratio were found to have an optimal threshold value with high TPR (= 1.00) and low FPR (≤ 0.06) for categorizing under-insertion and over-insertion, respectively. Curvature analysis is a useful tool to assess if a perimodiolar electrode array has been inserted deep enough into the cochlea. Independent of critical anatomical landmarks, over-inserted arrays and under-inserted arrays could be well categorized by using individual curvature variables. The results need to be validated using additional data sets.

COVID-19 severity detection using chest X-ray segmentation and deep learning

COVID-19 has resulted in a significant global impact on health, the economy, education, and daily life. The disease can range from mild to severe, with individuals over 65 or those with underlying medical conditions being more susceptible to severe illness. Early testing and isolation are vital due to the virus’s variable incubation period. Chest radiographs (CXR) have gained importance as a diagnostic tool due to their efficiency and reduced radiation exposure compared to CT scans. However, the sensitivity of CXR in detecting COVID-19 may be lower. This paper introduces a deep learning framework for accurate COVID-19 classification and severity prediction using CXR images. U-Net is used for lung segmentation, achieving a precision of 0.9924. Classification is performed using a Convulation-capsule network, with high true positive rates of 86% for COVID-19, 93% for pneumonia, and 85% for normal cases. Severity assessment employs ResNet50, VGG-16, and DenseNet201, with DenseNet201 showing superior accuracy. Empirical results, validated with 95% confidence intervals, confirm the framework’s reliability and robustness. This integration of advanced deep learning techniques with radiological imaging enhances early detection and severity assessment, improving patient management and resource allocation in clinical settings.

Leak detection in water distribution networks based on deep learning and kriging interpolation method

ABSTRACT The burgeoning growth of urban areas has escalated the necessity for efficient and precise leak detection in water distribution networks. Automatic detection methods based on deep learning are a state-of-the-art research topic. In this paper, a methodology that combines deep learning and data imaging is proposed. The framework employs pressure monitoring data and is anchored on the following three pillars: (1) the generation of a comprehensive dataset, encompassing one year of leak-free demand data derived from Fourier Series analysis and monitoring pressure under normal and leak conditions, (2) the transformation of pressure time series into images using kriging interpolation, (3) establishing convolution neural network (CNN) and evaluating its performance of abnormal identification. The effectiveness of the proposed methodology is assessed in different image sets under various leak conditions. The findings reveal that this method meets dependable and effective outputs for leak detection, with the deep learning model achieving a high true positive rate (TPR) of 98% and an area under the curve (AUC) of 94%. This study provides invaluable information for strategic action planning and the enhancement of water loss management protocols, especially in situations where water utilities and regulatory authorities grappling with limited budgets and diminishing revenues.

MSKI-RADS: An MRI-based Musculoskeletal Infection Reporting and Data System for the Diagnosis of Extremity Infections.

Background Current terms used to describe the MRI findings for musculoskeletal infections are nonspecific and inconsistent. Purpose To develop and validate an MRI-based musculoskeletal infection classification and scoring system. Materials and Methods In this retrospective cross-sectional internal validation study, a Musculoskeletal Infection Reporting and Data System (MSKI-RADS) was designed. Adult patients with radiographs and MRI scans of suspected extremity infections with a known reference standard obtained between June 2015 and May 2019 were included. The scoring categories were as follows: 0, incomplete imaging; I, negative for infection; II, superficial soft-tissue infection; III, deeper soft-tissue infection; IV, possible osteomyelitis (OM); V, highly suggestive of OM and/or septic arthritis; VI, known OM; and NOS (not otherwise specified), nonspecific bone lesions. Interreader agreement for 20 radiologists from 13 institutions (intraclass correlation coefficient [ICC]) and true-positive rates of MSKI-RADS were calculated and the accuracy of final diagnoses rendered by the readers was compared using generalized estimating equations for clustered data. Results Among paired radiographs and MRI scans from 208 patients (133 male, 75 female; mean age, 55 years ± 13 [SD]), 20 were category I; 34, II; 35, III; 30, IV; 35, V; 18, VI; and 36, NOS. Moderate interreader agreement was observed among the 20 readers (ICC, 0.70; 95% CI: 0.66, 0.75). There was no evidence of correlation between reader experience and overall accuracy (P = .94). The highest true-positive rate was for MSKI-RADS I and NOS at 88.7% (95% CI: 84.6, 91.7). The true-positive rate was 73% (95% CI: 63, 80) for MSKI-RADS V. Overall reader accuracy using MSKI-RADS across all patients was 65% ± 5, higher than final reader diagnoses at 55% ± 7 (P < .001). Conclusion MSKI-RADS is a valid system for standardized terminology and recommended management of imaging findings of peripheral extremity infections across various musculoskeletal-fellowship-trained reader experience levels. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Schweitzer in this issue.

Assessment of Effectiveness of the Algorithm for Automated Quantitative Analysis of Metallic Strut Tissue Short-Term Coverage with Intravascular Optical Coherence Tomography.

Background: Due to its high resolution, optical coherence tomography (OCT) is the most suitable modality for neointimal coverage assessments. Evaluation of stent healing seems crucial to accurately define their safety profile since delayed healing is connected with stent thrombosis. This study aimed to present an algorithm for automated quantitative analysis of stent strut coverage at the early stages of vessel healing in intravascular OCT. Methods: A set of 592 OCT frames from 24 patients one month following drug-eluting stent implantation was used to assess the algorithm's effectiveness. Struts not covered on any side or covered but only on one side were categorized as uncovered. The algorithm consists of several key steps: preprocessing, vessel lumen segmentation, automatic strut detection, and measurement of neointimal thickness. Results: The proposed algorithm proved its efficiency in lumen and stent area estimation versus manual reference. It showed a high positive predictive value (PPV) (89.7%) and true positive rate (TPR) (91.4%) in detecting struts. A qualitative assessment for covered and uncovered struts was characterized by high TPR (99.1% and 80%, respectively, for uncovered and covered struts) and PPV (77.3% and 87%). Conclusions: The proposed algorithm demonstrated good agreement with manual measurements. Automating the stent coverage assessment might facilitate imaging analysis, which might be beneficial in experimental and clinical settings.

Identifying topologically associating domains using differential kernels.

Chromatin is a polymer complex of DNA and proteins that regulates gene expression. The three-dimensional (3D) structure and organization of chromatin controls DNA transcription and replication. High-throughput chromatin conformation capture techniques generate Hi-C maps that can provide insight into the 3D structure of chromatin. Hi-C maps can be represented as a symmetric matrix [Formula: see text], where each element represents the average contact probability or number of contacts between chromatin loci i and j. Previous studies have detected topologically associating domains (TADs), or self-interacting regions in [Formula: see text] within which the contact probability is greater than that outside the region. Many algorithms have been developed to identify TADs within Hi-C maps. However, most TAD identification algorithms are unable to identify nested or overlapping TADs and for a given Hi-C map there is significant variation in the location and number of TADs identified by different methods. We develop a novel method to identify TADs, KerTAD, using a kernel-based technique from computer vision and image processing that is able to accurately identify nested and overlapping TADs. We benchmark this method against state-of-the-art TAD identification methods on both synthetic and experimental data sets. We find that the new method consistently has higher true positive rates (TPR) and lower false discovery rates (FDR) than all tested methods for both synthetic and manually annotated experimental Hi-C maps. The TPR for KerTAD is also largely insensitive to increasing noise and sparsity, in contrast to the other methods. We also find that KerTAD is consistent in the number and size of TADs identified across replicate experimental Hi-C maps for several organisms. Thus, KerTAD will improve automated TAD identification and enable researchers to better correlate changes in TADs to biological phenomena, such as enhancer-promoter interactions and disease states.

Multi‐Model Machine Learning Approach Accurately Predicts Lake Dissolved Oxygen With Multiple Environmental Inputs

AbstractAs a key water quality parameter, dissolved oxygen (DO) concentration, and particularly changes in bottom water DO is fundamental for understanding the biogeochemical processes in lake ecosystems. Based on two machine learning (ML) models, Gradient Boost Regressor (GBR) and long‐short‐term‐memory (LSTM) network, this study developed three ML model approaches: direct GBR; direct LSTM; and a 2‐step mixed ML model workflow combining both GBR and LSTM. They were used to simulate multi‐year surface and bottom DO concentrations in five lakes. All approaches were trained with readily available environmental data as predictors. Indices of lake thermal structure and mixing provided by a one‐dimensional (1‐D) hydrodynamic model were also included as predictors in the ML models. The advantages of each ML approach were not consistent for all the tested lakes, but the best one of them was defined that can estimate DO concentration with coefficient of determination (R2) up to 0.6–0.7 in each lake. All three approaches have normalized mean absolute error (NMAE) under 0.15. In a polymictic lake, the 2‐step mixed model workflow showed better representation of bottom DO concentrations, with a highest true positive rate (TPR) of hypolimnetic hypoxia detection of over 90%, while the other workflows resulted in, TPRs are around 50%. In most of the tested lakes, the predicted surface DO concentrations and variables indicating stratified conditions (i.e., Wedderburn number and the temperature difference between surface and bottom water) are essential for simulating bottom DO. The ML approaches showed promising results and could be used to support short‐ and long‐term water management plans.

An approach-based machine learning and automated thermal images to predict the dark-cutting incidence in cattle management of healthcare supply chain

The healthcare supply chain is a network made up of various systems, processes, and elements that function and interact seamlessly to offer healthcare services and products. Food safety is an essential component of the healthcare supply chain. In the cattle industry, the healthcare supply chain contributes positively to the global economy through providing high-quality products such as milk and meat. Stress in cattle is one of main factor that cause low quality meat called “dark meat”. Numerous studies have been conducted on the development of different non-invasive methods based on Infrared Thermography Technology (IRT) to enhance the meat quality by detecting stress in cattle pre-slaughtering. These studies have the following issues: lack of automating in detecting body temperature of cattle and ignoring detecting stress with prediction dark meat incidence. The present study endeavors a new fully automated system for detecting stress, and dark meat incidence, incorporating the following new approaches: Multiview face detecting, Automatic eye localisation for detecting body temperature automatically employing computer vision and image processing, respectively. Furthermore, machine learning algorithms like Support Vector Machine (SVM), Logistic Regression (LR), Naïve Bayes (NB), and Decision Tree (DT) have been developed specifically for stress detection and the prediction of dark meat. To develop automated system, two forms of data were collected: statistical temperature and infrared thermal images. Infrared thermal images are used to develop Multiview face detection and Automatic eye localisation. Temperature data used to develop the machine learning model. Results reveal that Multiview face detection better than the current methods in term of Precision 0.99, Recall 0.91, F-score 0.95 with high True positive rate 0.90 and zero False-positive rate. Automatic eye localisation has high accuracy, with the following values: sensitivity 0.9780, precision 0.7212, F measure of 0.8024, and misclassification 0.0455. Lastly, results elaborate that the decision tree model can attain a notable level of accuracy in terms of specificity, recall, F-measure, and Area Under the Curve (AUC), all at an optimal rate of 98%.

HES V2.0 outperforms GALAD for detection of HCC: A phase 3 biomarker study in the United States.

The original hepatocellular carcinoma early detection screening (HES) score, which combines alpha-fetoprotein (AFP) with age, alanine aminotransferase, and platelets, has better performance than AFP alone for early HCC detection. We have developed HES V2.0 by adding AFP-L3 and des-gamma-carboxy prothrombin to the score and compared its performance to GALAD and ASAP scores among patients with cirrhosis. We conducted a prospective-specimen collection, retrospective-blinded-evaluation phase 3 biomarker cohort study in patients with cirrhosis enrolled in imaging and AFP surveillance. True-positive rate (TPR)/sensitivity and false-positive rate for any or early HCC were calculated for GALAD, ASAP, and HES V2.0 scores within 6, 12, and 24 months of HCC diagnosis. We calculated the AUROC curve and estimated TPR based on an optimal threshold at a fixed false-positive rate of 10%. We analyzed 2331 patients, of whom 125 developed HCC (71% in the early stages). For any HCC, HES V2.0 had higher TPR than GALAD overall (+7.2%), at 6 months (+3.6%), at 12 months (+7.2%), and 24 months (+13.0%) before HCC diagnosis. HES V2.0 had higher TPR than ASAP for all time points (+5.9% to +12.0%). For early HCC, HES V2.0 had higher sensitivity/TPR than GALAD overall (+6.7%), at 12 months (+6.3%), and 24 months (+14.6%) but not at 6 months (+0.0%) and higher than ASAP for all time points (+13.4% to +18.0%). In a prospective cohort study, HES V2.0 had a significantly higher performance for identifying new HCC, including early stage, than GALAD or ASAP.

Childhood Environmental Instabilities and Their Behavioral Implications: A Machine Learning Approach to Studying Adverse Childhood Experiences.

Adverse childhood experiences (ACEs) include a range of abusive, neglectful, and dysfunctional household behaviors that are strongly associated with long-term health problems, mental health conditions, and societal difficulties. The study aims to uncover significant factors influencing ACEs in children aged 0-17 years and to propose a predictive model that can be used to forecast the likelihood of ACEs in children. Machine learning models are applied to identify and analyze the relationships between several predictors and the occurrence of ACEs. Key performance metrics such as AUC, F1 score, recall, and precision are used to evaluate the predictive strength of different factors on ACEs. Family structures, especially non-traditional forms such as single parenting, and the frequency of relocating to a new address are determined as key predictors of ACEs. The final model, a neural network, achieved an AUC of 0.788, a precision score of 0.683, and a recall of 0.707, indicating its effectiveness in accurately identifying ACE cases. The model's ROC and PR curves showed a high true positive rate for detecting children with two or more ACEs while also pointing to difficulties in classifying single ACE instances accurately. Furthermore, our analysis revealed the intricate relationship between the frequency of relocation and other predictive factors. The findings highlight the importance of familial and residential stability in children's lives, with substantial implications for child welfare policies and interventions. The study emphasizes the need for targeted educational and healthcare support to promote the well-being and resilience of at-risk children.

Fast and Lightweight Vision-Language Model for Adversarial Traffic Sign Detection

Several attacks have been proposed against autonomous vehicles and their subsystems that are powered by machine learning (ML). Road sign recognition models are especially heavily tested under various adversarial ML attack settings, and they have proven to be vulnerable. Despite the increasing research on adversarial ML attacks against road sign recognition models, there is little to no focus on defending against these attacks. In this paper, we propose the first defense method specifically designed for autonomous vehicles to detect adversarial ML attacks targeting road sign recognition models, which is called ViLAS (Vision-Language Model for Adversarial Traffic Sign Detection). The proposed defense method is based on a custom, fast, lightweight, and salable vision-language model (VLM) and is compatible with any existing traffic sign recognition system. Thanks to the orthogonal information coming from the class label text data through the language model, ViLAS leverages image context in addition to visual data for highly effective attack detection performance. In our extensive experiments, we show that our method consistently detects various attacks against different target models with high true positive rates while satisfying very low false positive rates. When tested against four state-of-the-art attacks targeting four popular action recognition models, our proposed detector achieves an average AUC of 0.94. This result achieves a 25.3% improvement over a state-of-the-art defense method proposed for generic image attack detection, which attains an average AUC of 0.75. We also show that our custom VLM is more suitable for an autonomous vehicle compared to the popular off-the-shelf VLM and CLIP in terms of speed (4.4 vs. 9.3 milliseconds), space complexity (0.36 vs. 1.6 GB), and performance (0.94 vs. 0.43 average AUC).