Pre-defined Categories Research Articles

A Survey on Open-Vocabulary Detection and Segmentation: Past, Present, and Future.

As the most fundamental scene understanding tasks, object detection and segmentation have made tremendous progress in deep learning era. Due to the expensive manual labeling cost, the annotated categories in existing datasets are often small-scale and pre-defined, i.e., state-of-the-art fully-supervised detectors and segmentors fail to generalize beyond the closed vocabulary. To resolve this limitation, in the last few years, the community has witnessed an increasing attention toward Open-Vocabulary Detection (OVD) and Segmentation (OVS). By "open-vocabulary", we mean that the models can classify objects beyond pre-defined categories. In this survey, we provide a comprehensive review on recent developments of OVD and OVS. A taxonomy is first developed to organize different tasks and methodologies. We find that the permission and usage of weak supervision signals can well discriminate different methodologies, including: visual-semantic space mapping, novel visual feature synthesis, region-aware training, pseudo-labeling, knowledge distillation, and transfer learning. The proposed taxonomy is universal across different tasks, covering object detection, semantic/instance/panoptic segmentation, 3D and video understanding. The main design principles, key challenges, development routes, methodology strengths, and weaknesses are thoroughly analyzed.

Strategies to optimise the health equity impact of digital pain self-reporting tools: a series of multi-stakeholder focus groups

BackgroundThere are avoidable differences (i.e., inequities) in the prevalence and distribution of chronic pain across diverse populations, as well as in access to and outcomes of pain management services. Digital pain self-reporting tools have the potential to reduce or exacerbate these inequities. This study aimed to better understand how to optimise the health equity impact of digital pain self-reporting tools on people who are experiencing (or are at risk of) digital pain inequities.MethodsThis was a qualitative study, guided by the Health Equity Impact Assessment tool—digital health supplement (HEIA-DH). We conducted three scoping focus groups with multiple stakeholders to identify the potential impacts of digital pain self-reporting tools and strategies to manage these impacts. Each group focused on one priority group experiencing digital pain inequities, including older adults, ethnic minorities, and people living in socio-economically deprived areas. A fourth consensus focus group was organised to discuss and select impact management strategies. Focus groups were audio-recorded, transcribed verbatim, and analysed using a framework approach. We derived codes, grouped them under four pre-defined categories from the HEIA-DH, and illustrated them with participants’ quotes.ResultsA total of fifteen people living with musculoskeletal pain conditions and thirteen professionals took part. Participants described how digital pain self-reports can have a positive health equity impact by better capturing pain fluctuations and enriching patient-provider communication, which in turn can enhance clinical decisions and self-management practices. Conversely, participants identified that incorrect interpretation of pain reports, lack of knowledge of pain terminologies, and digital (e.g., no access to technology) and social (e.g., gender stereotyping) exclusions may negatively impact on people’s health equity. The participants identified 32 strategies, of which 20 were selected as being likely to mitigate these negative health equity impacts. Example strategies included, e.g., option to customise self-reporting tools in line with users’ personal preferences, or resources to better explain how self-reported pain data will be used to build trust.ConclusionLinked to people’s personal and social characteristics, there are equity-based considerations for developing accessible digital pain self-reporting tools, as well as resources and skills to enable the adoption and use of these tools among priority groups. Future research should focus on implementing these equity-based considerations or strategies identified by our study and monitoring their impact on the health equity of people living with chronic pain.

Accurate Prediction of Bangla Text Article Categorization by Utilizing Novel Bangla Stemmer

Text categorization involves assigning predefined category labels to an unlabeled document. With the exponential growth in the accessibility and availability of digital documents over the past decade, this field significantly attracted the scientific community that immensely demands rapid and accurate categorization of these documents. Relying on experts for manual classification is time-consuming and resource-intensive. Consequently, labeling unlabeled digital documents faster more accurately, and more efficiently is inescapable. One promising approach to addressing this demand is the use of machine learning algorithms. Training these algorithms on a large dataset of labeled texts lets them learn patterns and predicted unlabeled documents. This strategy might greatly expedite the categorizing process while retaining a substantial level of accuracy through leveraging artificial intelligence. These algorithms have also enhanced natural language processing techniques, making them more accurate at classifying unlabeled digital documents. In this study, we propose a novel machine-learning computational framework to address this challenge. Our framework incorporates a novel Bangla stemmer, which reduces words to their stems. We then employed TF-IDF for document vectorization, a statistical measure assessing word relevance for categorization purposes. Experimental results reveal that our framework significantly enhances prediction performance, achieving an impressive 95.3% prediction accuracy.

Abstract A007: Beyond continental ancestry categories: trans-population genomic similarity continuum reveals overlooked somatic-germline interactions across cancer types

Abstract A growing body of research highlighted a strong presence of continuums that describe human genomic diversity. Historical scholarship has also extensively investigated the use of discrete classification systems in human genomics further questioning their validity. To investigate the limiting effect of pre-defined stratification, we performed a wide comparative driver GWAS analysis at the trans-population (i.e. without stratification) vs continent-based population levels across 16 cancer types. Results demonstrate that in the MSK-IMPACT cohort (n = 56,826 samples), the analysis of the trans-population genomic similarity continuum captures significant somatic-germline interactions (p < 5x10-8) in cancer type-specific and pancancer cohorts, with a set of interactions overlooked by the discrete continent-based paradigm. We demonstrate that overlooked germline SNPs have similar allele frequencies and directional relationships with somatic alterations across pre-defined discrete categories, and that earlier stratification prevents their detection due to lowered statistical power. PheWAS analysis of detected germline variants revealed associations with cancer and non-cancer phenotypes reported in previous studies as well as novel hits. To expand the scope of the trans-population lens, we carried out a comparative analysis on 17 whole-exome TCGA cohorts to identify potential cancer drivers based on mutational frequencies. Results similarly revealed known and potential cancer drivers at the trans-population level that would be overlooked when pre-defined continent-based stratification is applied. The lens we adopt further highlights the necessity for new quantitative perspectives to comprehensively detect genomic patterns and to leverage entire genomic datasets without pre-defined stratification schemes. Citation Format: Hussein Mohsen, Tomin Perea-Chamblee, Jian Carrot-Zhang. Beyond continental ancestry categories: trans-population genomic similarity continuum reveals overlooked somatic-germline interactions across cancer types [abstract]. In: Proceedings of the 17th AACR Conference on the Science of Cancer Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; 2024 Sep 21-24; Los Angeles, CA. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2024;33(9 Suppl):Abstract nr A007.

Feasibility of big data analytics in disaster psychiatry: Impact of Seoul Itaewon tragedy on sentiment distribution on Twitter

Numerous studies have highlighted the significant impact of disasters on mental health, often leading to psychiatric disorders among affected individuals. Timely identification of disaster-related mental health problems is crucial to prevent long-term negative consequences and improve individual and community resilience. To address the limitations of prior research that has focused solely on isolated incidents, we analyzed the impact of a recurring Halloween event in Itaewon, South Korea, which culminated tragically in a crowd crush incident in 2022. We conducted sentiment analysis on big data from Korean Twitter to gauge the impact of this disaster on public sentiment. We collected tweets 2 weeks before and after the annual festival from 2020 to 2022, allowing for the consideration of variability across years and days before the disaster. Using a pre-trained RoBERTa neural network model fine-tuned with public sentiment datasets, we categorized tweets into seven pre-defined emotional categories: Anger, sadness, happiness, disgust, fear, surprise, and neutrality. These sentiments were then analyzed as daily time-series data. The overall tweet volume across all sentiment categories increased, particularly showing an increase in the number of tweets indicating “Sadness” in 2022 compared with that in previous years. Post-disaster, a substantial increase was noted in the proportion of tweets expressing “Sadness” and “Fear.” This trend was confirmed by Seasonal Autoregressive Integrated Moving Average with Exogenous Regressor models. Notably, there was an increase in the number of tweets expressing all sentiments, including “Happy.” However, significant changes in proportions were observed only in tweets categorized as expressing “Sadness” [0.046 (95% CI: 0.024–0.068, P < 0.0001)] and “Fear” [0.033 (95% CI: 0.014–0.051, P < 0.0001)]. Our study demonstrates the feasibility of using sentiment data from social media, combined with sentiment classification, to assess distinct public mental health features following disasters. This approach provides valuable insights into the emotional impact of each event.

UNRAVELLING THE DETERMINANTS OF SOCIAL MEDIA PAYMENT PLATFORM (SMPP) USAGE: A QUALITATIVE STUDY ON USER INTENTIONS AND ADOPTION

Lately, social media companies worldwide have introduced payment features on their social media platforms. This has given social media users the option to make payments to their near and dear ones without changing the platform while engaged in chatting. The adoption of Social Media Payment Platforms (SMPP) varies across user groups and is influenced by socio-demographic characteristics, convenience, and perceived risk. Understanding these factors is essential for enhancing user engagement and trust, thereby addressing the gaps in current SMPP adoption research. The purpose of this study is to ascertain the behavioural intention and perception level of users toward Social Media Payment Platforms (SMPP) adoption and to identify the factors which motivate and inhibit from using social payment apps. To understand the intention of the users to behave and adopt SMPP, we conducted in-depth structured interviews. Qualitative analysis was used to analyse the data, using pre-defined categories (deductive) and code emerging from the respondent discussions (inductively) to identify the relevant influencing factors. Through data analysis, we identified three key themes that categorised user behaviour: socio-demographic, convivence and vulnerability to risks, which is further divided into several sub-themes such as age, gender, education, occupation under socio-demographic, while perceived usefulness (PU) and perceived ease of use (PEOU) comes under the theme of convivence and lastly, under Vulnerability of risk includes privacy, security and platform features. The findings of the study can be validated and explored in future research, which will provide a clear understanding of identifying factors that influence SMPP adoption and behaviour intention.

Towards Open Vocabulary Learning: A Survey.

In the field of visual scene understanding, deep neural networks have made impressive advancements in various core tasks like segmentation, tracking, and detection. However, most approaches operate on the close-set assumption, meaning that the model can only identify pre-defined categories that are present in the training set. Recently, open vocabulary settings were proposed due to the rapid progress of vision language pre-training. These new approaches seek to locate and recognize categories beyond the annotated label space. The open vocabulary approach is more general, practical, and effective than weakly supervised and zero-shot settings. This paper thoroughly reviews open vocabulary learning, summarizing and analyzing recent developments in the field. In particular, we begin by juxtaposing open vocabulary learning with analogous concepts such as zero-shot learning, open-set recognition, and out-of-distribution detection. Subsequently, we examine several pertinent tasks within the realms of segmentation and detection, encompassing long-tail problems, few-shot, and zero-shot settings. As a foundation for our method survey, we first elucidate the fundamental principles of detection and segmentation in close-set scenarios. Next, we examine various contexts where open vocabulary learning is employed, pinpointing recurring design elements and central themes. This is followed by a comparative analysis of recent detection and segmentation methodologies in commonly used datasets and benchmarks. Our review culminates with a synthesis of insights, challenges, and discourse on prospective research trajectories. To our knowledge, this constitutes the inaugural exhaustive literature review on open vocabulary learning.

Eliminating and mining strategies for open-world object proposal

Object proposal serves as a crucial pre-task of many image and video understanding applications. However, modern approaches for object proposal are typically based on closed-world assumptions, focusing only on pre-defined categories. This approach cannot meet the diverse needs of real-world applications. To address this limitation, we introduce two strategies, namely the eliminating strategy and the mining strategy, to robustly train the Object Localization Network (OLN) for open-world object proposal. The eliminating strategy takes into account the spatial configuration between labeled boxes, thereby eliminating box anchors that overlap with multiple objects. The mining strategy employs a pseudo-label guided self-training scheme, enabling the mining of object boxes in novel categories. Without bells and whistles, our proposed method outperforms previous state-of-the-art methods on large-scale benchmarks, including COCO, Objects365, and UVO. The source codes are available at https://github.com/hustvl/EM-OLN.

#1441 COVID-19 vaccination and risk of infection and subsequent death in patients with kidney disease

Abstract Background and Aims The effect of vaccinations against SARS-CoV-2 on risk of infection and subsequent adverse outcomes in patients with kidney dysfunction beyond end-stage kidney disease remains uncertain. Based on nationwide data from multiple health care registers, the study aims to evaluate vaccination effect on rates of SARS-CoV-2 infection and subsequent risk of associated mortality in patients with kidney disease beyond end-stage kidney disease. Method A population-based retrospective cohort study on all Danish residents ≥18 years with an eGFR ≤90 ml/min/1.73 m2 identified between February 1st 2021 and June 27th 2022 corresponding to predominance of Alpha-, Delta-, and Omicron-variant. Outcomes compared based on vaccination state (unvaccinated, vaccinated, and booster vaccinated) across predefined eGFR categories (eGFR 60-90 mL/min/1.73 m2; eGFR 45-59 mL/min/1.73 m2, eGFR 30-44 mL/min/1.73 m2; eGFR &amp;lt;30 mL/min/1.73 m2). Risk of infection reported as age- and gender-standardized rate ratios of positive SARS-CoV-2 PCR and subsequent risk of death reported as age- and gender standardized 30-day mortality risk. Standardized risks computed based on hazards obtained in multiple cause-specific Cox regression models. Results A total of 982,047 adults with eGFR ≤90 ml/min/1.73 m2 identified. Gender distribution was 43.9% male, median age was 71 [IQR 61–78] years, median eGFR 75 [IQR 63-84], and 13.1% diabetes. Population rates of vaccination were 3.5%, 80.1%, and 95.4% for Alpha-, Delta-, and Omicron-variant, with 66.2% booster vaccinated at Omicron-variant baseline (Pfizer BionTech 88.3%, Moderna 9.4%, other vaccinations 2.3%). In total 359,428 SARS-CoV-2 infections were identified (Alpha: n = 7,861, Delta: n = 27,868, Omicron: n = 323,699), corresponding to a crude rate of positive SARS-CoV-2 PCR test of 629.0 per 100,000 person-weeks overall. Vaccination and booster vaccination were associated with progressively lower rate of SARS-CoV-2 infection. Age- and gender-standardized rate ratios are shown in Fig. 1. Overall, mean 30-day risks of death were 7.1‰ [95% CI 6.7-7.6], 3.8‰ [95% CI 3.5-4.0], and 1.4‰ [95% CI 1.3-1.4] in unvaccinated, vaccinated, and booster vaccinated persons, respectively. Vaccination and booster vaccination were associated with benefit on 30-day mortality across all strata of eGFR despite progressive increase in risk of death observed with declining kidney function. Age- and gender standardized 30-day mortality rates following a positive PCR for SARS-CoV-2 across granular eGFR are shown in Fig. 2. Conclusion Based on comprehensive nationwide data covering Alpha-, Delta-, and Omicron-variant predominance in Denmark, vaccination and booster vaccination against COVID-19 were associated with lower rates of SARS-CoV-2 infection and benefit on subsequent risk of 30-day mortality across all strata of eGFR.

NEM-GNN: DAC/ADC-less, Scalable, Reconfigurable, Graph and Sparsity-Aware Near-Memory Accelerator for Graph Neural Networks

Graph neural networks (GNNs) are of great interest in real-life applications such as citation networks and drug discovery owing to GNN’s ability to apply machine learning techniques on graphs. GNNs utilize a two-step approach to classify the nodes in a graph into pre-defined categories. The first step uses a combination kernel to perform data-intensive convolution operations with regular memory access patterns. The second step uses an aggregation kernel that operates on sparse data having irregular access patterns. These mixed data patterns render CPU/GPU-based compute energy-inefficient. Von Neumann based accelerators like AWB-GCN [ 7 ] suffer from increased data movement, as the data-intensive combination requires large data movement to/from memory to perform computations. ReFLIP [ 8 ] performs resistive random access memory based in-memory (PIM) compute to overcome data movement costs. However, ReFLIP suffers from increased area requirement due to dedicated accelerator arrangement, and reduced performance due to limited parallelism and energy due to fundamental issues in ReRAM-based compute. This article presents a scalable (non-exponential storage requirement), DAC/ADC-less PIM-based combination, with (i) early compute termination and (ii) pre-compute by reconfiguring SOC components. Graph and sparsity-aware near-memory aggregation using the proposed compute-as-soon-as-ready (CAR) broadcast approach improves performance and energy further. NEM-GNN achieves ∼80–230x, ∼80–300x, ∼850–1,134x, and ∼7–8x improvement over ReFLIP, in terms of performance, throughput, energy efficiency, and compute density.

Analisis Pengendalian Kualitas dengan Metode Six Sigma Pada Produksi Packaging PT. ABC

The development of industry in this current era is progressing rapidly, marked by increasing competition among companies. The method used is the Six Sigma method with the DMAIC approach (Define, Measure, Analyze, Improve, and Control). With this method, data obtained includes a total number of both good and defective products, which are then classified into several predefined defect categories by the company. Based on the data analysis results from July 2022 to December 2022, PT ABC identified four types of defects in packaging products. These defects include sticking with a defect percentage of 54%, cracking at 27%, coloring at 18%, and scratching at 2%. Improvement efforts can be made through training, guidance, and supervision of operators to enhance product quality. Additionally, strict supervision of raw materials in packaging production is necessary to ensure their quality is maintained. Furthermore, the machine setting process also requires supervision and guidance, with the implementation of clear Standard Operating Procedures (SOP) for each product. Keywords: DMAIC, Quality Control, Six Sigma

A Comprehensive Analysis of Exploring the Efficacy of Machine Learning Algorithms in Text, Image, and Speech Analysis

Our study demonstrates a comprehensive investigation of multiple machine learning models based on text, image, and speech analysis. More specifically, concerning text analysis, we studied the application of such models as Recurrent Neural Networks, Gated Recurrent Units, and Bidirectional Encoder Representations from Transformers in classifying documents into pre-defined categories. The outcomes revealed that RNNs observed the highest precision, recall, F1 score, as well as the highest accuracy. Importantly, the models are well-suited to detect sequential dependencies and create semantic representations on the basis of textual data. With regard to image analysis, we found out that Convolutional Neural Networks were the best model. At the same time, VGG16 and GANs also demonstrated rather promising results suggesting that deep learning is paramount to extract significant data features. As far as speech analysis is concerned, we found out that CNNs are exceptional in terms of accuracy to recognize speech patterns in comparison with the other models. Simultaneously, LSTM also observed a high level of accuracy allowing to capture temporal dependencies in audio signals. In conclusion, the findings of our study suggest that it is exigent to identify an appropriate machine learning model depending on the task and the selected dataset. It is also crucial to understand the nature of each of the studied models to assess their applicability for a specific task. Moreover, our study might be valuable for other researchers given that it contributes to the development of the field of deep learning and, thus, promotes the emergence of new applications in different domains.

Debiased Novel Category Discovering and Localization

In recent years, object detection in deep learning has experienced rapid development. However, most existing object detection models perform well only on closed-set datasets, ignoring a large number of potential objects whose categories are not defined in the training set. These objects are often identified as background or incorrectly classified as pre-defined categories by the detectors. In this paper, we focus on the challenging problem of Novel Class Discovery and Localization (NCDL), aiming to train detectors that can detect the categories present in the training data, while also actively discover, localize, and cluster new categories. We analyze existing NCDL methods and identify the core issue: object detectors tend to be biased towards seen objects, and this leads to the neglect of unseen targets. To address this issue, we first propose an Debiased Region Mining (DRM) approach that combines class-agnostic Region Proposal Network (RPN) and class-aware RPN in a complementary manner. Additionally, we suggest to improve the representation network through semi-supervised contrastive learning by leveraging unlabeled data. Finally, we adopt a simple and efficient mini-batch K-means clustering method for novel class discovery. We conduct extensive experiments on the NCDL benchmark, and the results demonstrate that the proposed DRM approach significantly outperforms previous methods, establishing a new state-of-the-art.

Compound Text-Guided Prompt Tuning via Image-Adaptive Cues

Vision-Language Models (VLMs) such as CLIP have demonstrated remarkable generalization capabilities to downstream tasks. However, existing prompt tuning based frameworks need to parallelize learnable textual inputs for all categories, suffering from massive GPU memory consumption when there is a large number of categories in the target dataset. Moreover, previous works require to include category names within prompts, exhibiting subpar performance when dealing with ambiguous category names. To address these shortcomings, we propose Compound Text-Guided Prompt Tuning (TGP-T) that significantly reduces resource demand while achieving superior performance. We introduce text supervision to the optimization of prompts, which enables two benefits: 1) releasing the model reliance on the pre-defined category names during inference, thereby enabling more flexible prompt generation; 2) reducing the number of inputs to the text encoder, which decreases GPU memory consumption significantly. Specifically, we found that compound text supervisions, i.e., category-wise and content-wise, is highly effective, since they provide inter-class separability and capture intra-class variations, respectively. Moreover, we condition the prompt generation on visual features through a module called Bonder, which facilitates the alignment between prompts and visual features. Extensive experiments on few-shot recognition and domain generalization demonstrate that TGP-T achieves superior performance with consistently lower training costs. It reduces GPU memory usage by 93% and attains a 2.5% performance gain on 16-shot ImageNet. The code is available at https://github.com/EricTan7/TGP-T.

Automated Cardiovascular Disease Diagnosis from X-ray Images using Convolutional Neural Networks

Problem Cardiovascular Disease (CVD) is a leading cause of mortality worldwide, and its early and accurate diagnosis is crucial for effective treatment and patient care. Medical imaging, particularly X-ray imaging, plays a crucial role in the detection and assessment of cardiovascular abnormalities. In recent years, Convolutional Neural Networks (CNNs) have emerged as a powerful tool in medical image analysis, demonstrating promising results in various diagnostic tasks. Proposed Approach This research paper investigates the application of CNNs for the automated diagnosis of CVD from X-ray images. The CVD diagnosis framework proposed in this study consists of three key modules. The first module is an X-ray feature extractor built using a state-of-the-art CNN architecture. The second module is an age prediction component, which accurately estimates the age of the patients from the X-ray images. Finally, the third module is the CVD classifier, which categorizes the input X-ray images into four predefined severity categories of Cardiovascular Disease. Result Through extensive experiments, the proposed method has demonstrated its capability to offer novel insights into the potential use of X-ray images for predicting systemic biomarkers in the diagnosis of CVD. I expect that the proposed CVD diagnosis method can provide a significant advancement in the field of cardiovascular healthcare by offering an accurate, efficient, and automated solution for early detection of CVD.

Classification the Characteristics of Traffic Accident Victims in Pariaman Using the Chi-square Automatic Interaction Detection Algorithm

Traffic accidents are incidents that occur when motor vehicles collide on the road, resulting in damage to vehicles and road infrastructure, as well as the potential for material losses, injuries, physical damage, and even death for those involved. Data from the Indonesian National Police show that the number of traffic accident victims between 2010 and 2020 ranged from 147.798 to 197.560 people, with fatalities predominantly occurring among individuals aged 15-34. The high number of traffic accident victims has negative impacts on various aspects of life, ranging from material losses to physical damage to the victims. Classification is a technique used to group objects or data into pre-defined classes or categories based on their attributes or features. One method in the field of classification is Chi-Square Automatic Interaction Detection (CHAID). The results of the classification using this method indicate that the age of the victims and the type of accident are the most significant variables influencing the condition of traffic accident victims. The evaluation of the model using a confusion matrix yielded an accuracy rate of 92%. This indicates that the model performs well in overall data classification.

Internal Medicine Resident Addiction Training at the Veteran's Health Administration: A Qualitative Evaluation of Site Directors' Response to the 2022 ACGME Requirements.

Substance use disorders (SUDs) are prevalent in the USA yet remain dramatically undertreated. To address this care gap, the Accreditation Council for Graduate Medical Education (ACGME) approved revisions to the Program Requirements for Graduate Medical Education (GME) in Internal Medicine, effective July 1, 2022, requiring addiction medicine training for all internal medicine (IM) residents. The Veterans Health Administration (VHA) is a clinical training site for many academic institutions that sponsor IM residencies. This focus group project evaluated VHA IM residency site directors' perspectives about providing addiction medical education within VHA IM training sites. To better understand the current state, barriers to, and facilitators of IM resident addiction medicine training at VHA sites. This was a qualitative evaluation based on semi-structured video-based focus groups. Participants were VHA IM site directors based at a VHA hospital or clinic throughout the USA. Focus groups were conducted using a semi-structured group interview guide. Two investigators coded each focus group independently, then met to create a final adjudicated coding scheme. Thematic analysis was used to identify key themes. Forty-three participants from 38 VHA sites participated in four focus groups (average size: 11 participants). Six themes were identified within four pre-defined categories. Current state of training: most VHA sites offered no formal training in addiction medicine for IM residents. Barriers: addiction experts are often located outside of IM settings, and ACGME requirements were non-specific. Facilitators: clinical champions help support addiction training. Desired next steps: participants desired incentives to train or hire local champions and a pre-packaged didactic curriculum. Developing competent clinical champions and leveraging VHA addiction specialists from non-IM settings would create more addiction training opportunities for IM trainees at VHA sites. These insights can likely be applied to IM training at non-VHA sites.

Effectiveness and Cost-Effectiveness of Self-Management Interventions for Adults Living with Heart Failure to Improve Patient-Important Outcomes: An Evidence Map of Randomized Controlled Trials

Self-management interventions (SMIs) may enhance heart failure (HF) outcomes and address challenges associated with disease management. This study aims to review randomized evidence and identify knowledge gaps in SMIs for adult HF patients. Within the COMPAR-EU project, from 2010 to 2018, we conducted searches in the databases MEDLINE, CINAHL, Embase, Cochrane, and PsycINFO. We performed a descriptive analysis using predefined categories and developed an evidence map of randomized controlled trials (RCTs). We found 282 RCTs examining SMIs for HF patients, comparing two to four interventions, primarily targeting individual patients (97%) globally (34 countries, only 31% from an European country). These interventions involved support techniques such as information sharing (95%) and self-monitoring (62%), often through a mix of in-person and remote sessions (43%). Commonly assessed outcomes included quality of life, hospital admissions, mortality, exercise capacity, and self-efficacy. Few studies have focused on lower socio-economic or minority groups. Nurses (68%) and physicians (30%) were the primary providers, and most studies were at low risk of bias in generating a random sequence for participant allocation; however, the reporting was noticeably unclear of methods used to conceal the allocation process. Our analysis has revealed prevalent support techniques and delivery methods while highlighting methodological challenges. These findings provide valuable insights for researchers, clinicians, and policymakers striving to optimize SMIs for individuals living with HF.

Siamese-DETR for Generic Multi-Object Tracking.

The ability to detect and track the dynamic objects in different scenes is fundamental to real-world applications, e.g., autonomous driving and robot navigation. However, traditional Multi-Object Tracking (MOT) is limited to track objects belonging to the pre-defined closed-set categories. Recently, Generic MOT (GMOT) is proposed to track interested objects beyond pre-defined categories and it can be divided into Open-Vocabulary MOT (OVMOT) and Template-Image-based MOT (TIMOT). Taking the consideration that the expensive well pre-trained (vision-)language model and fine-grained category annotations are required to train OVMOT models, in this paper, we focus on TIMOT and propose a simple but effective method, Siamese-DETR. Only the commonly used detection datasets (e.g., COCO) are required for training. Different from existing TIMOT methods, which train a Single Object Tracking (SOT) based detector to detect interested objects and then apply a data association based MOT tracker to get the trajectories, we leverage the inherent object queries in DETR variants. Specifically: 1) The multi-scale object queries are designed based on the given template image, which are effective for detecting different scales of objects with the same category as the template image; 2) A dynamic matching training strategy is introduced to train Siamese-DETR on commonly used detection datasets, which takes full advantage of provided annotations; 3) The online tracking pipeline is simplified through a tracking-by-query manner by incorporating the tracked boxes in the previous frame as additional query boxes. The complex data association is replaced with the much simpler Non-Maximum Suppression (NMS). Extensive experimental results show that Siamese-DETR surpasses existing MOT methods on GMOT-40 dataset by a large margin.

The Value of Generative AI for Qualitative Research: A Pilot Study

This mixed methods approach study investigates the potential of introducing Generative AI (ChatGPT 4 and Bard) as part of a deductive qualitative research design that requires coding, focusing on possible gains in cost-effectiveness, coding throughput time and inter-coder reliability (Cohen’s Kappa). This study involved semi-structured interviews with five domain experts and analyzed a dataset of 122 respondents that required categorization into six pre-defined categories. The results from using Generative AI coders were compared with those from a previous study where human coders carried out the same task. In this comparison, we evaluated the performance of AI-based coders against two groups of human coders, comprising three experts and three non-experts. Our findings support the replacement of human coders with Generative AI ones, specifically ChatGPT for deductive qualitative research methods of limited scope. The experimental group, consisting of three independent Generative AI coders, outperformed both control groups in coding effort, with a fourfold (4x) efficiency and throughput time (15x) advantage. The latter could be explained by leveraging parallel processing. Concerning expert vs. non-expert coders, minimal evidence suggests a preference for experts. Although experts code slightly faster (17%), their inter-coder reliability showed no substantial advantage. A hybrid approach, combing ChatGPT and domain experts shows the most promise. This approach reduces costs, shortens project timelines, and enhances inter-coder reliability, as indicated by higher Cohen's Kappa values. In conclusion, Generative AI, exemplified by ChatGPT, offers a viable alternative to human coders, in combination with human research involvement, delivering cost savings and faster research completion without sacrificing notably reliability. These insights, while limited in scope, show potential for further studies with lager datasets, more inductive qualitative research designs and other research domains.