Task Domain Research Articles

Graph correlated discriminant embedding for multi-source domain adaptation

As a main branch of domain adaptation (DA), multi-source DA (MSDA) has attracted increasing attention for exploiting information from multi-source domain data. However, how to effectively explore useful information from each source domain for target tasks is still a key problem. In this paper, to fully explore multiple information of different domain data, we propose a graph correlated discriminant embedding (GCDE) method for MSDA. In GCDE, the category-discriminative information, manifold structure, and correlation learning are fully considered. Specifically, GCDE encodes the within- and between- class information of each domain data, preserves the local and global structure information of the data, and extracts the maximization correlative features from different domains by designing a novel correlative learning scheme. We also extend GCDE to a nonlinear case and obtain kernel GCDE (KGCDE). We have conducted extensive experiments on four public data benchmarks to verify the performance of GCDE and KGCDE. The promising performance on the databases prove the efficiency of our methods with the comparison of the advanced approaches.

Metallic surface defect recognition network based on global feature aggregation and dual context decoupled head

Surface defect detection is a crucial inspection phase in ensuring industrial product quality and reliability, particularly in the context of metallic manufacturing components. While existing deep learning-based approaches have demonstrated some effectiveness, the design of certain network structures still lacks consideration for industrial scenarios. This paper proposed a complex metallic surface defect detection neural network which incorporates two innovatively designed modules along with other targeted enhancements. Firstly, the spatial pyramid pooling (SPP) structure is redesigned for richer global-local information fusion. A global feature fusion and redistribution module (GFAR) is designed to fully leverage feature information across different scales. With the combination of a novel global positional attention mechanism (GPAM), GFAR is capable of distinguishing defects from complex backgrounds. Moreover, the asymmetric convolution blocks (ACB) are employed to process the final features, enhancing the representational capacity for intricate defect shapes. Subsequently, the detection head is decoupled in both spatial and task domains to addresses the problems posed by inconsistent focus between classification and regression tasks. Finally, the reassignment of loss functions is undertaken for each respective task to enhance their alignment with the specific requirements of the defect detection task. Extensive ablation experiments are conducted on two steel surface defect datasets, NEU-DET and GC10-DET, to demonstrate the effectiveness of each module. The defects detection accuracy of our method respectively reaches 80.9 % and 84.8 % for the two datasets, which outperforms the majority of state-of-the-art detection neural networks.

Contextual design requirements for decision-support tools involved in weaning patients from mechanical ventilation in intensive care units

Weaning patients from ventilation in intensive care units (ICU) is a complex task. There is a growing desire to build decision-support tools to help clinicians during this process, especially those employing Artificial Intelligence (AI). However, tools built for this purpose should fit within and ideally improve the current work environment, to ensure they can successfully integrate into clinical practice. To do so, it is important to identify areas where decision-support tools may aid clinicians, and associated design requirements for such tools. This study analysed the work context surrounding the weaning process from mechanical ventilation in ICU environments, via cognitive task and work domain analyses. In doing so, both what cognitive processes clinicians perform during weaning, and the constraints and affordances of the work environment itself, were described. This study found a number of weaning process tasks where decision-support tools may prove beneficial, and from these a set of contextual design requirements were created. This work benefits researchers interested in creating human-centred decision-support tools for mechanical ventilation that are sensitive to the wider work system.

Spectral Prompt Tuning: Unveiling Unseen Classes for Zero-Shot Semantic Segmentation

Recently, CLIP has found practical utility in the domain of pixel-level zero-shot segmentation tasks. The present landscape features two-stage methodologies beset by issues such as intricate pipelines and elevated computational costs. While current one-stage approaches alleviate these concerns and incorporate Visual Prompt Training (VPT) to uphold CLIP's generalization capacity, they still fall short in fully harnessing CLIP's potential for pixel-level unseen class demarcation and precise pixel predictions. To further stimulate CLIP's zero-shot dense prediction capability, we propose SPT-SEG, a one-stage approach that improves CLIP's adaptability from image to pixel. Specifically, we initially introduce Spectral Prompt Tuning (SPT), incorporating spectral prompts into the CLIP visual encoder's shallow layers to capture structural intricacies of images, thereby enhancing comprehension of unseen classes. Subsequently, we introduce the Spectral Guided Decoder (SGD), utilizing both high and low-frequency information to steer the network's spatial focus towards more prominent classification features, enabling precise pixel-level prediction outcomes. Through extensive experiments on two public datasets, we demonstrate the superiority of our method over state-of-the-art approaches, performing well across all classes and particularly excelling in handling unseen classes.

Cerebello-Hippocampal Interactions in the Human Brain: A New Pathway for Insights Into Aging.

The cerebellum is recognized as being important for optimal behavioral performance across task domains, including motor function, cognition, and affect. Decades of work have highlighted cerebello-thalamo-cortical circuits, from both structural and functional perspectives. However, these circuits of interest have been primarily (though not exclusively) focused on targets in the cerebral cortex. In addition to these cortical connections, the circuit linking the cerebellum and hippocampus is of particular interest. Recently, there has been an increased interest in this circuit, thanks in large part to novel findings in the animal literature demonstrating that neuronal firing in the cerebellum impacts that in the hippocampus. Work in the human brain has provided evidence for interactions between the cerebellum and hippocampus, though primarily this has been in the context of spatial navigation. Given the role of both regions in cognition and aging, and emerging evidence indicating that the cerebellum is impacted in age-related neurodegenerative disease such as Alzheimer's, I propose that further attention to this circuit is warranted. Here, I provide an overview of cerebello-hippocampal interactions in animal models and from human imaging and outline the possible utility of further investigations to improve our understanding of aging and age-related cognitive decline.

Type of bilingualism, task domain, and attention demands impact children's cognitive performance.

The present study investigated the relation between continuous measures of two qualitatively different types of bilingual experience and outcome measures that varied in domain (verbal or nonverbal) and processing demands (degree of conflict). Participants were 195 English-speaking children, 7 years old, who were enrolled in French immersion programs. Children were assessed for their degree of "home bilingualism," reflecting language use patterns at home, and "school bilingualism," reflecting progress in learning French. None of the children spoke French at home. Participants completed verbal fluency, n-back, and global-local tasks, with conditions within tasks varying in degree of attentional processing required. Type of bilingual experience affected verbal and nonverbal tasks differently. Home bilingualism was positively associated with performance on attentionally demanding conditions of the n-back and global-local tasks, the two nonverbal tasks, whereas school bilingualism was positively associated with performance on English verbal fluency, the linguistic task. In both cases, results were modulated by the degree of bilingual experience. These results underline the importance of specifying the type and degree of bilingual experience and the details of the outcome tasks to understand the impact of bilingualism on children's development. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Optimizing distortion magnitude for data augmentation in few-shot remote sensing scene classification

ABSTRACT In the domain of few-shot classification tasks for remote sensing images, the impact of data augmentation is often overlooked, and conventional methods offer only modest gains in classification performance. To better exploit the potential of data augmentation in few-shot learning, we propose an innovative data augmentation method tailored to optimize few-shot remote sensing scene classification. We begin by dissecting the relationship between various types of feature distortions and classification performance, introducing an optimal distortion magnitude estimation method for different feature types. Subsequently, we integrate multiple distortion magnitude optimization pathways into the model learning process, achieving a dual optimization of model parameters and distortion magnitudes. The extended data provide the model with distorted samples from multiple feature perspectives, and the dynamic optimization of distortion magnitudes enhances the effectiveness of the extended data for classification. Ultimately, we evaluate the performance of our method on general remote sensing datasets, demonstrating a significant advantage in classification accuracy over baseline methods. This approach offers a new perspective in robustness research for few-shot remote sensing scene classification models based on data augmentation.

The Efficacy of Autoencoders in the Utilization of Tabular Data for Classification Tasks

This article explores the effectiveness of applying auto encoders in the context of classical machine learning tasks, particularly classification tasks on tabular data. Significant emphasis is placed on a detailed comparison of this approach with other widely adopted methods in the field. As an applied task, the creation of a look-alike model for targeted advertising is pursued. Research into the requirements for this model reveals that conventional approaches to processing tabular data and model training cannot yield a solution that qualitatively meets all specified criteria. Evaluation is conducted on a closed dataset provided by a telecommunications operator. This de-identified dataset includes subscriber descriptions, their tariff plans, and mobile devices. The research goal is to assess non-traditional approaches from the domain of deep learning and neural networks in the context of classification tasks on tabular data and compare them with established methods in this context. The research findings demonstrate that the investigated approach yields a solution that surpasses traditional approaches across all specified criteria. Furthermore, the study provides insights into the universality of this approach, indicating its applicability in various domains for tabular data classification tasks. The ability to compress information into vector representations without significant loss and identify new hidden nonlinear dependencies in data enhances the practical value of the approach. Future research will address the formation of multiple autoencoders in different knowledge domains and their integration for specific tasks. Additionally, efforts will focus on identifying the most effective algorithm for finding similar subscribers based on the vector representations obtained from autoencoders.

An Intersubject Brain-Computer Interface Based on Domain-Adversarial Training of Convolutional Neural Network.

Attention decoding plays a vital role in daily life, where electroencephalography (EEG) has been widely involved. However, training a universally effective model for everyone is impractical due to substantial interindividual variability in EEG signals. To tackle the above challenge, we propose an end-to-end brain-computer interface (BCI) framework, including temporal and spatial one-dimensional (1D) convolutional neural network and domain-adversarial training strategy, namely DA-TSnet. Specifically, DA-TSnet extracts temporal and spatial features of EEG, while it is jointly supervised by task loss and domain loss. During training, DA-TSnet aims to maximize the domain loss while simultaneously minimizing the task loss. We conduct an offline analysis, simulate online experiments on a self-collected dataset of 85 subjects, and real online experiments on 22 subjects. DA-TSnet achieves a leave-one-subject-out (LOSO) cross-validation (CV) classification accuracy of 89.40% ± 9.96%, outperforming several state-of-the-art attention EEG decoding methods. In simulated online experiments, DA-TSnet achieves an outstanding accuracy of 88.07% ± 11.22%. In real online experiments, it achieves an average accuracy surpassing 86%. An end-to-end network framework does not rely on elaborate preprocessing and feature extraction steps, which saves time and human workload. Moreover, our framework utilizes domain-adversarial training neural network (DANN) to tackle the challenge posed by the high interindividual variability in EEG signals, which has significant reference value for handling other EEG signal decoding issues. Last, the performance of the DA-TSnet framework in offline and online experiments underscores its potential to facilitate more reliable applications.

Nursing students’ awareness and performance of patient medical information protection

Purpose: This study aims to understand the awareness, and practices of nursing students regarding patient medical information protection (PMIP). Methods: This study is a descriptive correlational research to determine the relationship between awareness and practice of PMIP. Data collection took place from August 1 to 15, 2022, with 119 nursing students from a nursing college in Daegu, South Korea. Differences in awareness and practice of PMIP according to general characteristics were analyzed using independent t-tests and one-way ANOVA. The correlation between awareness and practices was analyzed using Pearson correlation. Results: The awareness for PMIP of nursing students (4.35±0.43) was significantly higher than the performance score (4.20±0.57). By domain, the Awareness in the management domain was the highest, linked with the performance in the referral of tasks domain. However, the perception and performance scores were lower in the communication domain. PMIP awareness showed significant differences based on education type, direct lectures, and religion. Performance were higher when the education venue was the university. There was a strong correlation between nursing students’ awareness and practice of PMIP. Conclusion: To enhance nursing students’ awareness and practices in patient personal information protection, the development of educational programs reflecting the higher effectiveness of school-based education compared to hospital-based education is necessary.

Do teacher beliefs mediate leadership and teacher behaviors? Testing teacher self-efficacy's mediation role between leadership for learning and teacher outcomes

PurposeLeadership for learning emerged as an integrated leadership framework; however, attempts to establish an empirical measurement model have been limited. Critically, not much is known about how much teachers' beliefs (e.g. self-efficacy) can mediate leadership for learning impact on teacher behaviors. This study establishes a leadership for learning measurement model and examines whether teacher self-efficacy mediates the effect of leadership for learning tasks on teacher collaboration, instructional quality, intention to leave current schools and their confidence in equitable teaching practice.Design/methodology/approachDrawing on the most recent 2018 Teaching and Learning International Survey (TALIS), the study employed a structural equation modeling mediation approach.FindingsResults suggested that teacher self-efficacy statistically significantly mediated 16 out of 20 of the relationships between leadership for learning task domains and teacher outcomes. Especially, in explaining the variance in instructional quality and teacher confidence in implementing equitable teaching practices, considerable proportions of the predictive power of leadership for learning tasks were accounted for (i.e. mediated) by teacher self-efficacy.Research limitations/implicationsSchool-wide efforts to craft the school vision for learning must be coupled with enhancing teacher self-efficacy. Critically, leadership efforts may fall short of implementing equitable teaching practice and quality instruction without addressing teacher confidence in their ability in instruction, classroom management and student engagement.Originality/valueThis study is the first of its kind to evidence teacher self-efficacy mediates leadership for learning practice impact on teacher behaviors.

Small Target-YOLOv5: Enhancing the Algorithm for Small Object Detection in Drone Aerial Imagery Based on YOLOv5.

Object detection in drone aerial imagery has been a consistent focal point of research. Aerial images present more intricate backgrounds, greater variation in object scale, and a higher occurrence of small objects compared to standard images. Consequently, conventional object detection algorithms are often unsuitable for direct application in drone scenarios. To address these challenges, this study proposes a drone object detection algorithm model based on YOLOv5, named SMT-YOLOv5 (Small Target-YOLOv5). The enhancement strategy involves improving the feature fusion network by incorporating detection layers and implementing a weighted bidirectional feature pyramid network. Additionally, the introduction of the Combine Attention and Receptive Fields Block (CARFB) receptive field feature extraction module and DyHead dynamic target detection head aims to broaden the receptive field, mitigate information loss, and enhance perceptual capabilities in spatial, scale, and task domains. Experimental validation on the VisDrone2021 dataset confirms a significant improvement in the target detection accuracy of SMT-YOLOv5. Each improvement strategy yields effective results, raising the average precision by 12.4 percentage points compared to the original method. Detection improvements for large, medium, and small targets increase by 6.9%, 9.5%, and 7.7%, respectively, compared to the original method. Similarly, applying the same improvement strategies to the low-complexity YOLOv8n results in SMT-YOLOv8n, which is comparable in complexity to SMT-YOLOv5s. The results indicate that, relative to SMT-YOLOv8n, SMT-YOLOv5s achieves a 2.5 percentage point increase in average precision. Furthermore, comparative experiments with other enhancement methods demonstrate the effectiveness of the improvement strategies.

Generation-based Differential Fuzzing for Deep Learning Libraries

Deep learning (DL) libraries have become the key component in developing and deploying DL-based software nowadays. With the growing popularity of applying DL models in both academia and industry across various domains, any bugs inherent in the DL libraries can potentially cause unexpected server outcomes. As such, there is an urgent demand for improving the software quality of DL libraries. Although there are some existing approaches specifically designed for testing DL libraries, their focus is usually limited to one specific domain, such as computer vision (CV). It is still not very clear how the existing approaches perform in detecting bugs of different DL libraries regarding different task domains and to what extent. To bridge this gap, we first conduct an empirical study on four representative and state-of-the-art DL library testing approaches. Our empirical study results reveal that it is hard for existing approaches to generalize to other task domains. We also find that the test inputs generated by these approaches usually lack diversity, with only a few types of bugs. What is worse, the false-positive rate of existing approaches is also high ( up to 58% ). To address these issues, we propose a guided differential fuzzing approach based on generation , namely, Gandalf . To generate testing inputs across diverse task domains effectively, Gandalf adopts the context-free grammar to ensure validity and utilizes a Deep Q-Network to maximize the diversity. Gandalf also includes 15 metamorphic relations to make it possible for the generated test cases to generalize across different DL libraries. Such a design can decrease the false positives because of the semantic difference for different APIs. We evaluate the effectiveness of Gandalf on nine versions of three representative DL libraries, covering 309 operators from computer vision, natural language processing, and automated speech recognition. The evaluation results demonstrate that Gandalf can effectively and efficiently generate diverse test inputs. Meanwhile, Gandalf successfully detects five categories of bugs with only 3.1% false-positive rates. We report all 49 new unique bugs found during the evaluation to the DL libraries’ developers, and most of these bugs have been confirmed. Details about our empirical study and evaluation results are available on our project website. 1

FDNet: Feature decoupling for single-stage pose estimation in complex scenes

Severe error accumulation is a key reason that makes existing multi-person pose estimation challenging in complex scenes. Single-stage pose estimation methods detect human and keypoints in parallel, avoiding serial error accumulation compared to two-stage methods. However, two distinct tasks share identical features in the existing method, which causes error accumulation. Therefore, we propose a Feature Decoupling Network (FDNet) to further extend the pipeline of single-stage methods and reconstruct task-specific features, including Feature Decoupling Module (FDM), Human Aware Loss (HAL) and Keypoint Aware Loss (KAL). FDM can adaptively perceive spatial and channel features and allocate separate feature domains for various tasks. To learn distinctive feature representation for compact human bodies, HAL and KAL measure and suppress feature similarities and distances for different human and keypoints, thus alienating feature interference. Experiments on crowded datasets show that our method is superior, outperforming the state-of-the-art method CID by 1.5%, 0.8% on OCHuman and CrowdPose.

Distinct Contributions of the Cerebellum and Basal Ganglia to Arithmetic Procedures.

Humans exhibit complex mathematical skills attributed to the exceptional enlargement of neocortical regions throughout evolution. In the current work, we initiated a novel exploration of the ancient subcortical neural network essential for mathematical cognition. Using a neuropsychological approach, we report that degeneration of two subcortical structures, the cerebellum and basal ganglia, impairs performance in symbolic arithmetic. We identify distinct computational impairments in male and female participants with cerebellar degeneration (CD) or Parkinson's disease (PD). The CD group exhibited a disproportionate cost when the arithmetic sum increased, suggesting that the cerebellum is critical for iterative procedures required for calculations. The PD group showed a disproportionate cost for equations with increasing addends, suggesting that the basal ganglia are critical for chaining multiple operations. In Experiment 2, the two patient groups exhibited intact practice gains for repeated equations at odds with an alternative hypothesis that these impairments were related to memory retrieval. Notably, we discuss how the counting and chaining operations relate to cerebellar and basal ganglia function in other task domains (e.g., motor processes). Overall, we provide a novel perspective on how the cerebellum and basal ganglia contribute to symbolic arithmetic. Our studies demonstrate the constraints on the computational role of two subcortical regions in higher cognition.

Interdisciplinary Aspects of Synaesthesia

Synaesthesia is a multifaceted phenomenon encompassing genetic, neurochemical, and cognitive elements, contributing to a deeper understanding of how our senses and perceptions are intricately intertwined. This paper approaches an interdisciplinary approach to a current subject in cognitive psychology, examined by specialists from diverse fields: cognitive psychologists, neurologists, philologists, musicologists, etc. Nevertheless, the diverse manifestations of synaesthesia and its phenomenological nature have posed challenges in verifying and studying its origins and mechanisms. This article adopts a medical standpoint in its examination of synesthesia, with a particular focus on its evolutionary trends and its influence within the domain of musical tasks.

Expertise in Non-Well-Defined Task Domains: The Case of Reading

ABSTRACT In this article, we discuss expertise by considering the activity of reading. Cognitive scientists have traditionally conceptualised reading as a single, well-defined task, namely the decoding of letter sequences into meaningful sequences of speech sounds. This definition captures a core feature of the reading activity at the computational level, but it is an overly narrow model of how reading behaviour occurs in the real world. We propose a more expansive model of expertise. In our view, expertise in general is best conceptualised as a distributed process that takes place within a cultural-cognitive ecosystem. Our model allows for a more inclusive view of expertise in reading. We argue that reading is better understood as a manifold task domain that admits multiple reasonable criteria for evaluating performance. We draw on ethnographic data to show how our model allows for a wider appreciation of what expertise in reading amounts to in educational contexts.

A position-aware attention network with progressive detailing for land use semantic segmentation of Remote Sensing images

ABSTRACT Deep learning has achieved remarkable success in the semantic segmentation of remote sensing images (RSIs).In the domain of semantic segmentation, where classification and localization tasks need to be performed simultaneously, it is crucial to consider both global and local spatial relationships in RSIs. This is especially important for the recognition of ground objects that have a slim and elongated appearance. However, existing methods for land use semantic segmentation lack an effective mechanism to coordinate and address these two aspects, resulting in limitations on the recognition of slim targets and the continuity of land object identification. Here, a specific attention-based network called PaANet is developed for semantic segmentation. Our proposed framework builds upon the Swin transformer by incorporating two key modules: the position-aware attention (PaA) module and the pyramid pooling expectation-maximization (PPEM) module. These modules provide significant improvements in recognition accuracy and the continuity of ground object recognition while preserving structural classification details. Furthermore, we propose a multiresolution data augmentation method that utilizes scale-related information to guide the encoder. This approach leads to improved performance and generalization ability for the model. In experiments, the mIoU of our approach for the BLU and GID datasets is 2.37% and 3.94% higher than that of the baseline model, respectively. Our results also show significant superior to those of other methods regarding the continuity of ground object recognition.

Adverse Life Experiences and Brain Function

Adverse life experiences have been proposed to contribute to diverse mental health problems through an association with corticolimbic functioning. Despite compelling evidence from animal models, findings from studies in humans have been mixed; activation likelihood estimation (ALE) meta-analyses have failed to identify a consistent association of adverse events with brain function. To investigate the association of adversity exposure with altered brain reactivity using multilevel kernel density analyses (MKDA), a meta-analytic approach considered more robust than ALE to small sample sizes and methodological differences between studies. Searches were conducted using PsycInfo, Medline, EMBASE, and Web of Science from inception through May 4, 2022. The following search term combinations were used for each database: trauma, posttraumatic stress disorder (PTSD), abuse, maltreatment, poverty, adversity, or stress; and functional magnetic resonance imaging (fMRI) or neuroimaging; and emotion, emotion regulation, memory, memory processing, inhibitory control, executive functioning, reward, or reward processing. Task-based fMRI studies within 4 domains (emotion processing, memory processing, inhibitory control, and reward processing) that included a measure of adverse life experiences and whole-brain coordinate results reported in Talairach or Montreal Neurological Institute space were included. Conference abstracts, books, reviews, meta-analyses, opinions, animal studies, articles not in English, and studies with fewer than 5 participants were excluded. Using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses reporting guideline, 2 independent reviewers assessed abstracts and full-text articles for entry criteria. A third reviewer resolved conflicts and errors in data extraction. Data were pooled using a random-effects model and data analysis occurred from August to November 2022. Peak activation x-axis (left-right), y-axis (posterior-anterior), and z-axis (inferior-superior) coordinates were extracted from all studies and submitted to MKDA meta-analyses. A total of 83 fMRI studies were included in the meta-analysis, yielding a combined sample of 5242 participants and 801 coordinates. Adversity exposure was associated with higher amygdala reactivity (familywise error rate corrected at P < .001; x-axis = 22; y-axis = -4; z-axis = -17) and lower prefrontal cortical reactivity (familywise error rate corrected at P < .001; x-axis = 10; y-axis = 60; z-axis = 10) across a range of task domains. These altered responses were only observed in studies that used adult participants and were clearest among those who had been exposed to severe threat and trauma. In this meta-analysis of fMRI studies of adversity exposure and brain function, prior adversity exposure was associated with altered adult brain reactivity to diverse challenges. These results might better identify how adversity diminishes the ability to cope with later stressors and produces enduring susceptibility to mental health problems.

Continuing professional development training needs for primary care doctors in central Uganda.

Continuing professional development (CPD) activities relevant to medical doctors and their patients should be informed by current assessed training needs. The CPD provision is expected to improve the quality of professional practice and ethics. However, the Uganda Medical and Dental Practitioners' Council still receives about 40 reports of malpractice every month. The study aimed to describe the CPD training needs of doctors working in public primary care facilities in central Uganda. The district health system of central Uganda comprised 10 General Hospitals (GH) and 37 Health Center IVs (HC IVs) with a staffing norm of six and two doctors, respectively. This was a cross-sectional survey of 100 doctors working in public primary care facilities using the World Health Organization (WHO) Hennessy-Hicks questionnaire. Descriptive statistics of the importance, current performance, and training need of each skilled activity were calculated. Content analysis was applied to data from the open-ended questions. The response rate was 91%, majority were males, 80 (87.9%) from 7 GHs and 24 HC IVs with an average age of 37.9 years. The domain with the highest CPD training need for the doctors was research and audit, with a mean score (standard deviation [s.d.]) of 1.94 (±1.69), followed by administration 1.58 (±1.61) and clinical tasks 1.28 (±1.29). The clinical tasks domain had the most suggested CPD topics. Research and audit and clinical tasks were identified as important domains for CPD training for doctors in this setting.Contribution: The results give insight into CPD training needs of primary care doctors and guide various CPD providers.