Automated bacterial colony counting from images is an important technique to obtain data required for the development of vaccines and antibiotics. However, bacterial colonies present unique machine vision challenges that affect counting, including (1) small physical size, (2) object clustering, (3) high data annotation cost, and (4) limited cross-species generalisation. While FamNet is an established object counting technique effective for clustered objects and costly data annotation, its effectiveness for small colony sizes and cross-species generalisation remains unknown. To address the first three challenges, we propose ACFamNet, an extension of FamNet that handles small and clustered objects using a novel region of interest pooling with alignment and optimised feature engineering. To address all four challenges above, we introduce ACFamNet Pro, which augments ACFamNet with multi-head attention and residual connections, enabling dynamic weighting of objects and improved gradient flow. Experiments show that ACFamNet Pro achieves a mean normalised absolute error (MNAE) of 9.64

Automatic bacterial colony counting is a highly sought-after technology in modern biological laboratories because it eliminates manual counting effort. Previous work has observed that MicrobiaNet, currently the best-performing cardinality classification model for colony counting, has difficulty distinguishing colonies of three or more individuals. However, it is unclear if this is due to properties of the data together with inherent characteristics of the MicrobiaNet model. By analysing MicrobiaNet with explainable artificial intelligence (XAI), we demonstrate that XAI can provide insights into how data properties constrain cardinality classification performance in colony counting. Our results show that high visual similarity across classes is the key issue hindering further performance improvement, revising prior assertions about MicrobiaNet. These findings suggest future work should focus on models that explicitly incorporate visual similarity or explore density estimation approaches, with broader implications for neural network classifiers trained on imbalanced datasets.

To mitigate the economic losses and safety risks caused by reduced alertness of individuals in the context of an aging workforce, mental fatigue among the elderly is an issue worthy of in-depth exploration. Despite convergent studies on cognitive aging, the differential alterations in brain network topology between the elderly and young individuals during vigilance decline remain unclear. Here, a prolonged 30-min psychomotor vigilance task (PVT) was employed to induce mental fatigue, where both behavioral performance and electroencephalography (EEG) data were collected from healthy elderly (n =30) and young participants (n =40). Subsequently, EEG functional connectivity was constructed and the differences in network topological properties between the two groups were quantitatively evaluated based on global and nodal metrics. Both groups an exhibited age-independent significant decline in behavioral performance with time on task. Moreover, age-related dysconnectivity pattern was revealed over a wide frequency range ( $1-45$ Hz) in the elderly group, which further developed toward less optimal network architecture. Specifically, significant deficits in nodal efficiency were revealed in most of the brain regions, and the frontal area exhibited significant age-by-time interaction effect, which was attributed to a significant decline in the elderly group. Statistically significant correlation between behavioral and network metrics was also found. Overall, our results provide some of the first quantitative insights for revealing the neural mechanisms of age differences during mental fatigue, which may contribute to the rational arrangement of personnel in real-world scenarios with high alertness demands.

Transformer-based architectures have made significant progress in speech emotion recognition (SER). However, most published SER research trained and tested models on data from the same corpus, resulting in poor generalisation ability to unseen data collected from different corpora. To address this, we applied the HuBERT model to a combined training set consisting of five publicly available datasets (IEMOCAP, RAVDESS, TESS, CREMA-D, and 80% CMU-MOSEI) and conducted cross-corpus testing on the Strong Emotion (StrEmo) Dataset (a natural dataset collected by the authors) and two publicly available datasets (SAVEE and 20% CMU-MOSEI). Our best result achieved an F1 score of 0.78 over the three test sets, with an F1 score of 0.86 for StrEmo specifically. Additionally, we are pleased to release the spreadsheet of key information on the StrEmo dataset as supplementary material to the conference.

Sentiment Analysis could be performed on textual data and indicates the general 'tone' or emotional state of the writing. It is important in business, for instance in marketing, to determine customer opinions and trends, and in analysing social media to help weed out inappropriate or discriminatory language. Recently improved performance has been obtained by first converting the text to a grayscale image and then using a BLSTM and deep CNN, specifically ResNet, to classify the data. This paper investigates the addition of more context to the original text using a pre-trained BERT model to produce contextualised textual images. This produces a marked improvement over the previous results. The proposed BERT-BLSTM-ResNet model outperforms the BERT model on smaller datasets and above a threshold data size, the BERT performance is comparable.

The human brain connectome has long been recognized as a crucial component for various cognitive functions. While connectome-based predictive modeling (CPM) has been extensively explored for predicting behavior outcomes at the individual-level, its application to electroencephalogram (EEG) remains limited due to the inherent diversity and complexity of EEG frequency information. In the present work, we aim to address this issue by developing a filter bank CPM (FBCPM) framework that leverages narrowband EEG functional connectivity (FC) for individual prediction. Four independent datasets comprising 280 healthy subjects with 392 EEG recordings during the psychomotor vigilance test (PVT), were adopted here. Using the discovery dataset (i.e., Dataset 1) with 137 recordings, the feasibility of FBCPM was evaluated via predicting mean reaction time (RT) measures within a 15-min PVT task. The results showed that FBCPM framework achieved notable prediction accuracy and outperformed four benchmark approaches. Subsequent comprehensive internal and external validation analyses further affirmed its robustness across various hyper-parameters and generalizability to another three independent datasets (i.e., Dataset 2 to Dataset 4) with divergent recording or preprocessing settings. Moreover, the FBCPM framework exhibited satisfactory performance when generalized to time-on-task (TOT) effect measures (i.e., $\mathit {\Delta RT}$ and $\mathit {TOT_{slope}}$). Further investigation of contributing features to mean RT prediction indicated the remarkable predictive ability of negative features, manifesting as a pattern of low-frequency (below 8 Hz) predominance and complex topological distributions. Overall, these findings indicated that FBCPM provided a significant methodological advance in EEG-based individual prediction approaches, moving a step forward towards practical application in cognitive neuroscience.

The use of Machine Learning to monitor old people is crucial in providing immediate assistance and potentially life-saving interventions. With the rapid innovation in the field of Artificial Intelligence and Computer Vision, fall detection has seen significant improvements in accuracy and efficiency. Traditionally, 2D Convolutional Neural Networks (CNN) have been the main focus in fall detection research. However, these approaches have several drawbacks, including that 2D CNNs are primarily designed for spatial feature extraction and may not fully capture the temporal dynamics across multiple frames. This is because, for 2D CNN, the video frames are averaged out on time dimension within a time window. This project aims to explore and validate the use of 3D Convolutional Neural Networks (CNN) for fall detection, specifically in care home settings. The proposed 3D CNN keeps all frames in the time dimension (without averaging out video frames) and therefore can capture spatiotemporal dynamics of fall events more effectively, potentially enhancing detection accuracy. Experiment results indicate that the proposed 3D CNN achieved a G-Means, the geometric mean of recall and specificity, of 96.92

NLRP3 inflammasome activation is a pivotal area of research in innate immunity, yet the precise priming and activation signal remain unclear. In this study, we demonstrate that glycolysis inhibitor 2-Deoxy-D-glucose (2DG) triggers NLRP3-driven pyroptosis in human leukemia monocyte THP-1 cells by interfering glycosylation rather than glycolysis, which occurs independent of potassium efflux but requires the involvement of glycolysis rate-limiting enzyme PFKP. Using a CRISPR-Cas9 mediated large-scale screen, with 2DG as a new tool for probing NLRP3 activation, we identified that TLR2, rather than TLR4, initiates a rapid and robust priming signal for NLRP3 inflammasome activation. Importantly, both TLR2 and TLR4 depend entirely on MyD88, but not TRIF, for signal transduction. Furthermore, we discovered that TAK1, IKKβ and NEMO, but not IKKα, are essential for the priming signal. Additionally, we observed that deficiency in the linear ubiquitin assembly complex (LUBAC) subunits HOIP and HOIL-1, but not SHARPIN, is sufficient to inhibit 2DG-induced pyroptotic cell death. Collectively, our study reveals some common mechanism in the NLRP3 priming signals, as well as specific mechanisms upstream of NLRP3 triggered by 2DG, and underscores the potential of 2DG as a trigger to facilitate further detailed analysis of the underlying mechanisms of NLRP3 inflammasome activation.One Sentence Summary: Priming signal by IKKβ is essential for NLRP3 activation.

With the progression of artificial intelligence, there has been substantial advancement in autonomous driving technology. However, even the most advanced systems may confront failures in certain corner cases, necessitating enhanced analytical approaches. Traditional approaches focused on the numerical analysis of isolated sensor data, are often insufficient for deriving meaningful insights in such situations. To address this inadequacy, we propose a visual analytics approach, crafted to aid domain experts in performing analyses and extracting system improvements from cases with unexpected behaviors. This approach intricately integrates extensive driving scenarios and low-level module behaviors into the autonomous driving decision-making process, utilizing rich visualizations and an interface for interactive exploration and systematic synthesis of findings. Uniquely, our system opens the "black box" of modules in the decision-making pipeline during corner cases, taking into account both the overall decision-making pipeline and the fine-grained behaviors of the modules in the pipeline, setting our approach apart from previous works. To validate our system’s effectiveness, we perform two case studies, inviting domain experts for evaluation, and the results confirm our system’s efficacy in allowing experts to obtain crucial insights into autonomous driving systems.

Background:The unfolded protein response (UPR) is a critical biological process related to a variety of physiological functions and cardiac disease. However, the role of UPR-related genes in acute myocardial infarction (AMI) has not been well characterized. Therefore, this study aims to elucidate the mechanism and role of the UPR in the context of AMI. Methods:Gene expression profiles related to AMI and UPR pathway were downloaded from the Gene Expression Omnibus database and PathCards database, respectively. Differentially expressed genes (DEGs) were identified and then functionally annotated. The random forest (RF) and least absolute shrinkage and selection operator (LASSO) regression analysis were conducted to identify potential diagnostic UPR-AMI biomarkers. Furthermore, the results were validated by using external data sets, and discriminability was measured by the area under the curve (AUC). A nomogram based on the feature genes was developed to predict the AMI-risk rate. Then we utilized two algorithms, CIBERSORT and MCPcounter, to investigate the relationship between the key genes and immune microenvironment. Additionally, we performed uniform clustering of AMI samples based on the expression of UPR pathway-related genes. The weighted gene co-expression network analysis was conducted to identify the key modules in various clusters, enrichment analysis was performed for the genes existing in different modules. Results:A total of 14 DEGs related to the UPR pathway were identified. Among the 14 DEGs, CEBPB, ATF3, EIF2S3, and TSPYL2 were subsequently identified as biomarkers by the LASSO and RF algorithms. A diagnostic model was constructed with these four genes, and the AUC was 0.939. The calibration curves, receiver operating characteristic (ROC) curves, and the decision curve analysis of the nomogram exhibited good performance. Furthermore, immune cell infiltration analysis revealed that four feature genes were linked with the infiltration of immune cells such as neutrophils. The cluster analysis of the AMI samples identified two distinct clusters, each with differential expression of genes related to the UPR pathway, immune cell infiltration, and inflammatory cytokine secretion. Weighted gene coexpression network analysis and enrichment analysis showed that both clusters were associated with the UPR. Conclusions:Our study highlights the importance of the UPR pathway in the pathogenesis of myocardial infarction, and identifies four genes CEBPB, ATF3, EIF2S3, and TSPYL2 as diagnostic biomarkers for AMI, providing new ideas for the clinical diagnosis and treatment of AMI.

Idiopathic REM sleep behavior disorder (iRBD) is recognized as a prodromal stage of neuro-degenerative disease. While brain network analysis is a well-documented approach for characterizing disease-related dysfunctions, the specific patterns in iRBD, particularly those related to hemispheric aberrations remain largely unexplored. To address this gap, this study investigated the topological abnormalities of multi-band EEG networks in patients with iRBD. Specifically, eyes-open resting-state EEG signals were collected from 30 iRBD patients and 30 matched health control (HC) participants. Graph theoretical analysis was then employed to explore network properties at the whole-brain and the hemispheric level. At the whole-brain level, we found aberrant increased local and global efficiency along with a distinct pattern of increased frontal and decreased parietal nodal efficiency in alpha band of iRBD patients. At the hemispheric level, iRBD networks displayed more efficient topological properties in the left hemisphere. Additionally, significant hemispheric asymmetry was observed in alpha-band iRBD network compared to that of HC. In sum, these findings provide novel insights into the disrupted network reorganization in iRBD and suggest aberrant hemispheric asymmetry as a potential neural biomarker for early diagnosis and monitoring of the disease.