In this work, we explore various computer vision techniques, with a focus on texture recognition approaches, for the task of plant species detection. We particularly emphasize the study of a challenging dataset consisting of 50 Brazilian plant species' leaf midrib cross-sections using microscope images. The research focuses on a recent method named Random Encoding of Aggregated Deep Activation Maps (RADAM) that leverages deep features from pre-trained Convolutional Neural Networks (CNNs) for improved plant species identification. This method demonstrates significant advancement over traditional texture analysis and deep learning approaches, showcasing the potential of combining deep feature engineering with texture analysis for accurate plant species recognition.

Timely assessment and prediction of changes in microbial compositions leading to activated sludge settling problems, such as filamentous bulking (FB), can reduce water resource recovery facilities (WRRFs) upsets, operational challenges, and negative environmental impacts. This study presents a computer vision approach to assess activated sludge-settling characteristics based on Microscopy Images (MIs). We utilize MIs to train deep convolutional neural networks (CNN) using transfer learning to investigate the morphological properties of flocs and filaments. The methodology was tested on the offline MI dataset collected over two years at a full-scale industrial WRRF in Belgium. Various CNN architectures were tested, including Inception v3, ResNet18, ResNet152, ConvNeXt-nano, and ConvNeXt-S. The sludge volume index (SVI) was used as the final prediction variable, but the method can be easily adjusted to predict any other settling metric of choice. The best-performing CNN, ConvNeXt-nano, could predict SVI values with MAE (37.51 ± 4.02), MTD (11.65 ± 1.94), MAPE (0.18 ± 0.02), and R2 (0.75 ± 0.05). The model was tested in real-life FB events, where it identified early indicators of bulking by predictive surges in SVI values. We used an explainable AI technique, Eigen-CAM, to discover key morphological indicators of sludge bulking transitions. The findings highlight the SVI multimodality issue, where SVI readings as a unidimensional metric could not capture delicate shifts from good to poor sludge settling, while the model detected these subtle changes. The key morphological attributes of threshold conditions leading to FB were identified, which can provide actionable insight for preemptive WRRF management.

Chronic systemic immune activation significantly influences human immunodeficiency virus (HIV) disease progression. Despite evidence of a pro-inflammatory environment in the genital tract of HIV-infected women, comprehensive investigations into cervical tissue from this region remain limited. Similarly, the consequences of chronic HIV infection on the integrity of the female genital epithelium are poorly understood, despite its importance in HIV transmission and replication. Ectocervical biopsies were obtained from HIV-seropositive (n = 14) and HIV-seronegative (n = 47) female Kenyan sex workers. RNA sequencing and bioimage analysis of epithelial junction proteins (E-cadherin, desmoglein-1, claudin-1, and zonula occludens-1) were conducted, along with CD4 staining. RNA sequencing revealed upregulation of immunoregulatory genes in HIV-seropositive women, primarily associated with heightened T cell activity and interferon signaling, which further correlated with plasma viral load. Transcription factor analysis confirmed the upregulation of pro-inflammatory transcription factors, such as RELA, NFKB1, and IKZF3, which facilitates HIV persistence in T cells. Conversely, genes and pathways associated with epithelial barrier function and structure were downregulated in the context of HIV. Digital bioimage analysis corroborated these findings, revealing significant disruption of various epithelial junction proteins in ectocervical tissues of the HIV-seropositive women. Thus, chronic HIV infection associated with ectocervical inflammation, characterized by induced T cell responses and interferon signaling, coupled with epithelial disruption. These alterations may influence HIV transmission and heighten susceptibility to other sexually transmitted infections. These findings prompt exploration of therapeutic interventions to address HIV-related complications and mitigate the risk of sexually transmitted infection transmission.

We analyse and mutually compare time series of covid-19-related data and mobility data across Belgium's 43 arrondissements (NUTS 3). In this way, we reach three conclusions. First, we could detect a decrease in mobility during high-incidence stages of the pandemic. This is expressed as a sizeable change in the average amount of time spent outside one's home arrondissement, investigated over five distinct periods, and in more detail using an inter-arrondissement "connectivity index"(CI). Second, we analyse spatio-temporal covid-19-related hospitalisation time series, after smoothing them using a generalise additive mixed model (GAMM). We confirm that some arrondissements are ahead of others and morphologically dissimilar to others, in terms of epidemiological progression. The tools used to quantify this are time-lagged cross-correlation (TLCC) and dynamic time warping (DTW), respectively. Third, we demonstrate that an arrondissement's CI with one of the three identified first-outbreak arrondissements is correlated to a substantial local excess mortality some five to six weeks after the first outbreak. More generally, we couple results leading to the first and second conclusion, in order to demonstrate an overall correlation between CI values on the one hand, and TLCC and DTW values on the other. We conclude that there is a strong correlation between physical movement of people and viral spread in the early stage of the sars-cov-2 epidemic in Belgium, though its strength weakens as the virus spreads.

Silk fibers derived from the cocoon of silk moths and the wide range of silks produced by spiders exhibit an array of features, such as extraordinary tensile strength, elasticity, and adhesive properties. The functional features and mechanical properties can be derived from the structural composition and organization of the silk fibers. Artificial recombinant protein fibers based on engineered spider silk proteins have been successfully made previously and represent a promising way towards the large-scale production of fibers with predesigned features. However, for the production and use of protein fibers, there is a need for reliable objective quality control procedures that could be automated and that do not destroy the fibers in the process. Furthermore, there is still a lack of understanding the specifics of how the structural composition and organization relate to the ultimate function of silk-like fibers. In this study, we develop a new method for the categorization of protein fibers that enabled a highly accurate prediction of fiber tensile strength. Based on the use of a common light microscope equipped with polarizers together with image analysis for the precise determination of fiber morphology and optical properties, this represents an easy-to-use, objective non-destructive quality control process for protein fiber manufacturing and provides further insights into the link between the supramolecular organization and mechanical functionality of protein fibers.

The extinction of ecosystems and the mechanisms that support or limit species coexistence have long been studied by scientists. It has been shown that competition and cyclic dominance among species promote species coexistence, such as in the classic Rock-Paper-Scissors (RPS) game. However, individuals’ mobility and the underlying topology that defines the neighbourhood relations between individuals also play an important role in maintaining biodiversity. Typically, square grids are used for simulating such interactions. However, these constrain the individuals’ spatial degrees of freedom. In this work, we investigate the effect of the underlying topology on the RPS dynamics. For that purpose, we considered networks with varying node degree distributions and generated according to different theoretical models. We analyzed the time to the first extinction and the patchiness of the in silico ecosystem over time. In general, we observed a distinct large effect of the network topology on the RPS dynamics. Moreover, leaving regular networks aside, the probability of extinction is very high for some network models due to their inherent long-range connections. On the other hand, spatial arrangements characterized by nearest neighbors interactions have fewer long-range correlations, which is essential for biodiversity.

Evidence suggests that networks trained on large datasets generalize well not solely because of the numerous training examples, but also class diversity which encourages learning of enriched features. This raises the question of whether this remains true when data is scarce - is there an advantage to learning with additional labels in low-data regimes In this work, we consider a task that requires difficult-To-obtain expert annotations: Tumor segmentation in mammography images. We show that, in low-data settings, performance can be improved by complementing the expert annotations with seemingly uninformative labels from non-expert annotators, turning the task into a multi-class problem. We reveal that these gains increase when less expert data is available, and uncover several interesting properties through further studies. We demonstrate our findings on CSAW-S, a new dataset that we introduce here, and confirm them on two public datasets.

Evidence suggests that networks trained on large datasets generalize well not solely because of the numerous training examples, but also class diversity which encourages learning of enriched features. This raises the question of whether this remains true when data is scarce – is there an advantage to learning with additional labels in low-data regimes? In this work, we consider a task that requires difficult-to-obtain expert annotations: tumor segmentation in mammography images. We show that, in low-data settings, performance can be improved by complementing the expert annotations with seemingly uninformative labels from non-expert annotators, turning the task into a multi-class problem. We reveal that these gains increase when less expert data is available, and uncover several interesting properties through further studies. We demonstrate our findings on CSAW-S, a new dataset that we introduce here, and confirm them on two public datasets.

This paper outlines a hybrid modeling approach to facilitate weather-based operation and energy optimization for the largest Italian wastewater treatment plant (WWTP). Two clustering methods,K-means algorithm and Gaussian mixture model (GMM) based on the expectation-maximization (EM) algorithm, were applied to an extensive dataset of historical and meteorological records. This study addresses the problem of determining the intrinsic structure of clustered data when no information other than the observed values is available. Two quantitative indexes, namely the Bayesian information criterion (BIC) and the Silhouette coefficient using Euclidean distance, as well as two general criteria, were implemented to assess the clustering quality. Furthermore, seven weather-based influent scenarios were introduced to the process simulation model, and sets of aeration strategies are proposed. The results indicate that incorporating weather-based aeration strategies in the operation of the WWTP improves plant energy efficiency.