RESUMO
Competition affects mixed-mating strategies by limiting available abiotic or biotic resources such as nutrients, water, space, or pollinators. Cleistogamous species produce closed (cleistogamous, CL), obligately selfed, simultaneously with open (chasmogamous, CH), potentially outcrossed flowers. The effects of intraspecific competition on fitness and cleistogamy variation can range from limiting the production of costly CH flowers because of resource limitation, to favouring CH production because of fitness advantages of outcrossed, CH offspring. Moreover, the effects of competition can be altered when it co-occurs with other environmental variations. We grew plants from seven populations of the ruderal Lamium amplexicaule, originating from different climates and habitats, in a common garden experiment combining drought, interspecific competition, and seasonal variation. All these parameters have been shown to influence the degree of cleistogamy in the species on their own. In spring, competition and drought negatively impacted fitness, but the CL proportion only increased when plants were exposed to both treatments combined. We did not observe the same results in autumn, which can be due to non-adaptive phenotypic variation, or to differences in soil compactness between seasons. The observed responses are largely due to phenotypic plasticity, but we also observed phenotypic differentiation between populations for morphological, phenological, and cleistogamy traits, pointing to the existence of different ecotypes. Our data do not support the hypothesis that CL proportion should decrease when resources are scarce, as plants with reduced growth had relatively low CL proportions. We propose that variation in cleistogamy could be an adaptation to pollinator abundance, or to environment-dependent fitness differences between offspring of selfed and outcrossed seeds, two hypotheses worth further investigation. This opens exciting new possibilities for the study of the maintenance of mixed-mating systems using cleistogamous species as models that combine the effects of inbreeding and reproductive costs.
RESUMO
Computer Tomography (CT) is an imaging procedure that combines many X-ray measurements taken from different angles. The segmentation of areas in the CT images provides a valuable aid to physicians and radiologists in order to better provide a patient diagnose. The CT scans of a body torso usually include different neighboring internal body organs. Deep learning has become the state-of-the-art in medical image segmentation. For such techniques, in order to perform a successful segmentation, it is of great importance that the network learns to focus on the organ of interest and surrounding structures and also that the network can detect target regions of different sizes. In this paper, we propose the extension of a popular deep learning methodology, Convolutional Neural Networks (CNN), by including deep supervision and attention gates. Our experimental evaluation shows that the inclusion of attention and deep supervision results in consistent improvement of the tumor prediction accuracy across the different datasets and training sizes while adding minimal computational overhead.