Biology subjects

Cusack, J.

Publications and source records attributed to Cusack, J..

3 recordsLinked to original sources

bioRxiv2024

Variable responses of individual species to tropical forest degradation

The functional stability of ecosystems depends greatly on interspecific differences in responses to environmental perturbation. However, responses to perturbation are not necessarily invariant among populations of the same species, so intraspecific variation in responses might also contribute. Such inter-population response diversity has recently been shown to occur spatially across species ranges, but we lack estimates of the extent to which individual populations across an entire community might have perturbation responses that vary through time. We assess this using 524 taxa that have been repeatedly surveyed for the effects of tropical forest logging at a focal landscape in Sabah, Malaysia. Just 39 % of taxa - all with non-significant responses to forest degradation - had invariant responses. All other taxa (61 %) showed significantly different responses to the same forest degradation gradient across surveys, with 6 % of taxa responding to forest degradation in opposite directions across multiple surveys. Individual surveys had low power (< 80 %) to determine the correct direction of response to forest degradation for one-fifth of all taxa. Recurrent rounds of logging disturbance increased the prevalence of intra-population response diversity, while uncontrollable environmental variation and/or turnover of intraspecific phenotypes generated variable responses in at least 44 % of taxa. Our results show that the responses of individual species to local environmental perturbations are remarkably flexible, likely providing an unrealised boost to the stability of disturbed habitats such as logged tropical forests.

ecology↗

bioRxiv2022

Weakly supervised Unet: an image classifier which learns to explain itself

BackgroundExplainability is a major limitation of current convolutional neural network (CNN) image classifiers. A CNN is required which supports its image-level prediction with a voxel-level segmentation. MethodsA weakly-supervised Unet architecture (WSUnet) is proposed to model voxel classes, by training with image-level supervision. WSUnet computes the image-level class prediction from the maximal voxel class prediction. Thus, voxel-level predictions provide a causally verifiable saliency map for the image-level decision. WSUnet is applied to explainable lung cancer detection in CT images. For comparison, current model explanation approaches are also applied to a standard CNN. Methods are compared using voxel-level discrimination metrics and a clinician preference survey. ResultsIn test data from two external institutions, WSUnet localised the tumour precisely at voxel-level (Precision: 0.93 [0.93-0.94]), achieving superior voxel-level discrimination to the best comparator (AUPR: 0.55 [0.54-0.55] vs. 0.36 [0.35-0.36]). Clinicians preferred WSUnet predictions in most test instances (Clinician Preference Rate: 0.72 [0.68-0.77]). ConclusionsWSUnet is a simple extension of the Unet, which facilitates voxel-level modelling from image-level labels. As WSUnet supports its image-level prediction with a causative voxel-level segmentation, it functions as a self-explaining image classifier. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=194 SRC="FIGDIR/small/507144v1_ufig1.gif" ALT="Figure 1"> View larger version (35K): org.highwire.dtl.DTLVardef@52e3ccorg.highwire.dtl.DTLVardef@1e981f0org.highwire.dtl.DTLVardef@151f31eorg.highwire.dtl.DTLVardef@13046a2_HPS_FORMAT_FIGEXP M_FIG Graphical Abstract The weakly-supervised Unet converts voxel-level predictions to image-level predictions using a global max-pooling layer. Thus, loss is computed at image-level. Following training with image-level labels, voxel-level predictions are extracted from the voxel-level output layer. C_FIG FundingAuthors acknowledge funding support from the UK Research & Innovation London Medical Imaging and Artificial Intelligence Centre; Wellcome/Engineering and Physical Sciences Research Council Centre for Medical Engineering at Kings College London [WT 203148/Z/16/Z]; National Institute for Health Research Biomedical Research Centre at Guys & St Thomas Hospitals and Kings College London; National Institute for Health Research Biomedical Research Centre at Guys & St Thomas Hospitals and Kings College London; Cancer Research UK National Cancer Imaging Translational Accelerator [C1519/A28682]. For the purpose of open access, authors have applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission. HIGHLIGHTSO_LIWSUnet is a weakly supervised Unet architecture which can learn semantic segmentation from data labelled only at image-level. C_LIO_LIWSUnet is a convolutional neural network image classifier which provides a causally verifiable voxel-level explanation to support its image-level prediction. C_LIO_LIIn application to explainable lung cancer detection, WSUnets voxel-level output localises tumours precisely, outperforming current model explanation methods. C_LIO_LIWSUnet is a simple extension of the standard Unet architecture, requiring only the addition of a global max-pooling layer to the output. C_LI

bioinformatics↗

bioRxiv2017

Limited evidence for proactive and reactive spatial responses of prey to an active predator

O_LIThe extent to which prey space use actively minimises predation risk continues to ignite controversy. Methodological reasons that have hindered consensus include inconsistent measurements of predation risk, biased spatiotemporal scales at which responses are measured, and lack of robust null expectations.\nC_LIO_LIWe addressed all three challenges in a comprehensive analysis of the spatiotemporal responses of adult female elk (Cervus elaphus) to the risk of predation by grey wolves (Canis lupus) during winter in northern Yellowstone, USA.\nC_LIO_LIWe quantified spatial overlap between the winter home ranges of GPS-collared elk and three measures of predation risk: the intensity of wolf space use, the distribution of wolf-killed elk and vegetation openness. We also assessed whether elk varied their use of areas characterised by more or less predation risk across hours of the day, and estimated encounter rates between simultaneous elk and wolf pack trajectories. We determined whether observed values were significantly lower than expected if elk movements were random with reference to predation risk using a null model approach.\nC_LIO_LIAlthough a small proportion of elk did show a tendency to minimise use of open vegetation at specific times of the day, overall we highlight a notable absence of spatiotemporal response by female elk to the risk of predation posed by wolves in northern Yellowstone.\nC_LIO_LIOur results suggest that predator-prey interactions may not always result in strong spatiotemporal patterns of avoidance.\nC_LI

ecology↗