Non-electroencephalogram seizure detection holds promise for the early identification of generalised onset seizures. However, existing methods often suffer from high false alarm rates and difficulty distinguishing normal movements from seizure manifestations. To address this, we obtained exclusive access to the Open Seizure Database and selected a representative dataset of 94 events (42 generalised tonic-clonic seizures, 19 focal seizures, and 33 labelled as Other), totaling approximately 5 hours and 29 minutes. Each event contains acceleration and heart rate data, which were expertly annotated by a clinician in 5 second timesteps, with each timestep assigned a class label of Normal, Pre-Ictal, or Ictal. We introduce AMBER (Attention-guided Multi-Branching pipeline with Enhanced Residual Fusion), a multimodal seizure detection model designed for Ictal-Phase Detection. AMBER constructs multiple branches to form independent feature extraction pipelines for each sensing modality. The outputs of each branch are passed to a Residual Fusion layer, where the extracted features are combined into a fused representation and propagated through two densely connected blocks. The results of these experiments highlight the effectiveness of Ictal Phase Detection, with the model recording an accuracy and f1-score of 0.9027 and 0.9035, respectively, on unseen test data. Further experiments recorded True Positive Rate of 0.8342, 0.9485, and 0.9118 for the Normal, Pre-Ictal, and Ictal phases, respectively, with an average False Positive Rate of 0.0502. This study presents a novel Ictal Phase Detection technique that enhances seizure phase classification while showing reduced false alarms, laying the groundwork for further advancements in non-electroencephalogram-based seizure detection research.

Elaboration of Bayesian phylogenetic inference methods has continued at pace in recent years with major new advances in nearly all aspects of the joint modelling of evolutionary data. It is increasingly appreciated that some evolutionary questions can only be adequately answered by combining evidence from multiple independent sources of data, including genome sequences, sampling dates, phenotypic data, radiocarbon dates, fossil occurrences, and biogeographic range information among others. Including all relevant data into a single joint model is very challenging both conceptually and computationally. Advanced computational software packages that allow robust development of compatible (sub-)models which can be composed into a full model hierarchy have played a key role in these developments. Developing such software frameworks is increasingly a major scientific activity in its own right, and comes with specific challenges, from practical software design, development and engineering challenges to statistical and conceptual modelling challenges. BEAST 2 is one such computational software platform, and was first announced over 4 years ago. Here we describe a series of major new developments in the BEAST 2 core platform and model hierarchy that have occurred since the first release of the software, culminating in the recent 2.5 release.

MOTIVATION: The multispecies coalescent model provides a formal framework for the assignment of individual organisms to species, where the species are modeled as the branches of the sp tree. None of the available approaches so far have simultaneously co-estimated all the relevant parameters in the model, without restricting the parameter space by requiring a guide tree and/or prior assignment of individuals to clusters or species. RESULTS: We present DISSECT, which explores the full space of possible clusterings of individuals and species tree topologies in a Bayesian framework. It uses an approximation to avoid the need for reversible-jump Markov Chain Monte Carlo, in the form of a prior that is a modification of the birth-death prior for the species tree. It incorporates a spike near zero in the density for node heights. The model has two extra parameters: one controls the degree of approximation and the second controls the prior distribution on the numbers of species. It is implemented as part of BEAST and requires only a few changes from a standard *BEAST analysis. The method is evaluated on simulated data and demonstrated on an empirical dataset. The method is shown to be insensitive to the degree of approximation, but quite sensitive to the second parameter, suggesting that large numbers of sequences are needed to draw firm conclusions. AVAILABILITY AND IMPLEMENTATION: http://tree.bio.ed.ac.uk/software/beast/, http://www.indriid.com/dissectinbeast.html. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Unsupervised over-segmentation of an image into super-pixels is a common preprocessing step for image parsing algorithms. Superpixels are used as both regions of support for feature vectors and as a starting point for the final segmentation. In this paper we investigate incorporating a priori information into superpixel segmentations. We learn a probabilistic model that describes the spatial density of the object boundaries in the image. We then describe an over-segmentation algorithm that partitions this density roughly equally between superpixels whilst still attempting to capture local object boundaries. We demonstrate this approach using road scenes where objects in the center of the image tend to be more distant and smaller than those at the edge. We show that our algorithm successfully learns this foveated spatial distribution and can exploit this knowledge to improve the segmentation. Lastly, we introduce a new metric for evaluating vision labeling problems. We measure performance on a challenging real-world dataset and illustrate the limitations of conventional evaluation metrics.

Unsupervised over-segmentation of an image into superpixels is a common preprocessing step for image parsing algorithms. Ideally, every pixel within each superpixel region will belong to the same real-world object. Existing algorithms generate superpixels that forfeit many useful properties of the regular topology of the original pixels: for example, the nth superpixel has no consistent position or relationship with its neighbors. We propose a novel algorithm that produces superpixels that are forced to conform to a grid (a regular superpixel lattice). Despite this added topological constraint, our algorithm is comparable in terms of speed and accuracy to alternative segmentation approaches. To demonstrate this, we use evaluation metrics based on (i) image reconstruction (ii) comparison to human-segmented images and (iii) stability of segmentation over subsequent frames of video sequences.