Achieving accurate material segmentation for 3-channel RGB images is challenging due to the considerable variation in material appearance. Hyperspectral images, which are sets of spectral measurements sampled at multiple wavelengths, theoretically offer distinct information for material identification, as variations in the intensity of electromagnetic radiation reflected by a surface depend on the material composition of a scene. However, existing hyperspectral datasets are impoverished in terms of the number of images and material categories for the dense material segmentation task, and collecting and annotating hyperspectral images with a spectral camera is prohibitively expensive. To address this, we propose a novel model, the MatSpectNet to segment materials with recovered hyperspectral images from RGB images. The network leverages the principles of colour perception in modern cameras to constrain the reconstructed hyperspectral images and employs the domain adaptation method to generalise the hyperspectral reconstruction capability from a spectral recovery dataset to material segmentation datasets. The reconstructed hyperspectral images are further filtered using learnt response curves and enhanced with human perception. The performance of MatSpectNet is evaluated on the LMD dataset as well as the OpenSurfaces dataset. Our experiments demonstrate that MatSpectNet attains a 1.60% increase in average pixel accuracy and a 3.42% improvement in mean class accuracy compared with the most recent publication. Additional experiments and the project code are published on https://github.com/heng-yuwen/MatSpectNet.

Weight-sharing quantization has emerged as a technique to reduce energy expenditure during inference in large neural networks by constraining their weights to a limited set of values. However, existing methods often assume weights are treated solely based on value, neglecting the unique role of weight position. This paper proposes a probabilistic framework based on Bayesian neural networks (BNNs) and a variational relaxation to identify which weights can be moved to which cluster center and to what degree based on their individual position-specific learned uncertainty distributions. We introduce a new initialization setting and a regularization term, enabling the training of BNNs with complex dataset-model combinations. Leveraging the flexibility of weight values from probability distributions, we enhance noise resilience and compressibility. Our iterative clustering procedure demonstrates superior compressibility and higher accuracy compared to state-of-the-art methods on both ResNet models and the more complex transformer-based architectures. In particular, our method outperforms the state-of-the-art quantization method top-1 accuracy by 1.6\% on ImageNet using DeiT-Tiny, with its 5 million+ weights now represented by only 296 unique values. Code available at https://github.com/subiawaud/PWFN.

The UNet model consists of fully convolutional network (FCN) layers arranged as contracting encoder and upsampling decoder maps. Nested arrangements of these encoder and decoder maps give rise to extensions of the UNet model, such as UNete and UNet++. Other refinements include constraining the outputs of the convolutional layers to discriminate between segment labels when trained end to end, a property called deep supervision. This reduces feature diversity in these nested UNet models despite their large parameter space. Furthermore, for texture segmentation , pixel correlations at multiple scales contribute to the classification task ; hence, explicit deep supervision of shallower layers is likely to enhance performance. In this paper, we propose ADS_UNet, a stage-wise additive training algorithm that incorporates resource-efficient deep supervision in shallower layers and takes performance-weighted combinations of the sub-UNets to create the segmentation model. We provide empirical evidence on three histopathology datasets to support the claim that the proposed ADS_UNet reduces correlations between constituent features and improves performance while being more resource efficient. We demonstrate that ADS_UNet outperforms state-of-the-art Transformer-based models by 1.08 and 0.6 points on CRAG and BCSS datasets, and yet requires only 37% of GPU consumption and 34% of training time as that required by Transformers. The source code is available at: .

Recent advances in high throughput technologies have made large amounts of biomedical omics data accessible to the scientific community. Single omic data clustering has proved its impact in the biomedical and biological research fields. Multi-omic data clustering and multi-omic data integration techniques have shown improved clustering performance and biological insight. Cancer subtype clustering is an important task in the medical field to be able to identify a suitable treatment procedure and prognosis for cancer patients. State of the art multi-view clustering methods are based on non-convex objectives which only guarantee non-global solutions that are high in computational complexity. Only a few convex multi-view methods are present. However, their models do not take into account the intrinsic manifold structure of the data. In this paper, we introduce a convex graph regularized multi-view clustering method that is robust to outliers. We compare our algorithm to state of the art convex and non-convex multi-view and single view clustering methods, and show its superiority in clustering cancer subtypes on publicly available cancer genomic datasets from the TCGA repository. We also show our method's better ability to potentially discover cancer subtypes compared to other state of the art multi-view methods.

The behaviour of a high dimensional stochastic system described by a chemical master equation (CME) depends on many parameters, rendering explicit simulation an inefficient method for exploring the properties of such models. Capturing their behaviour by low-dimensional models makes analysis of system behaviour tractable. In this paper, we present low dimensional models for the noise-induced excitable dynamics in Bacillus subtilis, whereby a key protein ComK, which drives a complex chain of reactions leading to bacterial competence, gets expressed rapidly in large quantities (competent state) before subsiding to low levels of expression (vegetative state). These rapid reactions suggest the application of an adiabatic approximation of the dynamics of the regulatory model that, however, lead to competence durations that are incorrect by a factor of 2. We apply a modified version of an iterative functional procedure that faithfully approximates the time-course of the trajectories in terms of a two-dimensional model involving proteins ComK and ComS. Furthermore, in order to describe the bimodal bivariate marginal probability distribution obtained from the Gillespie simulations of the CME, we introduce a tunable multiplicative noise term in a two-dimensional Langevin model whose stationary state is described by the time-independent solution of the corresponding Fokker-Planck equation.

The prospect of emulating the impressive computational capabilities of biological systems has led to considerable interest in the design of analog circuits that are potentially implementable in very large scale integration CMOS technology and are guided by biologically motivated models. For example, simple image processing tasks, such as the detection of edges in binary and grayscale images, have been performed by networks of FitzHugh-Nagumo-type neurons using the reaction-diffusion models. However, in these studies, the one-to-one mapping of image pixels to component neurons makes the size of the network a critical factor in any such implementation. In this paper, we develop a simplified version of the employed reaction-diffusion model in three steps. In the first step, we perform a detailed study to locate this threshold using continuous Lyapunov exponents from dynamical system theory. Furthermore, we render the diffusion in the system to be anisotropic, with the degree of anisotropy being set by the gradients of grayscale values in each image. The final step involves a simplification of the model that is achieved by eliminating the terms that couple the membrane potentials of adjacent neurons. We apply our technique to detect edges in data sets of artificially generated and real images, and we demonstrate that the performance is as good if not better than that of the previous methods without increasing the size of the network.

The current trend in distributed systems is towards service-based integration. This article describes an ontology-driven framework implemented to provide knowledge management for data of different modalities, with multimedia processing, annotation, and reasoning provided by remote services. The framework was developed in, and is presented in the context of, the Medical Imaging and Advanced Knowledge Technologies (MIAKT) project that sought to support the Multidisciplinary Meetings (MDMs) that take place during breast cancer screening for diagnosing the patient. However, the architecture is entirely independent of the specific application domain and can be quickly prototyped into new domains. An Enterprise server provides resource access to a client-side presentation application which, in turn, provides knowledge visualization and markup of any supported media, as defined by a domain-dependent ontology-supported language.

A distributed decision support system involving multiple clinical centres is crucial to the diagnosis of rare diseases. Although sharing of valid diagnosed cases can facilitate later decision making, possibly from geographically different centres, the released information could reveal patient privacy if it is not properly protected. Clinical centres may have to impose their distinct regulations and rules that govern the use of their data externally. The collaboration of centres, therefore, must respect the collective policies and ideally, serve users the most appropriate and useful resources possible in the system according to the past experience. In this way, the system’s value is entrusted and even elevated through continuous collaboration. We present in this paper a link-anonymised data scheme and in addition to that, a security model that together enforce privacy data security and secure resource access for distributed clinical centres. Our illustration of the approach involves a prototype medical decision support system, HealthAgents, for brain tumour diagnosis.

This paper focuses on the problem of representing, in a meaningful way, the knowledge involved in the HealthAgents project. Our work is motivated by the complexity of representing electronic healthcare records in a consistent manner. We present HADOM (HealthAgents domain ontology) which conceptualises the required HealthAgents information and propose describing the sources knowledge by the means of conceptual graphs (CGs). This allows to build upon the existing ontology permitting for modularity and flexibility. The novelty of our approach lies in the ease with which CGs can be placed above other formalisms and their potential for optimised querying and retrieval.

Ontologies, as a knowledge representation paradigm, rely on semantics for performing logical inference. However, in the context of ontology query and reuse, the semantics are hard to be manipulated. When searching for ontologies and identifying potentially relevant candidates, a mechanism is required to exploit (i) the appropriate semantic restrictions as well as the (ii) meaning of concepts/properties. This paper proposes an approach that will address both requirements. The contribution of this paper is to conceptually detail how a set of ontology indices can be obtained. More specifically, to obtain a relevant set of keywords describing an ontology, we use logic-based constructors to generate an initial set of keywords. The meaning of these keywords is further explored by consulting a comprehensive text corpora serving as the background/domain knowledge. The novelty of our work lies in the combination of logic-based formalisms with a collaboratively contributed web repository. We illustrate our work by the means of a real world example.

The interpretation of medical evidence is normally presented in terms of a controlled, but diversely expressed specialist vocabulary and natural language phrases. Such informally expressed data require human intervention to ascertain its relevance in any specific case. In order to facilitate machine-based reasoning about the evidence gathered, additional interpretive semantics must be attached to the data; a shift from a merely data-intensive approach to a semantics-rich model of evidence. In this paper, we present a system to formally annotate medical images captured to aid the diagnosis and management of breast cancer, that enables a series of semantics-based operations to be performed. Our approach is grounded upon an imaging ontology specifying the domain knowledge and a description logic (DL) taxonomic inferential engine responsible for semantics-based reasoning and image retrieval.

We describe some high-level approaches to estimating confidence scores for the words output by a speech recognizer. By "high-level" we mean that the proposed measures do not rely on decoder specific "side information" and so should find more general applicability than measures that have been developed for specific recognizers. Our main approach is to attempt to decouple the language modeling and acoustic modeling in the recognizer in order to generate independent information from these two sources that can then be used for estimation of confidence. We isolate these two information sources by using a phone recognizer working in parallel with the word recognizer. A set of techniques for estimating confidence measures using the phone recognizer output in conjunction with the word recognizer output is described. The most effective of these techniques is based on the construction of "metamodels," which generate alternative word hypotheses for an utterance. An alternative approach requires no other recognizers or extra information for confidence estimation and is based on the notion that a word that is semantically "distant" from the other decoded words in the utterance is likely to be incorrect. We describe a method for constructing "semantic similarities" between words and hence estimating a confidence. Results using the UK version of the Wall Street Journal are given for each technique.

We describe an approach to confidence estimation that attempts to decouple the contributions of the acoustic and language model components to speech recognition output. The output of the acoustic models when decoding phonemes is itself modelled using HMMs to produce a set of models which we term meta-models. When benchmarked against a "standard" method for assigning confidence (the N-best score), the meta-models gave a relative improvement of 6.2%. Furthermore, it appears that the N-best and meta-models techniques are complementary, because they tend to fail on different words.

In previous work, we have argued that it is beneficial to find confidence measures (CM's) that are not dependent on use of "side information" from a specific recogniser. Here, we extend this philosophy to include the use of semantic information in estimating the confidence that a word is correct. We are motivated by the observation that sometimes the recogniser outputs a word which can easily be spotted (by humans) as incorrect, because it bears no relation to the semantics of the rest of the decoded sentence. Latent semantic analysis (LSA) was used as a method for estimating semantic "semantic similarity" between words in a text corpus. From these scores, an average semantic similarity of each decoded word to the other decoded words in an utterance could be estimated, and by thresholding this similarity measure, words were tagged as CORRECT or INCORRECT. We benchmarked the performance of this semantic CM against a tried-and-tested CM, the N-best CM. The precision of the semantic CM was inferior to that of N-best when the recall (the number of words considered) was high, but it out-performed N-best for low recall, and a combined classifier showed the benefits of using both techniques. An interesting and unexpected result was that the semantic CM was better at identifying correct words than incorrect words.