179566 research outputs found
Sort by
Exploring Robustness of Image Recognition Models on Hardware Accelerators
Get PDFAs the usage of Artificial Intelligence (AI) on resource-intensive and safety-critical tasks increases, a variety of Machine Learning (ML) compilers have been developed, enabling compatibility of Deep Neural Networks (DNNs) with a variety of hardware acceleration devices. However, given that DNNs are widely utilized for challenging and demanding tasks, the behavior of these compilers must be verified. To this direction, we propose MutateNN, a tool that utilizes elements of both differential and mutation testing in order to examine the robustness of image recognition models when deployed on hardware accelerators with different capabilities, in the presence of faults in their target device code - introduced either by developers, or problems in their compilation process. We focus on the image recognition domain by applying mutation testing to 7 well-established DNN models, introducing 21 mutations of 6 different categories. We deployed our mutants on 4 different hardware acceleration devices of varying capabilities and observed that DNN models presented discrepancies of up to 90.3% in mutants related to conditional operators across devices. We also observed that mutations related to layer modification, arithmetic types and input affected severely the overall model performance (up to 99.8%) or led to model crashes, in a consistent manner across devices
SCONe:a community-acquired retinal image repository enabling ocular, cardiovascular and neurodegenerative disease prediction
Get PDFOBJECTIVES: To safeguard Scotland's community-acquired retinal images (colour fundus photographs) in a secure, centrally held repository and support a variety of research including ocular, neurodegenerative and systemic disease prediction.DESIGN: Retinal images captured in optometry practices linked to national, routinely collected, longitudinal healthcare data.SETTING: Community optometry and the Public Health Scotland National Safe Haven.PARTICIPANTS: Adults (mostly aged 60+) who have attended their optometrist since 2006 for an eye examination during which a retinal image was captured.MAIN OUTCOME MEASURES: Successful retrieval of linkable colour fundus photographs from systems in use in practice and delivery to the Safe Haven for linkage and secure storage.RESULTS: Scottish Collaborative Optometry-Ophthalmology Network e-research (SCONe) currently contains over 367 000 retinal images matched to over 36 000 patients. Healthcare data (hospital inpatient and outpatient, general ophthalmic, death and prescribing) records were retrieved for patients with one or more images, providing demographic and healthcare information for 95% of the cohort. The linked data allow the application of condition labels or phenotypes at specific points in time, facilitating research into retinal manifestations of vascular and neural diseases. The cohort is representative of the Scottish 60+ population in terms of sex (54% female), and there is a slight over-representation of people of black, Asian and minority ethnic groups (2% vs 1%) and those living in areas of lower deprivation (30% vs 16% in lowest two categories). Early research work has begun and is focusing on ocular and neurodegenerative disease prediction.CONCLUSIONS: The SCONe retinal image repository has been successfully established. We believe it offers enormous potential to support research into earlier detection of disease.</p
Soaring with TRILLI:an HW/SW heterogeneous accelerator for multi-modal image registration
Get PDF3D rigid image registration is a pivotal procedure in computer vision that aligns a floating volume with a reference one to correct positional and rotational distortions. It serves either as a stand-alone process or as a pre-processing stepfor non-rigid registration, where the rigid part dominates the computational cost. Various hardware accelerators have been proposed to optimize its compute-intensive components: geometric transformation with interpolation and similarity metric computation. However, existing solutions fail to address both components effectively, as GPUs excel at image transformation, while FPGAs in similarity metric computation. To close this gap, we propose TRILLI, a novel Versal-based accelerator for image transformation and interpolation. TRILLI optimally maps each computational step on the proper heterogeneous hardware component. TRILLI achieves speedup of 5.32× against the top performing GPU-based solution, and an energy efficiency improvement of 36.75× against the most efficient one. Moreover, we integrate it with an FPGA-based similarity metric from literature to complete a rigid image registration step (i.e., transformation, interpolation, and similarity metric) attaining a speedup of 18.60× against the top performing GPU-based solution, whil
Expansion microscopy reveals nano-scale insights into the human neuromuscular junction
Get PDFThe neuromuscular junction (NMJ) is a specialized synapse that relays signals from the lower motor neuron to the skeletal muscle. Here, we detail the development and application of expansion microscopy (ExM) as a highly accessible, relatively cheap, powerful, and reproducible tool with which to obtain high-resolution insights into the subcellular structure and function of NMJs from whole-mount preparations, previously only achievable using super-resolution microscopy. ExM is equally applicable to both mouse and human tissue samples, facilitating high-resolution comparative analyses. Qualitative and quantitative analysis of ExM images reveals significant differences in the distribution of acetylcholine receptors, synaptic vesicles, and voltage-gated Na+ 1.4 (NaV1.4) channels between human and mouse NMJs that are not readily observable using conventional confocal microscopy. We conclude that ExM offers a cost-effective and adaptable approach to facilitate nano-scale imaging of the NMJ
A computational ontology framework for the synthesis of multi-level pathology reports from brain MRI scans
Get PDFBackgroundConvolutional neural network (CNN) based volumetry of MRI data can help differentiate Alzheimer's disease (AD) and the behavioral variant of frontotemporal dementia (bvFTD) as causes of cognitive decline and dementia. However, existing CNN-based MRI volumetry tools lack a structured hierarchical representation of brain anatomy, which would allow for aggregating regional pathological information and automated computational inference.ObjectiveDevelop a computational ontology pipeline for quantifying hierarchical pathological abnormalities and visualize summary charts for brain atrophy findings, aiding differential diagnosis.MethodsUsing FastSurfer, we segmented brain regions and measured volume and cortical thickness from MRI scans pooled across multiple cohorts (N = 3433; ADNI, AIBL, DELCODE, DESCRIBE, EDSD, and NIFD), including healthy controls, prodromal and clinical AD cases, and bvFTD cases. Employing the Web Ontology Language (OWL), we built a semantic model encoding hierarchical anatomical information. Additionally, we created summary visualizations based on sunburst plots for visual inspection of the information stored in the ontology.ResultsOur computational framework dynamically estimated and aggregated regional pathological deviations across different levels of neuroanatomy abstraction. The disease similarity index derived from the volumetric and cortical thickness deviations achieved an AUC of 0.88 for separating AD and bvFTD, which was also reflected by distinct atrophy profile visualizations.ConclusionsThe proposed automated pipeline facilitates visual comparison of atrophy profiles across various disease types and stages. It provides a generalizable computational framework for summarizing pathologic findings, potentially enhancing the physicians' ability to evaluate brain pathologies robustly and interpretably.</p
A Comparative Analysis of Technical Data:At-Home vs. In-Clinic Application of Transcranial Direct Current Stimulation in Depression
Get PDFOBJECTIVE: The application of transcranial direct current stimulation (tDCS) at home for the treatment of depression and other neuropsychiatric disorders presents both significant opportunities and inherent challenges. Ensuring safety and maintaining high-quality stimulation are paramount for the efficacy and safety of at-home tDCS. This study investigates tDCS quality based on its technical parameters as well as safety of at-home and in-clinic tDCS applications comparing the data from two randomized controlled trials in patients with major depressive disorder.METHODS: We analyzed 229 active stimulation sessions from the HomeDC study (at-home tDCS) and 835 sessions from the DepressionDC study (in-clinic tDCS). Notably, five adverse events (skin lesions) were reported exclusively in the at-home cohort, highlighting the critical need for enhanced safety protocols in unsupervised environments.RESULTS: The analysis revealed a significant difference in the average variability of impedances between at-home and in-clinic applications (F1,46 = 4.96, p = .031, η 2 = .097). The at-home tDCS sessions exhibited higher impedance variability (M = 837, SD = 328) compared to in-clinic sessions (M = 579, SD = 309). Furthermore, at-home tDCS sessions resulting in adverse events (AEs) were associated with significantly higher average impedances than sessions without such issues. CONCLUSION: The study demonstrates that monitoring the technical parameters of at-home tDCS used in this study is essential. However, it may be not sufficient for ensuring safety and promptly detecting or preventing adverse events. Quality control protocols including digital training and monitoring techniques should be systematically developed and tested for a reliable and safe application of at-home tDCS therapies.</p
Machine learning to optimize use of natriuretic peptides in the diagnosis of acute heart failure
Get PDFAIMS: B-type natriuretic peptide (BNP) and mid-regional pro-atrial natriuretic peptide (MR-proANP) testing are guideline-recommended to aid in the diagnosis of acute heart failure. Nevertheless, the diagnostic performance of these biomarkers is uncertain.METHODS: We performed a systematic review and individual patient-level data meta-analysis to evaluate the diagnostic performance of BNP and MR-proANP. We subsequently developed and externally validated a decision-support tool called CoDE-HF that combines natriuretic peptide concentrations with clinical variables using machine learning to report the probability of acute heart failure.RESULTS: Fourteen studies from 12 countries provided individual patient-level data in 8,493 patients for BNP and 3,899 patients for MR-proANP, in whom, 48.3% (4,105/8,493) and 41.3% (1,611/3,899) had an adjudicated diagnosis of acute heart failure, respectively. The negative predictive value (NPV) of guideline-recommended thresholds for BNP (100 pg/mL) and MR-proANP (120 pmol/L) was 93.6% (95% confidence interval 88.4-96.6%) and 95.6% (92.2-97.6%), respectively, whilst the positive predictive value (PPV) was 68.8% (62.9-74.2%) and 64.8% (56.3-72.5%). Significant heterogeneity in the performance of these thresholds was observed across important subgroups. CoDE-HF was well calibrated with excellent discrimination in those without prior acute heart failure for both BNP and MR-proANP (area under the curve of 0.914 [0.906-0.921] and 0.929 [0.919-0.939], and Brier scores of 0.110 and 0.094, respectively). CoDE-HF with BNP and MR-proANP identified 30% and 48% as low-probability (NPV of 98.5% [97.1-99.3%] and 98.5% [97.7-99.0%]), and 30% and 28% as high-probability (PPV of 78.6% [70.4-85.0%] and 75.1% [70.9-78.9%]), respectively, and performed consistently across subgroups.CONCLUSION: The diagnostic performance of guideline-recommended BNP and MR-proANP thresholds for acute heart failure varied significantly across patient subgroups. A decision-support tool that combines natriuretic peptides and clinical variables was more accurate and supports more individualised diagnosis.STUDY REGISTRATION: PROSPERO number, CRD42019159407.</p
A cross-sectional study of symptom prevalence, frequency, severity, and impact of long-COVID in Scotland:Part I
Get PDFBackgroundCommonly reported symptoms of long COVID may have different patterns of prevalence and presentation across different countries. While some limited data have been reported for the United Kingdom, national specificity for Scotland is less clear. We present a cross-sectional survey to examine the symptom prevalence, frequency, and severity of long COVID for people living with the condition in Scotland.MethodsAn online survey was created in the English language and was available between April 21, 2022 and August 5, 2022. Participants were included if they were ≥18 years old, living in Scotland, and had self-diagnosed or confirmed long COVID; and excluded if they were hospitalized during their initial infection. Within this article we quantify symptom prevalence, frequency, severity, and duration.ResultsParticipants (n = 253) reported the most prevalent long-COVID symptoms to be post-exertional malaise (95%), fatigue/tiredness (85%), and cognitive impairment (68%). Fatigue/tiredness, problems with activities of daily living (ADL), and general pain were most frequently occurring, while sleep difficulties, problems with ADL, and nausea were the most severe. Scottish Index of Multiple Deprivation associated with symptom number, severity, and frequency, whereas vaccine status, age, sex, and smoking status had limited or no association.ConclusionsThese findings outline the challenges faced for those living with long COVID and highlight the need for longitudinal research to ascertain a better understanding of the condition and its longer-term societal impact.<br/
Repairability as a condition of the world: Ernesto Oroza’s archive of dis/repair
Get PDFIn an age of apparent disrepair as the climate crisis takes hold and neoliberalism fails to liberate, as the cost of living rises and rights are retracted, the need for a reparative turn is overdue. But what is repair? If repair is contained in moments of total breakdown, then the reparative acts of care that sustain the world are denied. Countering these forces and the urgency prescribed by the crisis of disrepair and in what too often appears as the proprietary epistemology of repair, in this paper I offer an account of ‘repairability’. Structured in relation to the reparative gaze of feminist theory and poetic thinking, repairability assumes a material trace, I contend, through the vernacular archive of Cuban artist-ethnographer Ernesto Oroza. Oroza’s work offers a compelling case study through which to think the possibility of repair as an act of worldly becoming
Access Repository Dashboard
Do you manage Edinburgh Research Explorer? Access insider analytics, issue reports and manage access to outputs from your repository in the CORE Repository Dashboard!