Publications

VolumeScaler is an open-source Kubernetes controller that automates persistent-volume-claim scaling. The work focuses on improving storage management in container-orchestrated systems by detecting capacity pressure and applying controlled volume expansion without requiring manual intervention.

Paper

BibTeX

@article{Ghanem2026, doi={10.21105/joss.08846}, year={2026}, publisher={The Open Journal}, volume={11}, number={119}, pages={8846}, author={Ghanem, Ahmed and O'Reilly, Ruairi}, title={VolumeScaler: A Kubernetes Controller for Automated PVC Scaling}, journal={Journal of Open Source Software}}

This work develops an adaptive ECG signal-quality assessment designed to transfer across psychophysiological datasets without extensive dataset-specific tuning or large manually annotated training sets. It evaluates signal quality under varied recording conditions and examines how artefacts can affect downstream heart-rate-variability features and emotion classification.

Paper

BibTeX

@article{dair2026datasetagnostic, title={Towards Dataset Agnostic ECG Signal Quality Assessments for Psychophysiological Analysis}, author={Dair, Zachary and Dockray, Samantha and O'Reilly, Ruairi}, journal={Journal of Psychophysiology}, volume={40}, number={1}, pages={12--29}, year={2026}, doi={10.1027/0269-8803/a000357}}

Abstract

Biomedical image datasets can be imbalanced due to the rarity of targeted diseases. Generative Adversarial Networks play a key role in addressing this imbalance by enabling the generation of synthetic images to augment datasets. It is important to generate synthetic images that incorporate a diverse range of features to accurately represent the distribution of features present in the training imagery. Furthermore, the absence of diverse features in synthetic images can degrade the performance of machine learning classifiers. The mode collapse problem impacts Generative Adversarial Networks’ capacity to generate diversified images. Mode collapse comes in two varieties: intra-class and inter-class. In this paper, both varieties of the mode collapse problem are investigated, and their subsequent impact on the diversity of synthetic X-ray images is evaluated. This work contributes an empirical demonstration of the benefits of integrating the adaptive input-image normalization with the Deep Convolutional GAN and Auxiliary Classifier GAN to alleviate the mode collapse problems. Synthetically generated images are utilized for data augmentation and training a Vision Transformer model. The classification performance of the model is evaluated using accuracy, recall, and precision scores. Results demonstrate that the DCGAN and the ACGAN with adaptive input-image normalization outperform the DCGAN and ACGAN with un-normalized X-ray images as evidenced by the superior diversity scores and classification scores.

Paper Google Scholar

BibTeX

@article{saad2026adaptive, title={Adaptive input-image normalization for solving the mode collapse problem in GAN-based X-ray images}, author={Saad, Muhammad Muneeb and Rehmani, Mubashir Husain and O'Reilly, Ruairi}, journal={Biomedical Signal Processing and Control}, volume={111}, pages={108333}, year={2026}, doi={10.1016/j.bspc.2025.108333}}

Abstract

Alzheimer’s disease (AD) is a neurodegenerative disorder marked by memory loss and cognitive decline, making early detection vital for timely intervention. However, early diagnosis is challenging due to the heterogeneous presentation of symptoms. Resting-state functional magnetic resonance imaging (rs-fMRI) captures spontaneous brain activity and functional connectivity, which are known to be disrupted in AD and mild cognitive impairment (MCI). Traditional methods, such as Pearson’s correlation, have been used to calculate association matrices, but these approaches often overlook the dynamic and non-stationary nature of brain activity. In this study, we introduce a novel method that integrates discrete wavelet transform (DWT) and graph theory to model the dynamic behavior of brain networks. Our approach captures the time-frequency representation of brain activity, allowing for a more nuanced analysis of

Google Scholar

BibTeX

@article{khazaee2025multi, title={Multi-scale graph theoretical analysis of resting-state fMRI for classification of Alzheimer’s disease, mild cognitive impairment, and healthy controls}, author={Khazaee, Ali and Mohammadi, Abdolreza and O’Reilly, Ruairi}, journal={Signal, Image and Video Processing}, volume={19}, number={14}, pages={1192}, year={2025}, publisher={Springer London London}}

Abstract

Most research investigating relationships between the Big Five and emotional states has focused on how emotional attributes relate to Extraversion and Neuroticism. However, the potential for discrete emotional states to enable a richer understanding of the emotive nature of all Big Five traits and their subtraits has been neglected. Methods Participants (N = 203) completed the Big Five Aspects Scale, watched six emotionally stimulating video clips, and self‐reported their experience of basic emotions before (Baseline) and after (Reaction) each video. Spearman correlations identified state–trait relationships, followed by regression analyses to assess the unique contribution of each trait to emotional experiences. Results Conscientiousness negatively correlated with Baseline Sadness, while Agreeableness positively correlated with Reaction Disgust, Fear, and Sadness.

Google Scholar

BibTeX

@article{donovan2025investigating, title={Investigating the Relationships Between Basic Emotions and the Big Five Personality Traits and Their Sub-Traits}, author={Donovan, Ryan and Johnson, Aoife and De Roiste, Aine and O'Reilly, Ruairi}, journal={Journal of Personality}, year={2025}}

Abstract

In the context of Machine Learning (ML) and Artificial Intelligence (AI), the concepts of sufficiency and necessity of features offer nuanced perspectives on the cause-and-effect relationships underlying a model’s outputs. These concepts are, therefore, essential in Explainable AI (XAI) as they can provide a more holistic understanding of a ‘black-box’ AI model. Addressing this need, our study explored the relationships between the XAI’s explanations and the sufficiency and necessity of features in data. This is achieved by emphasising the impact of neighbourhoods, which are central in generating explanations. By analysing a diverse set of neighbourhoods, we highlighted how they influence the alignment between the feature rankings by XAI and the measures of sufficiency and necessity. This work offers two contributions. First, it provides a comprehensive discussion on how XAI frameworks relate to sufficiency and necessity with respect to their operating neighbourhoods; and second, it empirically demonstrates the effectiveness of these neighbourhoods in conveying the sufficiency and necessity of features by the XAI frameworks.

BibTeX

@article{pawar2024optimal, title={Optimal Neighborhood Contexts in Explainable AI: An Explanandum-Based Evaluation}, author={Pawar, Urja and O'Shea, Donna and O'Reilly, Ruairi and Costello, Maebh and Beder, Christian}, journal={IEEE Open Journal of the Computer Society}, year={2024}, volume={5}, pages={181--194}, doi={10.1109/OJCS.2024.3389781}, publisher={IEEE} }

Abstract

In biomedical image analysis, the applicability of deep learning methods is directly impacted by the quantity of image data available. This is due to deep learning models requiring large image datasets to provide high-level performance. Generative Adversarial Networks (GANs) have been widely utilized to address data limitations through the generation of synthetic biomedical images. GANs consist of two models. The generator, a model that learns how to produce synthetic images based on the feedback it receives. The discriminator, a model that classifies an image as synthetic or real and provides feedback to the generator. Throughout the training process, a GAN can experience several technical challenges that impede the generation of suitable synthetic imagery. First, the mode collapse problem whereby the generator either produces an identical image or produces a uniform image from distinct input features. Second, the non-convergence problem whereby the gradient descent optimizer fails to reach a Nash equilibrium. Thirdly, the vanishing gradient problem whereby unstable training behavior occurs due to the discriminator achieving optimal classification performance resulting in no meaningful feedback being provided to the generator. These problems result in the production of synthetic imagery that is blurry, unrealistic, and less diverse. To date, there has been no survey article outlining the impact of these technical challenges in the context of the biomedical imagery domain. This work presents a review and taxonomy based on solutions to the training problems of GANs in the biomedical imaging domain. This survey highlights important challenges and outlines future research directions about the training of GANs in the domain of biomedical imagery.

Paper Google Scholar

BibTeX

@article{saad2024survey, title={A survey on training challenges in generative adversarial networks for biomedical image analysis}, author={Saad, Muhammad Muneeb and O’Reilly, Ruairi and Rehmani, Mubashir Husain}, journal={Artificial Intelligence Review}, volume={57}, number={2}, pages={19}, year={2024}, publisher={Springer Netherlands Dordrecht} }

BibTeX

@inproceedings{bahamazava2025fair, title={FAIR-MED: Bias Detection and Fairness Evaluation in Healthcare Focused XAI}, author={Bahamazava, Katsiaryna and O’Reilly, Ruairi}, booktitle={World Conference on Explainable Artificial Intelligence}, pages={380--401}, year={2026}, organization={Springer Nature Switzerland Cham}}

Abstract

This paper presents a discussion on utilising a dashboard tool to enhance the interpretability of SHapley Additive exPlanations (SHAP) in healthcare Artificial Intelligence (AI)-based applications. Despite SHAP’s potential to demystify an AI model’s decisions, interpreting SHAP values remains challenging, especially when considering different data neighbourhoods [1]. This issue is particularly critical in healthcare, where decision-making requires high precision and clarity. We demonstrate three use cases that can effectively demonstrate the utility of interactive neighborhood exploration. The first compares SHAP explanations in two similar patient neighbourhoods with different classifications, offering unique insights into features that influence classification changes. The second use case focuses on “feature freezing” which isolates certain features to better understand their impact. This can enable highlighting diagnostic tests considered important by a Machine Learning (ML) model for a specific population of patients (eg, patients of the same age). The final use case demonstrates the relationship between sufficient features for a given classification and the importance ranking by SHAP.

Paper Google Scholar

BibTeX

@article{pawar2024dynamics, title={The Dynamics of Explainability: Diverse Insights from SHAP Explanations using Neighbourhoods}, author={Pawar, Urja and O’Reilly, Ruairi and Beder, Christian and O’Shea, Donna}, year={2024} }

Abstract

In the context of Machine Learning (ML) and Artificial Intelligence (AI), the concepts of sufficiency and necessity of features offer nuanced perspectives on the cause-and-effect relationships underlying a model’s outputs. These concepts are, therefore, essential in Explainable AI (XAI) as they can provide a more holistic understanding of a ‘‘black-box’’ AI model. Addressing this need, our study explored the relationships between the XAI’s explanations and the sufficiency and necessity of features in data. This is achieved by emphasising the impact of neighbourhoods, which are central in generating explanations. By analysing a diverse set of neighbourhoods, we highlighted how they influence the alignment between the feature rankings by XAI and the measures of sufficiency and necessity. This work offers two contributions. First, it provides a comprehensive discussion on how XAI frameworks relate to sufficiency and necessity with respect to their operating neighbourhoods; and second, it empirically demonstrates the effectiveness of these neighbourhoods in conveying the sufficiency and necessity of features by the XAI frameworks.

Paper Google Scholar

BibTeX

@inproceedings{pawar2024impact, title={On the Impact of Neighbourhood Sampling to Satisfy Sufficiency and Necessity Criteria in Explainable AI}, author={Pawar, Urja and Beder, Christian and Ruairi, O and Donna, O and others}, booktitle={Causal Learning and Reasoning}, pages={570--586}, year={2024}, organization={Proceedings of Machine Learning Research} }

Abstract

Generative Adversarial Networks (GANs) have high computational costs to train their complex architectures. Throughout the training process, GANs’ output is analyzed qualitatively based on the loss and synthetic images’ diversity and quality. Based on this qualitative analysis, training is manually halted once the desired synthetic images are generated. By utilizing an early stopping criterion, the computational cost and dependence on manual oversight can be reduced yet impacted by training problems such as mode collapse, non-convergence, and instability. This is particularly prevalent in biomedical imagery, where training problems degrade the diversity and quality of synthetic images, and the high computational cost associated with training makes complex architectures increasingly inaccessible. This work proposes a novel early stopping criteria to quantitatively detect training problems, halt training, and reduce the computational costs associated with synthesizing biomedical images. Firstly, the range of generator and discriminator loss values is investigated to assess whether mode collapse, non-convergence, and instability occur sequentially, concurrently, or interchangeably throughout the training of GANs. Secondly, utilizing these occurrences in conjunction with the Mean Structural Similarity Index (MS-SSIM) and Frechet Inception Distance (FID) scores of synthetic images forms the basis of the proposed early stopping criteria. This work helps identify the occurrence of training problems in GANs using low-resource computational cost and reduces training time to generate diversified and high-quality synthetic images.

Paper Google Scholar

BibTeX

@inproceedings{saad2024early, title={Early Stopping Criteria for Training Generative Adversarial Networks in Biomedical Imaging}, author={Saad, Muhammad Muneeb and Rehmani, Mubashir Husain and O'Reilly, Ruairi}, booktitle={2024 35th Irish Signals and Systems Conference (ISSC)}, year={2024}, doi={10.1109/ISSC61953.2024.10603268}}

Abstract

WebTransport is a new protocol which enables the opening of a persistent connection between a client and server for bi-directional, uni-directional and datagram data exchange. The WebTransport protocol is currently in a draft state within the Internet Engineering Task Force (IETF). WebTransport is designed to use the QUIC protocol as its transport mechanism, which provides packet ordering and delivery management. This article provides an overview of the WebTransport protocol, summarises prior work on benchmarking QUIC, and compares WebTransport and WebSockets. The findings indicate that Web-Transport does not match the performance of WebSockets with the current language implementations.

Google Scholar

BibTeX

@inproceedings{williamson2023webtransport, title={WebTransport and WebSockets: An Empirical Analysis of Connection Time, Message Response, and Payload Efficiency}, author={Williamson, Declan and O’Reilly, Ruairi}, booktitle={2023 34th Irish Signals and Systems Conference (ISSC)}, pages={1--6}, year={2023}, organization={IEEE}} 

Abstract

Glaucoma is an irreversible ocular disease and is the second leading cause of visual disability worldwide. Slow vision loss and the asymptomatic nature of the disease make its diagnosis challenging. Early detection is crucial for preventing irreversible blindness. Ophthalmologists primarily use retinal fundus images as a non-invasive screening method. Convolutional neural networks (CNN) have demonstrated high accuracy in the classification of medical images. Nevertheless, CNN’s translation-invariant nature and inability to handle the part-whole relationship between objects make its direct application unsuitable for glaucomatous fundus image classification, as it requires a large number of labelled images for training. This work reviews existing state of the art models and proposes InceptionCaps, a novel capsule network (CapsNet) based deep learning model having pre-trained InceptionV3 as its convolution base, for automatic glaucoma classification. InceptionCaps achieved an accuracy of 0.956, specificity of 0.96, and AUC of 0.9556, which surpasses several state-of-the-art deep learning model performances on the RIM-ONE v2 dataset. The obtained result demonstrates the robustness of the proposed deep learning model.

Paper Google Scholar

BibTeX

@inproceedings{manohar2023inceptioncaps, title={InceptionCaps: A Performant Glaucoma Classification Model for Data-Scarce Environment}, author={Manohar, Gyanendar and O'Reilly, Ruairi}, booktitle={2023 31st Irish Conference on Artificial Intelligence and Cognitive Science (AICS)}, pages={1--8}, year={2023}, organization={IEEE} }

Abstract

The diverse nature of emotion models presents challenges in psychophysiological analysis, notably affecting the comparability and reliability of research findings across studies. This exploratory work aims to address the diversity of emotion models by abstracting them into a higher-order perspective while preserving the specificity of the original emotions. This will be accomplished by integrating Complex Adaptive Systems utilising their inherent interactivity to foster rich interactions between different emotion models, adaptability to adjust to the nuances of each model, and self-organisation to autonomously form a coherent understanding of diverse emotional data. By analysing interactions within the system, this study aims to identify abstract emotion representations that capture dynamic patterns in physiological data that indicate emotional states. Combining these dynamic patterns with an analysis of system interactions offers a novel perspective in emotion modelling.

Google Scholar

BibTeX

@inproceedings{dair2023complex, title={Complex Adaptive Systems and Psychophysiological Data-An Exploratory Approach}, author={Dair, Zachary and Dockray, Samantha and O'Reilly, Ruairi}, booktitle={31st Irish Conference on Artificial Intelligence and Cognitive Science (AICS)}, pages={1--4}, year={2023}, organization={IEEE} }

Abstract

In biomedical image analysis, data imbalance is common across several imaging modalities. Data augmentation is one of the key solutions in addressing this limitation. Generative Adversarial Networks (GANs) are increasingly being relied upon for data augmentation tasks. Biomedical image features are sensitive to evaluating the efficacy of synthetic images. These features can have a significant impact on metric scores when evaluating synthetic images across different biomedical imaging modalities. Synthetically generated images can be evaluated by comparing the diversity and quality of real images. Multi-scale Structural Similarity Index Measure and Cosine Distance are used to evaluate intra-class diversity, while Frechet Inception Distance is used to evaluate the quality of synthetic images. Assessing these metrics for biomedical and non-biomedical imaging is important to investigate an informed strategy in evaluating the diversity and quality of synthetic images. In this work, an empirical assessment of these metrics is conducted for the Deep Convolutional GAN in a biomedical and non-biomedical setting. The diversity and quality of synthetic images are evaluated using different sample sizes. This research intends to investigate the variance in diversity and quality across biomedical and non-biomedical imaging modalities. Results demonstrate that the metrics scores for diversity and quality vary significantly across biomedical-to-biomedical and biomedical-to-non-biomedical imaging modalities.

Paper Google Scholar

BibTeX

@article{saad2023assessing, title={Assessing Intra-class Diversity and Quality of Synthetically Generated Images in a Biomedical and Non-biomedical Setting}, author={Saad, Muhammad Muneeb and Rehmani, Mubashir Husain and O'Reilly, Ruairi}, journal={arXiv preprint arXiv:2308.02505}, year={2023} }

Abstract

Imbalanced image datasets are commonly available in the domain of biomedical image analysis. Biomedical images contain diversified features that are significant in predicting targeted diseases. Generative Adversarial Networks (GANs) are utilized to address the data limitation problem via the generation of synthetic images. Training challenges such as mode collapse, non-convergence, and instability degrade a GAN’s performance in synthesizing diversified and high-quality images. In this work, MSG-SAGAN, an attention-guided multi-scale gradient GAN architecture is proposed to model the relationship between long-range dependencies of biomedical image features and improves the training performance using a flow of multi-scale gradients at multiple resolutions in the layers of generator and discriminator models. The intent is to reduce the impact of mode collapse and stabilize the training of GAN using an attention mechanism with multi-scale gradient learning for diversified X-ray image synthesis. Multi-scale Structural Similarity Index Measure (MS-SSIM) and Frechet Inception Distance (FID) are used to identify the occurrence of mode collapse and evaluate the diversity of synthetic images generated. The proposed architecture is compared with the multi-scale gradient GAN (MSG-GAN) to assess the diversity of generated synthetic images. Results indicate that the MSG-SAGAN outperforms MSG-GAN in synthesizing diversified images as evidenced by the MS-SSIM and FID scores.

Paper Google Scholar ResearchGate

BibTeX

@inproceedings{muneeb2023self, title={A Self-attention Guided Multi-scale Gradient GAN for Diversified X-ray Image Synthesis}, author={Muneeb Saad, Muhammad and Husain Rehmani, Mubashir and O'Reilly, Ruairi}, booktitle={Artificial Intelligence and Cognitive Science. AICS 2022}, volume={1662}, number={Springer CCIS}, year={2023}, organization={https://doi.org/10.1007/978-3-031-26438-2_2} }

Abstract

Psychophysiology investigates the causal relationship of physiological changes resulting from psychological states. There are significant challenges with machine learning-based momentary assessments of physiology due to varying data collection methods, physiological differences, data availability and the requirement for expertly annotated data. Advances in wearable technology have significantly increased the scale, sensitivity and accuracy of devices for recording physiological signals, enabling large-scale unobtrusive physiological data gathering. This work contributes an empirical evaluation of signal variances acquired from wearables and their associated impact on the classification of affective states by (i) assessing differences occurring in features representative of affective states extracted from electrocardiograms and photoplethysmography, (ii) investigating the disparity in feature importance between signals to determine signal-specific features, and (iii) investigating the disparity in feature importance between affective states to determine affect-specific features. Results demonstrate that the degree of feature variance between ECG and PPG in a dataset is reflected in the classification performance of that dataset. Additionally, beats-per-minute, inter-beat-interval and breathing rate are identified as common best-performing features across both signals. Finally feature variance per-affective state identifies hard-to-distinguish affective states requiring one-versus-rest or additional features to enable accurate classification.

Paper Google Scholar

BibTeX

@inproceedings{dair2022inter, title={Inter and intra signal variance in feature extraction and classification of affective state}, author={Dair, Zachary and Dockray, Samantha and O’Reilly, Ruairi}, booktitle={Irish Conference on Artificial Intelligence and Cognitive Science}, pages={3--17}, year={2022}, organization={Springer Nature Switzerland Cham} }

Abstract

In the domain of skin lesion classification using computer-aided diagnosis, machine learning approaches found in the literature are reported to be highly effective. However, state-of-the-art findings can prove challenging to reimplement due to inconsistencies and ambiguities in recorded methodologies. These ambiguities reduce the velocity at which future research advancements can be achieved. This paper proposes a machine learning configuration capture method that obtains a complete and faithful descriptor of a machine learning workflow. This descriptor is serialised into a sharable file format, enabling sub-sequent research to reimplement a cited model to a high degree of accuracy. Following this configuration capture, reproducing input data sources is an essential step in the faithful reimple-mentation of baseline models. A centralised data sourcing tool for the automated acquisition of highly cited skin lesion datasets from various sources is also delivered. The work contributes a standardised approach in creating reproducible and sharable machine learning-based workflows, enabling accelerated machine learning research in the domain of skin lesion classification.

Paper Google Scholar

BibTeX

@inproceedings{walshe2022fair, title={FAIR Skin Lesion Classification Workflows using Transfer Learning}, author={Walshe, David and O'Reilly, Ruairi}, booktitle={2022 33rd Irish Signals and Systems Conference (ISSC)}, pages={1--6}, year={2022}, organization={IEEE} }

Abstract

Sleep apnea is one of the most common sleep disorders. To diagnose sleep apnea, a patient must undertake a polysomnography where multiple physiological signals are recorded in a specialised sleep laboratory. Reducing the number of physiological signals necessary for a diagnosis and enabling data monitoring in a distributed fashion would assist in the detection of sleep apnea. Smartwatches are becoming more advanced, with the current generation capable of deriving blood oxygen saturation, which can indicate sleep apnea. This work evaluates the efficacy of sleep apnea classifiers in a simulated smartwatch environment. Results demonstrate that SpO2 is a performant signal for classifying sleep apnea. Naive Bayes trained with features extracted from a Long Short Term Memory Network is capable of classifying sleep apnea with an accuracy of 97.04%, outperforming state-of-the-art approaches. Classification within the simulated smartwatch environment demonstrates robustness up to a signal-to-noise ratio of 50 dB and maintains high levels of accuracy at sampling frequencies above 25 Hz. These encouraging results show substantial potential for smartwatches to provide timely, accessible sleep apnea screening and enable automated diagnostics reducing the reliance on specialist centres.

Google Scholar

BibTeX

@inproceedings{lyden2022classification, title={Classification of Sleep Apnea via SpO2 in a Simulated Smartwatch Environment}, author={Lyden, Brendan and Dair, Zachary and O'Reilly, Ruairi}, booktitle={2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)}, pages={1--5}, year={2022}, organization={IEEE} }

Abstract

Biomedical image datasets can be imbalanced due to the rarity of targeted diseases. Generative Adversarial Networks play a key role in addressing this imbalance by enabling the generation of synthetic images to augment datasets. It is important to generate synthetic images that incorporate a diverse range of features to accurately represent the distribution of features present in the training imagery. Furthermore, the absence of diverse features in synthetic images can degrade the performance of machine learning classifiers. The mode collapse problem impacts Generative Adversarial Networks’ capacity to generate diversified images. Mode collapse comes in two varieties: intra-class and inter-class. In this paper, the intra-class mode collapse problem is investigated, and its subsequent impact on the diversity of synthetic X-ray images is evaluated. This work contributes an empirical demonstration of the benefits of integrating the adaptive input-image normalization for the Deep Convolutional GAN to alleviate the intra-class mode collapse problem. Results demonstrate that the DCGAN with adaptive input-image normalization outperforms DCGAN with un-normalized X-ray images as evident by the superior diversity scores.

Paper Google Scholar

BibTeX

@inproceedings{saad2022addressing, title={Addressing the intra-class mode collapse problem using adaptive input image normalization in GAN-based X-ray images}, author={Saad, Muhammad Muneeb and Rehmani, Mubashir Husain and O'Reilly, Ruairi}, booktitle={2022 44th annual international conference of the IEEE Engineering in Medicine and Biology Society (EMBC)}, pages={2049--2052}, year={2022}, organization={IEEE}}

Abstract

Sleep apnea is one of the most common sleep disorders. It can have a variety of detrimental effects ranging from trouble breathing to increased risk of heart failure. Sleep apnea often goes undiagnosed due to the practicalities and cost of getting tested. To be diagnosed with sleep apnea, a patient must undertake a polysomnography (PSG) or sleep study where multiple physiological signals are recorded in a specialised sleep laboratory. Diagnosing sleep apnea using data gathered in a distributed manner could help to remove barriers to diagnosis. Prior work has largely focused on creating machine learning models trained on a subset of the physiological signals that are recorded in a medical setting. Activity trackers are becoming more advanced with the current generation being capable of deriving blood oxygen saturation (SpO2). SpO2 has been shown to be an important indicator for sleep apnea. Activity tracker data acquisition faces challenges, such as the sampling rate of data and noise associated with the signal. This work compares the performance of ML models in a simulated activity tracker environment. The results show that a Support Vector Machine (SVM) classifier trained with features extracted from a Convolutional Neural Network (CNN) is capable of classifying sleep apnea with a high degree of accuracy.

ResearchGate

Abstract

In the domain of skin lesion classification using computer-aided diagnosis (CAD), machine learning (ML) approaches found in the literature are reported to be highly effective. However, state-of-the-art findings can prove challenging to reimplement due to inconsistencies and ambiguities in recorded methodologies. These ambiguities reduce the velocity at which future research advancements can be achieved. This abstract proposes an ML configuration capture method that obtains a complete and faithful descriptor of an ML workflow. This descriptor is serialised into a sharable file format, enabling subsequent research to reimplement a cited model to a high degree of accuracy. Following this configuration capture, reproducing input data sources is an essential step in the faithful reimplementation of baseline models. This abstract also delivers a centralised data sourcing tool for the automated acquisition of highly cited skin lesion datasets from various sources. The work contributes a standardised approach in creating reproducible and sharable ML-based workflows, enabling accelerated machine learning research in the domain of skin lesion classification.

ResearchGate

Abstract

This work explores the effect of gender and linguistic-based vocal variations on the accuracy of emotive expression classification. Emotive expressions are considered from the perspective of spectral features in speech (Mel-frequency Cepstral Coefficient, Melspectrogram, Spectral Contrast). Emotions are considered from the perspective of Basic Emotion Theory. A convolutional neural network is utilised to classify emotive expressions in emotive audio datasets in English, German, and Italian. Vocal variations for spectral features assessed by (i) a comparative analysis identifying suitable spectral features, (ii) the classification performance for mono, multi and cross-lingual emotive data and (iii) an empirical evaluation of a machine learning model to assess the effects of gender and linguistic variation on classification accuracy. The results showed that spectral features provide a potential avenue for increasing emotive expression classification. Additionally, the accuracy of emotive expression classification was high within mono and cross-lingual emotive data, but poor in multi-lingual data. Similarly, there were differences in classification accuracy between gender populations. These results demonstrate the importance of accounting for population differences to enable accurate speech emotion recognition.

Paper

BibTeX

@article{dair2021linguistic, title={Linguistic and gender variation in speech emotion recognition using spectral features}, author={Dair, Zachary and Donovan, Ryan and O'Reilly, Ruairi}, journal={arXiv preprint arXiv:2112.09596}, year={2021} }

Abstract

The applicability and utility of Artificial Intelligence (AI) based solutions has been demonstrated widely in the healthcare domain via the automated analysis of medical information. However, the adoption rate of AI-based healthcare systems is inhibited due to their complicated nature. Also, the hierarchical nature of the medical settings adds a layer of complexity in understanding how a Machine Learning (ML) model functions when subjected to varying workflows. Variation in models’ performance, and individual features’ contribution, needs to be effectively quantified such that medical practitioners can understand the models, and validate their operation if widespread adoption is to be enabled. In this paper, a hierarchical medical workflow for understanding the operation of ML in a healthcare-based setting is proposed. Its utility is demonstrated in the context of heart disease classification. Explainable Artificial Intelligence (XAI) is incorporated in the form of Feature Importance (FI) scores and correlated with an ML model’s performance metrics (Accuracy, F1-score). This provides a multi-stakeholder perspective aligned with the hierarchy as experienced in a real-world medical setting. The paper contributes a methodology for accommodating an enhanced understanding of diverse hierarchical healthcare settings that would benefit from the adoption of AI-based systems.

Google Scholar ResearchGate

BibTeX

@INPROCEEDINGS{9474678, author={Pawar, Urja and Culbert, Christopher T and O'Reilly, Ruairi}, booktitle={2021 IEEE 34th International Symposium on Computer-Based Medical Systems (CBMS)}, title={Evaluating Hierarchical Medical Workflows using Feature Importance}, year={2021}, volume={}, number={}, pages={265-270}, doi={10.1109/CBMS52027.2021.00075}}

Abstract

This paper presents an approach for enabling emotive expression classification through speech analysis combining affective prosody (Mel-frequency Cepstral Coefficient, Zero Crossing Rate, Chroma Energy Normalised) and semantic analysis (Bag-of-words model). Two machine learning (ML) classifiers, a convolutional neural network (CNN) and logistic regression (LR) model, are combined to form an ensemble based approach for the classification of emotive expressions from multi-modal data (audio, text). The approach builds upon existing work in emotion classification, sentiment analysis, and natural language processing (NLP) techniques. The paper contributes a workflow for the comparative analysis of the efficacy in classifying emotion from multi-modal data. The results demonstrate mixed accuracy across varied data sources, indicating the limitations and considerations of a generalised approach. The approach utilised has direct benefits for Affective computing research as it enables (a) insight into the strengths and limitations of such models in correctly classifying emotion in relation to population differences (e.g. gender) and provides (b) a baseline for emotion classification in speech across the six canonical basic emotions (Anger, Fear, Disgust, Joy, Sadness, Surprise).

Google Scholar ResearchGate

BibTeX

@INPROCEEDINGS{9467869, author={Dair, Zachary and Donovan, Ryan and O’Reilly, Ruairi}, booktitle={2021 32nd Irish Signals and Systems Conference (ISSC)}, title={Classification of Emotive Expression Using Verbal and Non Verbal Components of Speech}, year={2021}, volume={}, number={}, pages={1-8}, doi={10.1109/ISSC52156.2021.9467869}}

Abstract

The Personality Emotion Mapping (PEM) model is a workflow for generating quantifiable and bi-directional mappings between 15 personality traits and the basic emotions. PEM utilises Affective computing methodology to map this relationship across the modalities of self-report, facial expressions, semantic analysis, and affective prosody. The workflow is an end-to-end solution integrating data collection, feature extraction, data analysis, and result generation. PEM results in a real-time model that provides a high-resolution correlated mapping between personality traits and the basic emotions. The robustness of PEM’s model is supported by the workflow’s ability to conduct meta-analytical and multimodal analysis; each state-to-trait mapping is dynamically updated in terms of its magnitude, direction, and statistical significance as data is processed. PEM provides a methodology that can contribute to long-standing research questions in the fields of Psychology and Affective computing. These research questions include (i) quantifying the emotive nature of personality, (ii) minimising the effects of context variance in basic emotions research, and (iii) investigating the role of emotion sequencing effects in relation to individual differences. PEM’s methodology enables direct applications in any domain that requires the provision of individualised and personalised services (e.g. advertising, clinical care, research).

Google Scholar ResearchGate

BibTeX

@inproceedings{donovan2021multimodal, title={A Multimodal Workflow for Modeling Personality and Emotions to Enable User Profiling and Personalisation.},  author={Donovan, Ryan and Johnson, Aoife},  year={2021}}

Abstract

This article describes an exploratory study that aimed to analyse the relationship between personality traits and emotions. In particular, it investigates to what extent the sub-traits of the Five Factor Model has an empirically quantifiable correlation with the Basic Emotions (Anger, Anxiety, Disgust, Fear, Joy, Sadness, Surprise). If links between these personality traits and the basic emotions can be found, then this would enable an emotional-state-to-personality-trait mapping. In this study, 38 participants answered a Big Five Aspects Scale (BFAS) questionnaire and then watched 12 emotionally provocative film clips along with answering 12 short emotional Likert-scales on their emotional experiences during each film clip. The results showed that (i) four of the seven Basic Emotions outright significantly correlated, while two emotions (Fear and Disgust) approached statistical significance,  with at least one of the personality traits and (ii) significant correlations between personality traits and basic emotions could only be identified at the sub-trait level, demonstrating the value in adopting a higher-resolution personality model. The results supports the long-term goal of this research, which is the enabling of state-to-trait inferences. A method for analysing and visualising such a mapping, that differentiates mappings based on the direction and magnitude of the effect size was also developed. The study contributes a blueprint towards utilising Affective Computing methodology to automatically map these phenomena.

Paper

Abstract

In the healthcare domain, Artificial Intelligence (AI) based systems are being increasingly adopted with applications ranging from surgical robots to automated medical diagnostics. While a Machine Learning (ML) engineer might be interested in the parameters related to the performance and accuracy of these AI-based systems, it is postulated that a medical practitioner would be more concerned with the applicability, and utility of these systems in the medical setting. However, medical practitioners are unlikely to have the prerequisite skills to enable reasonable interpretation of an AI-based system. This is a concern for two reasons. Firstly, it inhibits the adoption of systems capable of automating routine analysis work and prevents the associated productivity gains. Secondly, and perhaps more importantly, it reduces the scope of expertise available to assist in the validation, iteration, and improvement of AI-based systems in providing healthcare solutions. Explainable Artificial Intelligence (XAI) is a domain focused on techniques and approaches that facilitate the understanding and interpretation of the operation of ML models. Research interest in the domain of XAI is becoming more widespread due to the increasing adoption of AI-based solutions and the associated regulatory requirements . Providing an understanding of ML models is typically approached from a Computer Science (CS) perspective with a limited research emphasis being placed on supporting alternate domains.In this paper, a simple, yet powerful solution for increasing the explainability of AI-based solutions to individuals from non-CS domains (such as medical practitioners), is presented. The proposed solution enables the explainability of ML models and the underlying workflows to be readily integrated into a standard ML workflow.Central to this solution are feature importance techniques that measure the impact of individual features on the outcomes of AI-based systems. It is envisaged that feature importance can enable a high-level understanding of a ML model and the workflow used to train the model. This could aid medical practitioners in comprehending AI-based systems and enhance their understanding of ML models’ applicability and utility.

Paper Google Scholar ResearchGate

BibTeX

@inproceedings{pawar2020incorporating, title={Incorporating Explainable Artificial Intelligence (XAI) to aid the Understanding of Machine Learning in the Healthcare Domain.}, author={Pawar, Urja and Rea, Susan}}

Abstract

Academic performance across Computer Science (CS) courses in the Republic of Ireland is underwhelming. CS undergraduates are statistically the most likely cohort in the country not to progress past year one of their studies. Insufficient motivation to pursue CS studies has been demonstrated to be a significant cause of poor CS academic performance. Goal-setting programs are an efficient, cost-effective, and student empowering way to boost motivation. Goal-setting is the formulation of a set of activities intended to motivate an individual to the desired goal state. This paper provides an experimental design for assessing the effectiveness of a written goal-setting program on academic performance concerning individual differences. Participants are randomly assigned either to the written goal-setting program or an active control task via an online platform. The goal-setting program requires participants to articulate both a desired future life and a feared future life. The program also requires participants to: identify goals and sub-goals across several domains (e.g. family, health, study); the benefits that achieving their goals would have for themselves for their connected group; the daily habits they could develop to make their ideal future more likely to occur. This study also investigates how personality (via a high-resolution personality model) and cognitive differences influence goal-setting effectiveness. Differences in both Semester 1 performance and the number of students who progress to Semester 2 are assessed between experimental groups. An ANCOVA analysis will assess whether the effectiveness of an experimental task varied based on individual differences in personality and/or cognitive ability.

Paper ResearchGate

Abstract

Artificial Intelligence (AI) is an enabling technology that when integrated into healthcare applications and smart wearable devices such as Fitbits etc. can predict the occurrence of health conditions in users by capturing and analysing their health data. The integration of AI and smart wearable devices has a range of potential applications in the area of smart healthcare but there is a challenge in the black box operation of decisions made by AI models which have resulted in a lack of accountability and trust in the decisions made. Explainable AI (XAI) is a domain in which techniques are developed to explain predictions made by AI systems. In this paper, XAI is discussed as a technique that can used in the analysis and diagnosis of health data by AI-based systems and a proposed approach presented with the aim of achieving accountability. transparency, result tracing, and model improvement in the domain of healthcare.

Paper ResearchGate

Abstract

How does the relationship between personality traits and the basic emotions vary across the modalities of self-report and facial expression analysis? This article presents the results of an exploratory study that quantifies consistencies and differences in personality-emotion mappings across these two modalities. Twenty-four participants answered a personality questionnaire before watching twelve emotionally provocative videos. Participant’s self-reported their emotional reactions per video, while their facial expressions were being recorded for automated emotional analysis. The results indicated that overall there was greater consistency than differences in personality-emotion mappings across the two modalities. The robustness of this relationship enables direct applications of emotional-state-to-personality-trait in academic and industrial domains.

Paper Google Scholar ResearchGate

BibTeX

@inproceedings{donovan2020differentiation,  title={Differentiation in Personality Emotion Mappings From Self Reported Emotion and Automatically Classified Emotion.}, author={Donovan, Ryan and Johnson, Aoife}}

Abstract

The rapid emergence and spread of COVID-19 resulted in a surge in demand for laboratory-based testing globally. Currently, the gold standard diagnostic approach is large-scale molecular testing of biological samples which detect the SARS-CoV-2 virus RNA. Infrastructure limitations and supply shortages are limiting testing capacity with a growing demand for COVID-19 diagnostics across the EU. X-ray imagery is essential in establishing the severity of a multitude of diseases and monitoring responses of patients in the hospital setting. X-ray imagery should not be used to screen for or as a first-line test to detect COVID-19. However, X-ray presents an ideal opportunity to integrate additional screening measures into a pre-existing workflow. This paper investigates the utilisation of machine learning in automating the detection of COVID-19 induced pneumonia from X-Ray imagery. The approach will assist radiologists in the monitoring and differentiation of pneumonia caused by COVID-19 from other viral causes. A classification for the presence, or absence, of pneumonia caused by COVID-19 and other viral causes is derived. The paper contributes an initial investigation into an ensemble-learning based approach using transfer learning models VGG16, Inception and ResNet. The results of this work indicate an improved performance using ensemble-based learning as compared to an individual transfer learning model.

Abstract

Despite state of the art performance on object recognition and image classification problems, CNNs are considered to have two significant weaknesses. Firstly, their inability to cater for changes in object orientation, position or lighting. Secondly, their inability to deal with part-whole relationships between objects. Capsule Networks are an enhancement to CNNs to more closely model the viewpoint invariance capability of human vision. The application of Capsule Networks to well known datasets, such as MNIST and NORB, has achieved state of the art performance, while application to other datasets has had mixed results. The application of Capsule Networks to domains such as medical based imaging problems is of significant interest as they have been shown to train accurately on some datasets with limited training data. The contribution of this research is to compare the performance of a Capsule Network to a highly accurate CNN specifically developed for classification of malaria infected and uninfected cell images. It looks at how the accuracy of each model is affected by the volume of available training data, and at how robust each model is to classifying test images subjected to transformation such as rotation, shear and lighting change.

Abstract

The analysis of physiological data plays a significant role in medical diagnostics. While state-of-the-art machine learning models demonstrate high levels of performance in classifying physiological data clinicians are slow to adopt them. A contributing factor to the slow rate of adoption is the “black-box” nature of the underlying model whereby the clinician is presented with a prediction result, but the rationale for that result omitted or not presented in an interpretable manner. This gives rise to the need for interpretable machine learning models such that clinicians can verify, and rationalise, the predictions made by a model. If a clinician understands why a model makes a prediction, they will be more inclined to accept a models assistance in analysing physiological data. This paper discusses some of the latest findings in interpretable machine learning. Thereafter, based on these findings, three models are selected and implemented to analyse ECG data that are both accurate and exhibit a high level of interpretability.

Paper ResearchGate

Abstract

The quantity and quality of data available to an organisation plays an increasingly important role in its operation. This data can relate to a variety of subjects, from the internal logistics to consumer sentiment towards a product in a specific market. This data provides increasingly optimal behaviour derived from its analysis, e.g. improved decision making. As more advanced, application-specific, machine learning models are developed, the organisation with the largest share of data will gain an advantage over its competitors. It is postulated that smaller entities with minority shares in data within a domain possess only a fragmented view of a market. This fragmented view puts the smaller entity at a disadvantage and enables larger entities to reap unfair, competitive advantages. This unequal dynamic should be rectified. To that end, in this work, a model is proposed that enables the consensual sharing of data between multiple parties using blockchain.

Paper ResearchGate

Abstract

Electrocardiography (ECG) is a form of physiological data used to record the electrical activity of the heart. Numerous researchers have proposed and developed methods to extract features from the ECG signal (for example, R-R segment, P-R segment). These features can be used to analyse and classify various forms of heart arrhythmia. In this work, a method for ECG classification that employs a generalised signal pre-processing technique and uses a Multi-Layer Perceptron network to classify arrhythmia per the AAMI EC57 standard accurately is presented. The method is trained and evaluated using PhysioNet’s MIT-BIH dataset, and an average accuracy of 98.72% is achieved. The proposed methodology is comparable to state-of-the-art CNN models, both in terms of accuracy and efficiency.

Paper ResearchGate

Abstract

Cervical cancer is a severe concern for women’s health. Every year in the Republic of Ireland, approximately 300 women are diagnosed with cervical cancer, 30% for whom the diagnosis will prove fatal. It is the second most common cause of death due to cancer in women aged 25 to 39 years. Recently there has been a series of controversies concerning the mishandling of results from cervical screening tests, delays in processing said tests and the recalling of individuals to retake tests. The serious nature of the prognosis highlights the importance and need for the timely processing and analysis of data related to screenings. This work presents a comparative analysis of several classification techniques used for the automated analysis of known risk factors and screening tests with the aim of predicting cervical cancer outcomes via a Biopsy result. These techniques encompass methods such as tree-based, cluster-based, liner and ensemble techniques, and where applicable use parameter tuning to determine optimal model parameters. The dataset utilised for training and validation consists of 858 observations and 36 variables, including the binary target variable “Biopsy”. The data itself is heavily imbalanced with 803 negative and 55 positive observations with approximately 11.73% of the data points missing. These issues are addressed during pre-processing by methods such as mean or median imputation, as well as over-sampling, under-sampling and combination techniques which led to the creation of 6 augmented datasets of varying size, consisting of 34 variables including the response Biopsy. The results show that a SMOTE-Tomek combination resampling method in conjunction with a tuned Random Forest model produced an accuracy score of 99.69% with a recall and precision value of 0.99% for both positive and negative responses.

Paper ResearchGate

Abstract

Beats-Per-Minute (BPM) is a microservice-based platform that provides a monitoring solution for the continuous acquisition, analysis and visualisation of health related data. BPM combines Commercial Off-The-Self (COTS) Activity Trackers and a scalable cloud-based infrastructure. This paper demonstrates the efficacy, reliability and integrity of BPM when utilised as a monitoring solution for health conditions, such as Cardiovascular Disease. The results are indicative of the suitability of a microservice-based architecture for such a platform.

Paper ResearchGate

Abstract

This research paper investigates the running of object detection algorithms on low-end devices to detect individuals in images while leveraging cloud-based services to provide facial verification of the individuals detected. The performance of three computer vision object detection algorithms that utilize Convolutional Neural Networks (CNN) are compared: SSD MobileNet, Inception v2 and Tiny YOLO along with three cloud-based facial verification services: Kairos, Amazon Web Service Rekognition (AWS) and Microsoft Azure Vision API. The results contribute to the limitations of running CNN based algorithms to solve vision tasks on low-end devices and highlights the limitations of using such devices and models in an application domain such as a home-security solution.

Paper ResearchGate

Abstract

Seizures are both the most common neurological emergency afflicting neonates and the most difficult to detect clinically. Currently, the monitoring of a multi-channel electroencephalogram (EEG) is the gold standard for seizure detection. The accurate analysis of this physiological data requires a neurophysiologist with expertise in neonatal EEG. The provisioning of this expertise on a continuous basis can be challenging for medical facilities. In this paper, we describe a cloud-based platform capable of supporting clinicians through the creation of expert knowledge repositories. While the platform is considered general purpose, in this work it is applied specifically to neonatal EEG.

Paper ResearchGate

Abstract

A system, constructed as a “proof of concept”, for providing an Irish tele-neurophysiology service is presented. It is based on a distributed architecture and capable of handling synchronous data streams from multiple Irish Neonatal Intensive Care Units. It provides Ireland with an infrastructure for overcoming factors affecting the diagnosis of neurological disorders in neonates. The system supports collaborative efforts by neurophysiologists, removes geographical constraints on expert knowledge and allows for the creation of national data stores while simultaneously supporting the transAtlantic processing of EEG. Technical obstacles affecting its successful implementation are outlined and solutions proposed. The implementation of such a system could significantly improve the quality of care provided to neonates.

Paper ResearchGate

Abstract

The processing of a patient in a medical facility encompasses data acquisition, data analysis, diagnosis and medical reporting. The result is a program of treatment for the patient. It is postulated that the latter two stages can be facilitated by allowing acquisition locations to avail of data interpretation by multiple off-site analysts who interactively annotate patient data in a collaborative fashion. The concept of “meta annotations” is also introduced, whereby users can comment on an annotation or supply a confidence rating that indicates their assessment of the accuracy of an annotation. This concept enhances the utility of annotations by transforming them from simple textual labels to focal points for discussion and consensus building.

Paper ResearchGate

Abstract

In a critical care setting, delays in the diagnosis of neurological conditions can have a significant impact on patient outcomes. Recording the electrical activity of the brain (EEG) is often used to diagnose and monitor neurological conditions. A trained neurophysiologist is then required to analyse these signals. However, in many cases this expertise is not available on-site. A web-based remote monitoring system for EEG data is presented that reduces the delays that can arise when off-site expert analysis is required. The system allows EEG data to be streamed to a remote location in near-real-time and viewed while acquisition is ongoing. Data streamed earlier may also be reviewed, providing the user with a continually updating view of the entire EEG recording. All communications are performed using web technologies in order to minimise issues with firewalls and to enable analysis on most web-connected PCs. The only tool required for viewing EEG data is a modern web browser with the commonly-available Adobe Flash plugin installed. Since these are ubiquitous, data analysis is not limited by geographic location. Moreover, multiple users can independently view the data simultaneously.

Paper ResearchGate

Abstract

An agent framework that assists with the interpretation of streaming physiological data is presented. The framework operates as a module within an existing remote monitoring system for streaming physiological data. Agents complement the remote monitoring system by enhancing the users’ view of the signals through the addition of annotations and the creation of derived views of the signals.

Paper ResearchGate

Abstract

In healthcare, detecting stress and enabling individuals to monitor their mental health and well-being is challenging. Advancements in wearable technology now enable continuous physiological data collection. This data can provide insights into mental health and behavioural states through psychophysiological analysis. However, automated analysis is required to provide timely results due to the quantity of data collected. Machine learning has shown efficacy in providing an automated classification of physiological data for health applications in controlled laboratory environments. Ambulatory uncontrolled environments, however, provide additional challenges requiring further modelling to overcome. This work empirically assesses several approaches utilising machine learning classifiers to detect stress using physiological data recorded in an ambulatory setting with self-reported stress annotations. A subset of the training portion of the SMILE dataset enables the evaluation of approaches before submission. The optimal stress detection approach achieves 90.77% classification accuracy, 91.24 F1-Score, 90.42 Sensitivity and 91.08 Specificity, utilising an ExtraTrees classifier and feature imputation methods. Meanwhile, accuracy on the challenge data is much lower at 59.23% (submission #54 from BEaTS-MTU, username ZacDair). The cause of the performance disparity is explored in this work.

Paper Google Scholar

BibTeX

@incollection{dair2024classification, title={Classification of Stress via Ambulatory ECG and GSR Data}, author={Dair, Zachary and Saad, Muhammad and Pawar, Urja and O'Reilly, Ruairi and Dockray, Samantha}, booktitle={Human Activity and Behavior Analysis}, pages={183--196}, year={2024}, publisher={CRC Press}}

Abstract

Publicly available diabetic retinopathy (DR) datasets are imbalanced, containing limited numbers of images with DR. This imbalance contributes to overfitting when training machine learning classifiers. The impact of this imbalance is exacerbated as the severity of the DR stage increases, affecting the classifiers’ diagnostic capacity. The imbalance can be addressed using Generative Adversarial Networks (GANs) to augment the datasets with synthetic images. Generating synthetic images is advantageous if high-quality and diverse images are produced. To evaluate the quality and diversity of synthetic images, several evaluation metrics, such as Multi-Scale Structural Similarity Index (MS-SSIM), Cosine Distance (CD), and Fréchet Inception Distance (FID), are used. Understanding the effectiveness of each metric in evaluating the quality and diversity of synthetic images is critical to select images for augmentation. To date, there has been limited analysis of the appropriateness of these metrics in the context of biomedical imagery. This work contributes an empirical assessment of these evaluation metrics as applied to synthetic Proliferative DR imagery generated by a Deep Convolutional GAN (DCGAN). Furthermore, the metrics’ capacity to indicate the quality and diversity of synthetic images and their correlation with classifier performance are examined. This enables a quantitative selection of synthetic imagery and an informed augmentation strategy, which are often lacking in the literature. Results indicate that FID is suitable for evaluating the quality, while MS-SSIM and CD are suitable for evaluating the diversity of synthetic imagery. Furthermore, the superior performance of Convolutional Neural Network (CNN) and EfficientNet classifiers, as indicated by the and AUC scores, for the augmented datasets compared to the original dataset demonstrate the efficacy of synthetic imagery to augment the imbalanced dataset while improving the classification scores

Paper Google Scholar

BibTeX

 @article{dragan2022evaluating, title={Evaluating the Quality and Diversity of DCGAN-based Generatively Synthesized Diabetic Retinopathy Imagery}, author={Dragan, Cristina-Madalina and Saad, Muhammad Muneeb and Rehmani, Mubashir Husain and O'Reilly, Ruairi}, journal={arXiv:2208.05593v3}, year={2022}}

Abstract

Affective computing research aims to accurately detect users’ emotional states to enhance human-computer interaction. While existing studies demonstrate the ability of affective computing models to accurately detect posed expressions of emotions, there is limited investigation into their validity in detecting natural or spontaneous emotional expressions. This study examines the validity of several models in accurately detecting six basic emotional states from spontaneous expressions. Data were collected from 114 participants who viewed emotionally provocative video clips, while multi-modal information on their emotional experiences was captured (questionnaires, video recordings, audio recordings, and text). Open-source software was used to detect basic emotional states from facial expressions (Py-Feat), text (National Research Lexicon), and audio (an LTSM-DNN model). The construct validity of these measurements was assessed by correlating their relationship to self-reported emotional experiences (positive convergent validity) and Big Five personality traits (both discriminant and negative convergent validity). Data analyses demonstrated limited construct validity for automatic detection of emotion through facial expression and text, and poor construct validity through audio. These results suggest that existing models are not yet capable of accurately measuring basic emotional states in real-life settings, highlighting the critical need for improving affective computing models using spontaneous emotional data to enhance their accuracy. The Personality-Emotion Mapping (PEM) dataset is provided as a resource for training and testing these

Google Scholar

BibTeX

@article{donovan2024evaluating, title={Evaluating the Validity of Automated Emotion Detection from Spontaneous Expressions across Modalities: Insights from the PEM Dataset}, author={Donovan, Ryan and Johnson, Aoife and deRoiste, Aine and O’Reilly, Ruairi}, year={2024}, publisher={OSF}}

Abstract

Advances in wearable technology have significantly increased the sensitivity and accuracy of devices for recording physiological signals. Commercial off-the-shelf wearable devices can gather large quantities of physiological data un-obtrusively. This enables momentary assessments of human physiology, which provide valuable insights into an individual’s health and psychological state. Leveraging these insights provides significant benefits for human-to-computer interaction and personalised healthcare. This work contributes an analysis of variance occurring infeatures representative of affective states extracted from electro cardiograms and photoplethysmography; subsequently identifies the cardiac measures most descriptive of affective states from both signals and provides insights into signal and emotion-specific cardiac measures; finally baseline performance for automated affective state detection from physiological signals is established.

Google Scholar

BibTeX

@article{dair2022variance, title={Variance in Classifying Affective State via Electrocardiogram and Photoplethysmography.}, author={Dair, Zachary and Dockray, Samantha and O'Reilly, Ruairi}, journal={CoRR}, year={2022}}

Abstract

Systems for the monitoring and analysis of data, e.g. stock markets, smart grids and sensor networks, can be made up of massive numbers of individual components. These components can be geographically distributed yet may interact with one another via continuous data streams, which in turn may affect the state of the sender or receiver. This introduces a dynamic causality, which further complicates the overall system by introducing a temporal constraint that is difficult to accommodate. Practical approaches to realising the system described above have led to a multiplicity of analysis techniques, each of which concentrates on specific characteristics of the system being analysed and treats these characteristics as the dominant component affecting the results being sought. The multiplicity of analysis techniques introduces another layer of heterogeneity, that is heterogeneity of approach, partitioning the field to the extent that results from one domain are difficult to exploit in another. The question is asked can a generic solution for the monitoring and analysis of data that: accommodates temporal constraints; bridges the gap between expert knowledge and raw data; and enables data to be effectively interpreted and exploited in a transparent manner, be identified? The approach proposed in this dissertation acquires, analyses and processes data in a manner that is free of the constraints of any particular analysis technique, while at the same time facilitating these techniques where appropriate. Constraints are applied by defining a workflow based on the production, interpretation and consumption of data. This supports the application of different analysis techniques on the same raw data without the danger of incorporating hidden bias that may exist. To illustrate and to realise this approach a software platform has been created that allows for the transparent analysis of data, combining analysis techniques with a maintainable record of provenance so that independent third party analysis can be applied to verify any derived conclusions. In order to demonstrate these concepts, a complex real world example involving the near real-time capturing and analysis of neurophysiological data from a neonatal intensive care unit (NICU) was chosen. A system was engineered to gather raw data, analyse that data using different analysis techniques, uncover information, incorporate that information into the system and curate the evolution of the discovered knowledge. The application domain was chosen for three reasons: firstly because it is complex and no comprehensive solution exists; secondly, it requires tight interaction with domain experts, thus requiring the handling of subjective knowledge and inference; and thirdly, given the dearth of neurophysiologists, there is a real world need to provide a solution for this domain.

Paper ResearchGate