Pub Date : 2022-09-27DOI: 10.1109/BHI56158.2022.9926744
S. M. Deniz, Hamraz Javaheri, J. F. Vargas, Dogan Urgun, Fariza Sabit, Mahmut Tok, Mehmet Haklıdır, Bo Zhou, P. Lukowicz
In brain-computer interface and neuroscience, electroencephalography (EEG) signals have been well studied with not only cognitive activities but also physical activities. This work investigates if EEG can be used for detecting the motion as well as the variable weights a person is lifting. To this end, we used both commercial EEG headsets as well as open-source and open-protocol EEG hardware that is suitable for do-it-yourself designers. EEG data were obtained during performing biceps flexion-extension motions for different weight categories: lifting with no weight (empty), medium, and heavy lifting. Through two experiments of the bicep curl lifting scenario, we validated the concept with a study designed according to neuroscience standards and explored the pathway towards real-world applications with wearable sensing and smart garments. Both feature-based classification methods and deep learning models were designed and evaluated, showing accuracy up to 78% of differentiating three levels of weight (empty, medium, and heavy) consistently outperforming similar the state of the art. Our approach to predict different categories of lifted weight could be used in further optimizations in different research areas such as rehabilitation, sport as well as industrial applications. To encourage further research in this direction, the data sets acquired during this study will be publicly available.
{"title":"Prediction of Lifted Weight Category Using EEG Equipped Headgear","authors":"S. M. Deniz, Hamraz Javaheri, J. F. Vargas, Dogan Urgun, Fariza Sabit, Mahmut Tok, Mehmet Haklıdır, Bo Zhou, P. Lukowicz","doi":"10.1109/BHI56158.2022.9926744","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926744","url":null,"abstract":"In brain-computer interface and neuroscience, electroencephalography (EEG) signals have been well studied with not only cognitive activities but also physical activities. This work investigates if EEG can be used for detecting the motion as well as the variable weights a person is lifting. To this end, we used both commercial EEG headsets as well as open-source and open-protocol EEG hardware that is suitable for do-it-yourself designers. EEG data were obtained during performing biceps flexion-extension motions for different weight categories: lifting with no weight (empty), medium, and heavy lifting. Through two experiments of the bicep curl lifting scenario, we validated the concept with a study designed according to neuroscience standards and explored the pathway towards real-world applications with wearable sensing and smart garments. Both feature-based classification methods and deep learning models were designed and evaluated, showing accuracy up to 78% of differentiating three levels of weight (empty, medium, and heavy) consistently outperforming similar the state of the art. Our approach to predict different categories of lifted weight could be used in further optimizations in different research areas such as rehabilitation, sport as well as industrial applications. To encourage further research in this direction, the data sets acquired during this study will be publicly available.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116873477","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-09-27DOI: 10.1109/BHI56158.2022.9926957
V. Pezoulas, F. Kalatzis, T. Exarchos, Antreas Goules, A. Tzioufas, D. Fotiadis
Nowadays there is an intensive need to move towards a universal health data ecosystem by breaking down data silos. Faced with a wealth of dispersed health data, there are still critical open issues and unmet needs to make this feasible, varying from secure data sharing to data quality and heterogeneity. Considering these challenges, we propose a novel federated platform to unlock the full potential of data from health data intermediaries through the secure sharing, curation, and Natural Language Processing (NLP)-based harmonization of dispersed and complex clinical data structures. The platform was deployed to establish a first Pan-European data hub on rare autoimmune and chronic diseases with 7551 harmonized patient records across 21 European countries with a 90% terminology overlap. An advanced data driven imputer was built to predict missing records in the real patient data based on high-quality synthetic data profiles (with Kullback-Leibler divergence less than 0.01). with reduced fault detection rate (less than 2%) compared to conventional imputers, such as, the kNN imputer. Customized and explainable federated AI algorithms were trained on top of the established data hub for lymphomagenesis modeling with 0.87 sensitivity and 0.74 specificity along with a set of validated biomarkers for disease onset and progression.
{"title":"A federated AI-empowered platform for disease management across a Pan-European data driven hub","authors":"V. Pezoulas, F. Kalatzis, T. Exarchos, Antreas Goules, A. Tzioufas, D. Fotiadis","doi":"10.1109/BHI56158.2022.9926957","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926957","url":null,"abstract":"Nowadays there is an intensive need to move towards a universal health data ecosystem by breaking down data silos. Faced with a wealth of dispersed health data, there are still critical open issues and unmet needs to make this feasible, varying from secure data sharing to data quality and heterogeneity. Considering these challenges, we propose a novel federated platform to unlock the full potential of data from health data intermediaries through the secure sharing, curation, and Natural Language Processing (NLP)-based harmonization of dispersed and complex clinical data structures. The platform was deployed to establish a first Pan-European data hub on rare autoimmune and chronic diseases with 7551 harmonized patient records across 21 European countries with a 90% terminology overlap. An advanced data driven imputer was built to predict missing records in the real patient data based on high-quality synthetic data profiles (with Kullback-Leibler divergence less than 0.01). with reduced fault detection rate (less than 2%) compared to conventional imputers, such as, the kNN imputer. Customized and explainable federated AI algorithms were trained on top of the established data hub for lymphomagenesis modeling with 0.87 sensitivity and 0.74 specificity along with a set of validated biomarkers for disease onset and progression.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129353980","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-09-27DOI: 10.1109/BHI56158.2022.9926824
Michail Sarafidis, G. Lambrou, G. Matsopoulos, D. Koutsouris
Bladder cancer (BCa) is one of the most prevalent cancers worldwide and accounts for high socioeconomic impact. BCa can manifest in the form of nonaggressive and usually non-muscle invasive (NMIBC) tumors that recur and require chronic invasive surveillance, or aggressive and muscle invasive (MIBC) tumors with high associated mortality. These two subtypes exhibit distinct prognosis and require different therapeutic approaches. In the present study, we conducted an integrative bioinformatics analysis, combining transcriptomic data from various microarray experiments, in order to reveal a common signature of differentially expressed genes (DEGs) between the two subtypes. Subsequently, we constructed the protein-protein interaction (PPI) network of the DEGs and defined the hub genes based on 11 topological analysis methods. Then, the most significant hub genes were identified using LASSO logistic regression algorithm. The selected genes were finally used as features in supervised classification algorithms, namely support vector machines and random forests, for BCa subtype discrimination. The models' evaluation showed area under the curve (AUC) values up to 96% as regards separating NMIBC from MIBC tumors. Genes driving the separation between tumor subtypes may prove to be important biomarkers for BCa development and progression, and eventually candidates for therapeutic targeting.
{"title":"Integrative Bioinformatics Analysis of Transcriptomic Data Reveals Hub Genes as Diagnostic Biomarkers for Non-Muscle vs. Muscle Invasive Bladder Cancer","authors":"Michail Sarafidis, G. Lambrou, G. Matsopoulos, D. Koutsouris","doi":"10.1109/BHI56158.2022.9926824","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926824","url":null,"abstract":"Bladder cancer (BCa) is one of the most prevalent cancers worldwide and accounts for high socioeconomic impact. BCa can manifest in the form of nonaggressive and usually non-muscle invasive (NMIBC) tumors that recur and require chronic invasive surveillance, or aggressive and muscle invasive (MIBC) tumors with high associated mortality. These two subtypes exhibit distinct prognosis and require different therapeutic approaches. In the present study, we conducted an integrative bioinformatics analysis, combining transcriptomic data from various microarray experiments, in order to reveal a common signature of differentially expressed genes (DEGs) between the two subtypes. Subsequently, we constructed the protein-protein interaction (PPI) network of the DEGs and defined the hub genes based on 11 topological analysis methods. Then, the most significant hub genes were identified using LASSO logistic regression algorithm. The selected genes were finally used as features in supervised classification algorithms, namely support vector machines and random forests, for BCa subtype discrimination. The models' evaluation showed area under the curve (AUC) values up to 96% as regards separating NMIBC from MIBC tumors. Genes driving the separation between tumor subtypes may prove to be important biomarkers for BCa development and progression, and eventually candidates for therapeutic targeting.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"401 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131916932","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-09-27DOI: 10.1109/BHI56158.2022.9926945
J. López-Correa, Caroline König, A. Vellido
G protein-coupled receptors are a large super-family of cell membrane proteins that play an important physiological role as transmitters of extra-cellular signals. Signal transmission through the cell membrane depends on the conformational changes of the transmembrane region of the receptor and the investigation of the dynamics in these regions is therefore key. Molecular Dynamics (MD) simulations can provide information of the receptor conformational states at the atom level and machine learning (ML) methods can be useful for the analysis of these data. In this paper, Recurrent Neural Networks (RNNs) are used to evaluate whether the MD can be modeled focusing on the different regions of the receptor (intra-cellular, extra-cellular and each transmembrane regions (TM)). The best results, as measured by root-mean-square deviation (RMSD), are 0.1228 Å for TM4 of the 2rh1 (inactive state) and 0.1325 Å for TM4 of the 3p0g (active state), which are comparable to the state-of-the-art in non-dynamic 3-D predictions, showing the potential of the proposed approach.
{"title":"Molecular Dynamics forecasting of transmembrane Regions in GPRCs by Recurrent Neural Networks","authors":"J. López-Correa, Caroline König, A. Vellido","doi":"10.1109/BHI56158.2022.9926945","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926945","url":null,"abstract":"G protein-coupled receptors are a large super-family of cell membrane proteins that play an important physiological role as transmitters of extra-cellular signals. Signal transmission through the cell membrane depends on the conformational changes of the transmembrane region of the receptor and the investigation of the dynamics in these regions is therefore key. Molecular Dynamics (MD) simulations can provide information of the receptor conformational states at the atom level and machine learning (ML) methods can be useful for the analysis of these data. In this paper, Recurrent Neural Networks (RNNs) are used to evaluate whether the MD can be modeled focusing on the different regions of the receptor (intra-cellular, extra-cellular and each transmembrane regions (TM)). The best results, as measured by root-mean-square deviation (RMSD), are 0.1228 Å for TM4 of the 2rh1 (inactive state) and 0.1325 Å for TM4 of the 3p0g (active state), which are comparable to the state-of-the-art in non-dynamic 3-D predictions, showing the potential of the proposed approach.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123672065","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-09-27DOI: 10.1109/BHI56158.2022.9926929
Dimitris Zaridis, E. Mylona, N. Tachos, K. Marias, M. Tsiknakis, D. Fotiadis
Precise delineation of the prostate gland on MRI is the cornerstone for accurate prostate cancer diagnosis, detection, characterization and treatment. The present work proposes a meta-learner deep learning (DL) network that combines the complexity of 3 well-established DL models and fine tune them in order to improve the segmentation of the prostate compared to the base learners. The backbone of the meta-learner consist the original U-net, Dense2U-net and Bridged U-net models. A model was added on top of the three base networks that has four convolutions with different receptor fields. The meta-learner outperformed the base-learners in 4 out of 5 performance metrics. The median Dice Score for the meta-learner was 89% while for the second best model it was 83%. Except for Hausdorff distance, where the meta-learner and Dense2U-net performed equally well, the improvement achieved in terms of average sensitivity, balanced accuracy, dice score and rand error, compared to the best performing base-learner, was 6%, 3%, 5% and 4%, respectively.
{"title":"Fine-tuned feature selection to improve prostate segmentation via a fully connected meta-learner architecture","authors":"Dimitris Zaridis, E. Mylona, N. Tachos, K. Marias, M. Tsiknakis, D. Fotiadis","doi":"10.1109/BHI56158.2022.9926929","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926929","url":null,"abstract":"Precise delineation of the prostate gland on MRI is the cornerstone for accurate prostate cancer diagnosis, detection, characterization and treatment. The present work proposes a meta-learner deep learning (DL) network that combines the complexity of 3 well-established DL models and fine tune them in order to improve the segmentation of the prostate compared to the base learners. The backbone of the meta-learner consist the original U-net, Dense2U-net and Bridged U-net models. A model was added on top of the three base networks that has four convolutions with different receptor fields. The meta-learner outperformed the base-learners in 4 out of 5 performance metrics. The median Dice Score for the meta-learner was 89% while for the second best model it was 83%. Except for Hausdorff distance, where the meta-learner and Dense2U-net performed equally well, the improvement achieved in terms of average sensitivity, balanced accuracy, dice score and rand error, compared to the best performing base-learner, was 6%, 3%, 5% and 4%, respectively.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123690063","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-09-27DOI: 10.1109/BHI56158.2022.9926777
Elvin Johnson, Shreshta Mohan, Alex Gaudio, A. Smailagic, C. Faloutsos, A. Campilho
Advances in data-driven deep learning for chest X-ray image analysis underscore the need for explainability, privacy, large datasets and significant computational resources. We frame privacy and explainability as a lossy single-image compression problem to reduce both computational and data requirements without training. For Cardiomegaly detection in chest X-ray images, we propose HeartSpot and four spatial bias priors. HeartSpot priors define how to sample pixels based on domain knowledge from medical literature and from machines. HeartSpot privatizes chest X-ray images by discarding up to 97% of pixels, such as those that reveal the shape of the thoracic cage, bones, small lesions and other sensitive features. HeartSpot priors are ante-hoc explainable and give a human-interpretable image of the preserved spatial features that clearly outlines the heart. HeartSpot offers strong compression, with up to 32x fewer pixels and 11 $x$ smaller filesize. Cardiomegaly detectors using HeartSpot are up to 9x faster to train or at least as accurate (up to +.01 AUC ROC) when compared to a baseline DenseNet121. HeartSpot is post-hoc explainable by re-using existing attribution methods without requiring access to the original non-privatized image. In summary, HeartSpot improves speed and accuracy, reduces image size, improves privacy and ensures explainability.
{"title":"HeartSpot: Privatized and Explainable Data Compression for Cardiomegaly Detection","authors":"Elvin Johnson, Shreshta Mohan, Alex Gaudio, A. Smailagic, C. Faloutsos, A. Campilho","doi":"10.1109/BHI56158.2022.9926777","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926777","url":null,"abstract":"Advances in data-driven deep learning for chest X-ray image analysis underscore the need for explainability, privacy, large datasets and significant computational resources. We frame privacy and explainability as a lossy single-image compression problem to reduce both computational and data requirements without training. For Cardiomegaly detection in chest X-ray images, we propose HeartSpot and four spatial bias priors. HeartSpot priors define how to sample pixels based on domain knowledge from medical literature and from machines. HeartSpot privatizes chest X-ray images by discarding up to 97% of pixels, such as those that reveal the shape of the thoracic cage, bones, small lesions and other sensitive features. HeartSpot priors are ante-hoc explainable and give a human-interpretable image of the preserved spatial features that clearly outlines the heart. HeartSpot offers strong compression, with up to 32x fewer pixels and 11 $x$ smaller filesize. Cardiomegaly detectors using HeartSpot are up to 9x faster to train or at least as accurate (up to +.01 AUC ROC) when compared to a baseline DenseNet121. HeartSpot is post-hoc explainable by re-using existing attribution methods without requiring access to the original non-privatized image. In summary, HeartSpot improves speed and accuracy, reduces image size, improves privacy and ensures explainability.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124874105","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-09-27DOI: 10.1109/BHI56158.2022.9926944
Nagaraj Hegde, T. Swibas, E. Melanson, E. Sazonov
In this work we developed and validated a method to capture the activities of daily living (ADL), transitions between ADL, and the associated Energy Expenditure (EE) using a novel insole based wearable system (SmartStep). A 15-participant study was conducted in a controlled laboratory environment while participants wore the SmartStep and performed various ADL. Machine learning models were developed using 4-branched and 8-branched steady-state activities to estimate the total energy expenditure (TEE) and physical activity energy expenditure (PAEE). Additional models accounting for transitions between activities were also developed. These models were validated in an independent study with 8-participants, performed in a whole room indirect calorimeter. In the controlled study, the 8-branched models had a lower root mean square error (RMSE, 0.58 vs. 0.67 kcal/min) and lower total error (−1.5% vs. 3%). In the validation study, the 8-branched models also had a lower RMSE (0.9 kcal/min vs. 1.2 kcal/min) and lower total error (−4.5% vs 11%). Accounting for activity transitions reduced the total error in the EE estimation to −1.3%. The results suggested that SmartStep can be used to accurately monitor the EE of the wearers in their daily living. The validation study results suggested that 8-branched models more accurately predict EE than 4-branched models and that accounting for activity transitions improves the estimation of EE in daily living.
在这项工作中，我们开发并验证了一种方法来捕捉日常生活活动(ADL)， ADL之间的转换，以及相关的能量消耗(EE)使用一种新型的鞋垫可穿戴系统(SmartStep)。一项15名参与者的研究在受控的实验室环境中进行，参与者佩戴SmartStep并进行各种ADL。利用4支和8支稳态活动建立了机器学习模型，以估计总能量消耗(TEE)和身体活动能量消耗(PAEE)。另外还开发了考虑活动之间转换的其他模型。这些模型在一项有8名参与者的独立研究中得到了验证，该研究在整个房间的间接量热计中进行。在对照研究中，8支模型具有较低的均方根误差(RMSE, 0.58 vs. 0.67 kcal/min)和较低的总误差(- 1.5% vs. 3%)。在验证研究中，8支模型也具有较低的RMSE (0.9 kcal/min vs. 1.2 kcal/min)和较低的总误差(- 4.5% vs. 11%)。考虑活动转换将EE估计的总误差降低到- 1.3%。结果表明，SmartStep可以用来准确地监测佩戴者在日常生活中的情感表达。验证研究结果表明，8支模型比4支模型更准确地预测情感表达，并且考虑活动转换可以改善对日常生活中情感表达的估计。
{"title":"Development and Independent Validation of Energy Expenditure Models Using SmartStep","authors":"Nagaraj Hegde, T. Swibas, E. Melanson, E. Sazonov","doi":"10.1109/BHI56158.2022.9926944","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926944","url":null,"abstract":"In this work we developed and validated a method to capture the activities of daily living (ADL), transitions between ADL, and the associated Energy Expenditure (EE) using a novel insole based wearable system (SmartStep). A 15-participant study was conducted in a controlled laboratory environment while participants wore the SmartStep and performed various ADL. Machine learning models were developed using 4-branched and 8-branched steady-state activities to estimate the total energy expenditure (TEE) and physical activity energy expenditure (PAEE). Additional models accounting for transitions between activities were also developed. These models were validated in an independent study with 8-participants, performed in a whole room indirect calorimeter. In the controlled study, the 8-branched models had a lower root mean square error (RMSE, 0.58 vs. 0.67 kcal/min) and lower total error (−1.5% vs. 3%). In the validation study, the 8-branched models also had a lower RMSE (0.9 kcal/min vs. 1.2 kcal/min) and lower total error (−4.5% vs 11%). Accounting for activity transitions reduced the total error in the EE estimation to −1.3%. The results suggested that SmartStep can be used to accurately monitor the EE of the wearers in their daily living. The validation study results suggested that 8-branched models more accurately predict EE than 4-branched models and that accounting for activity transitions improves the estimation of EE in daily living.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116880592","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-09-27DOI: 10.1109/BHI56158.2022.9926950
Yilun Zhu, A. Mariakakis, E. de Lara, T. Falk
Recent work has shown the potential of using speech signals for remote detection of coronavirus disease 2019 (COVID-19). Due to the limited amount of available data, however, existing systems have been typically evaluated within the same dataset. Hence, it is not clear whether systems can be generalized to unseen speech signals and if they indeed capture COVID-19 acoustic biomarkers or only dataset-specific nuances. In this paper, we start by evaluating the robustness of systems proposed in the literature, including two based on hand-crafted features and two on deep neural network architectures. In particular, these systems are tested across two international COVID-19 detection challenge datasets (COMPARE and DICOVA2). Experiments show that the performance of the explored systems degraded to chance levels when tested on unseen data, especially those based on deep neural networks. To increase the generalizability of existing systems, we propose a new set of acoustic biomarkers based on speech modulation spectrograms. The new biomarkers, when used to train a simple linear classifier, showed substantial improvements in cross-dataset testing performance. Further interpretation of the biomarkers provides a better understanding of the acoustic properties of COVID-19 speech. The generalizability and inter-pretability of the selected biomarkers allow for the development of a more reliable and lower-cost COVID-19 detection system.
{"title":"How Generalizable and Interpretable are Speech-Based COVID-19 Detection Systems?: A Comparative Analysis and New System Proposal","authors":"Yilun Zhu, A. Mariakakis, E. de Lara, T. Falk","doi":"10.1109/BHI56158.2022.9926950","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926950","url":null,"abstract":"Recent work has shown the potential of using speech signals for remote detection of coronavirus disease 2019 (COVID-19). Due to the limited amount of available data, however, existing systems have been typically evaluated within the same dataset. Hence, it is not clear whether systems can be generalized to unseen speech signals and if they indeed capture COVID-19 acoustic biomarkers or only dataset-specific nuances. In this paper, we start by evaluating the robustness of systems proposed in the literature, including two based on hand-crafted features and two on deep neural network architectures. In particular, these systems are tested across two international COVID-19 detection challenge datasets (COMPARE and DICOVA2). Experiments show that the performance of the explored systems degraded to chance levels when tested on unseen data, especially those based on deep neural networks. To increase the generalizability of existing systems, we propose a new set of acoustic biomarkers based on speech modulation spectrograms. The new biomarkers, when used to train a simple linear classifier, showed substantial improvements in cross-dataset testing performance. Further interpretation of the biomarkers provides a better understanding of the acoustic properties of COVID-19 speech. The generalizability and inter-pretability of the selected biomarkers allow for the development of a more reliable and lower-cost COVID-19 detection system.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"72 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127080367","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-09-27DOI: 10.1109/BHI56158.2022.9926894
Siyue Song, Tianhua Chen, G. Antoniou
Currently, how to make a concrete and correct disease prediction is a popular research trend. Researchers made more efforts to develop various models to provide interpretations of medical area, however, there is still lack of human understandable explanations provided due to the non-transparency structure of some machine learning and deep learning models. According to this work, there is one combined model application we would like to adopt. After comparison experiments of classification and interpretation, it is found the combination model can address the issues from the latest interpretation models, and try to improve the trustworthiness of medical text interpretations.
{"title":"Improve the trustwortiness of medical text interpretations","authors":"Siyue Song, Tianhua Chen, G. Antoniou","doi":"10.1109/BHI56158.2022.9926894","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926894","url":null,"abstract":"Currently, how to make a concrete and correct disease prediction is a popular research trend. Researchers made more efforts to develop various models to provide interpretations of medical area, however, there is still lack of human understandable explanations provided due to the non-transparency structure of some machine learning and deep learning models. According to this work, there is one combined model application we would like to adopt. After comparison experiments of classification and interpretation, it is found the combination model can address the issues from the latest interpretation models, and try to improve the trustworthiness of medical text interpretations.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"198 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121748223","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-09-27DOI: 10.1109/BHI56158.2022.9926956
D. Petsani, E. Konstantinidis, Michalis Timoleon, Nicholaos Athanasopoulos, Georgios Nikolaos Tsakonas, S. Nifakos, Natalia Stathakarou, M. Doumas, P. Bamidis
Healthcare continuity and remote care are among the key components for tackling disease-related effects using technological solutions. People recovering from home need high-quality of care and timely monitoring, resembling hospital care. This study proposes the use of a new device for person - machine interaction for home monitoring. The system takes advantage of automatic interaction initiated by the device on detecting patients' symptoms and providing remote care in order to improve technology engagement features. The feasibility of the proposed system was tested in COVID-19 patients as a definitive case of stay-at-home care where the treatment depends on the current state of health and the severity of the symptoms. The study shows promising results in terms of usability. The vast majority of the answers are perceiving the system as useful (90.9%) and easy to use (95.5%) and the overall System Usability Score (SUS) of the system is 65.25. The system usage adherence was also promising for the quarantine period (on average 7.2 days) but dropped after that. However, the results from the clinical team interviews showed that there is a need for sufficient allocated time for clinicians to get acquainted with the system and for ED staff to explain the device to patients.
{"title":"Towards acceptable emerging technologies for homemonitoring and care: a feasibility study with COVID-19 patients","authors":"D. Petsani, E. Konstantinidis, Michalis Timoleon, Nicholaos Athanasopoulos, Georgios Nikolaos Tsakonas, S. Nifakos, Natalia Stathakarou, M. Doumas, P. Bamidis","doi":"10.1109/BHI56158.2022.9926956","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926956","url":null,"abstract":"Healthcare continuity and remote care are among the key components for tackling disease-related effects using technological solutions. People recovering from home need high-quality of care and timely monitoring, resembling hospital care. This study proposes the use of a new device for person - machine interaction for home monitoring. The system takes advantage of automatic interaction initiated by the device on detecting patients' symptoms and providing remote care in order to improve technology engagement features. The feasibility of the proposed system was tested in COVID-19 patients as a definitive case of stay-at-home care where the treatment depends on the current state of health and the severity of the symptoms. The study shows promising results in terms of usability. The vast majority of the answers are perceiving the system as useful (90.9%) and easy to use (95.5%) and the overall System Usability Score (SUS) of the system is 65.25. The system usage adherence was also promising for the quarantine period (on average 7.2 days) but dropped after that. However, the results from the clinical team interviews showed that there is a need for sufficient allocated time for clinicians to get acquainted with the system and for ED staff to explain the device to patients.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"63 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126442049","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：