Pub Date : 2022-09-27DOI: 10.1109/BHI56158.2022.9926845
C. Marrero, A. Briens, P. Fontaine, B. Rigaud, R. Crevoisier, O. Acosta
Prostate cancer has been typically treated with a total radiation dose of 74–80 Gy administered in 2 Gy fractions. However, about 20% of patients suffer biochemical recurrence. Hypofractionated treatments may have a positive effect on tumour control. Nevertheless, the choice of an optimal personalised therapy is still compromised by the limited knowledge of the response of patients to high irradiation fractions. The purposes of this work were i) to predict biochemical recurrence after standard fractionation using our previously developed mechanistic model and ii) to explore the impact of hypofractionated treatments for patients who suffered biochemical failure. A cohort of 279 patients with localised prostate adenocarcinoma was used. Analogous virtual tissues were built from pre-treatment MRIs. The prescribed standard irradiation schedules were simulated using the mechanistic model. Biochemical recurrence was predicted from the in silico number of tumour cells at the end of treatment (AUC = 0.68). Then, alternative 2.5 and 3 Gy fractionations were simulated for patients who suffered biochemical recurrence. Significantly lower numbers of tumour cells at the end of treatment were obtained after these hypofractionated schedules. Significant decreases in total doses assuring tumour control were also observed for these patients (median of -10.3 and -14.0 Gy for 2.5 and 3 Gy fractionations, respectively).
{"title":"Predicting the impact of standard and hypofractionated schedules in prostate cancer radiotherapy with a mechanistic model","authors":"C. Marrero, A. Briens, P. Fontaine, B. Rigaud, R. Crevoisier, O. Acosta","doi":"10.1109/BHI56158.2022.9926845","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926845","url":null,"abstract":"Prostate cancer has been typically treated with a total radiation dose of 74–80 Gy administered in 2 Gy fractions. However, about 20% of patients suffer biochemical recurrence. Hypofractionated treatments may have a positive effect on tumour control. Nevertheless, the choice of an optimal personalised therapy is still compromised by the limited knowledge of the response of patients to high irradiation fractions. The purposes of this work were i) to predict biochemical recurrence after standard fractionation using our previously developed mechanistic model and ii) to explore the impact of hypofractionated treatments for patients who suffered biochemical failure. A cohort of 279 patients with localised prostate adenocarcinoma was used. Analogous virtual tissues were built from pre-treatment MRIs. The prescribed standard irradiation schedules were simulated using the mechanistic model. Biochemical recurrence was predicted from the in silico number of tumour cells at the end of treatment (AUC = 0.68). Then, alternative 2.5 and 3 Gy fractionations were simulated for patients who suffered biochemical recurrence. Significantly lower numbers of tumour cells at the end of treatment were obtained after these hypofractionated schedules. Significant decreases in total doses assuring tumour control were also observed for these patients (median of -10.3 and -14.0 Gy for 2.5 and 3 Gy fractionations, respectively).","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"26 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":"115254271","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/BHI56158.2022.9926774
V. Pezoulas, Pashupati P. Mishra, Olli T. Raitakari, M. Kahonen, T. Lehtimaki, D. Fotiadis, A. Sakellarios
Carotid artery disease (CAD) may be responsible for a stroke with fatal consequences for the patients. Early and non-invasive diagnosis and prediction of significantly high carotid intima media thickness (IMT) can reduce the death rates caused by cardiovascular disease. Machine learning can be applied for the development of robust models for this purpose when adequate data are available. In this work, we utilized metabolomics data from 2,147 patients in the Young Finns Study clinical trial to predict the high intima media thickness as a prodromal stage of the atherosclerotic carotid disease. An explainable AI based pipeline was developed which includes a novel employment of the Gradient Boosted Trees (GBT). More specifically, a hybrid loss function was used to adjust the effect of the dropout rates in the ‘dart’ booster in the loss function topology. The results of our analysis demonstrate that the novel implementation of the GBT improves the results in terms of the sensitivity which is the most important requirement to our analysis (accuracy 0.80, sensitivity 0.86, AUC 0.85). Moreover, it is shown that metabolomics can be used to increase sensitivity in predicting the increased IMT.
{"title":"Metabolomics in the prediction of prodromal stages of carotid artery disease using a hybrid ML algorithm","authors":"V. Pezoulas, Pashupati P. Mishra, Olli T. Raitakari, M. Kahonen, T. Lehtimaki, D. Fotiadis, A. Sakellarios","doi":"10.1109/BHI56158.2022.9926774","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926774","url":null,"abstract":"Carotid artery disease (CAD) may be responsible for a stroke with fatal consequences for the patients. Early and non-invasive diagnosis and prediction of significantly high carotid intima media thickness (IMT) can reduce the death rates caused by cardiovascular disease. Machine learning can be applied for the development of robust models for this purpose when adequate data are available. In this work, we utilized metabolomics data from 2,147 patients in the Young Finns Study clinical trial to predict the high intima media thickness as a prodromal stage of the atherosclerotic carotid disease. An explainable AI based pipeline was developed which includes a novel employment of the Gradient Boosted Trees (GBT). More specifically, a hybrid loss function was used to adjust the effect of the dropout rates in the ‘dart’ booster in the loss function topology. The results of our analysis demonstrate that the novel implementation of the GBT improves the results in terms of the sensitivity which is the most important requirement to our analysis (accuracy 0.80, sensitivity 0.86, AUC 0.85). Moreover, it is shown that metabolomics can be used to increase sensitivity in predicting the increased IMT.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"29 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":"128368423","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/BHI56158.2022.9926939
Kexin Feng, Theodora Chaspari
Psychomotor retardation associated with depression has been linked with tangible differences in vowel production. This paper investigates a knowledge-driven machine learning (ML) method that integrates spectrotemporal information of speech at the vowel-level to identify the depression. Low-level speech descriptors are learned by a convolutional neural network (CNN) that is trained for vowel classification. The temporal evolution of those low-level descriptors is modeled at the high-level within and across utterances via a long short-term memory (LSTM) model that takes the final depression decision. A modified version of the Local Interpretable Model-agnostic Explanations (LIME) is further used to identify the impact of the low-level spectrotemporal vowel variation on the decisions and observe the high-level temporal change of the depression likelihood. The proposed method outperforms baselines that model the spectrotemporal information in speech without integrating the vowel-based information, as well as ML models trained with conventional prosodic and spectrotemporal features. The conducted explainability analysis indicates that spectrotemporal information corresponding to non-vowel segments less important than the vowel-based information. Explainability of the high-level information capturing the segment-by-segment decisions is further inspected for participants with and without depression. The findings from this work can provide the foundation toward knowledge-driven interpretable decision-support systems that can assist clinicians to better understand fine-grain temporal changes in speech data, ultimately augmenting mental health diagnosis and care.
{"title":"Toward Knowledge-Driven Speech-Based Models of Depression: Leveraging Spectrotemporal Variations in Speech Vowels","authors":"Kexin Feng, Theodora Chaspari","doi":"10.1109/BHI56158.2022.9926939","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926939","url":null,"abstract":"Psychomotor retardation associated with depression has been linked with tangible differences in vowel production. This paper investigates a knowledge-driven machine learning (ML) method that integrates spectrotemporal information of speech at the vowel-level to identify the depression. Low-level speech descriptors are learned by a convolutional neural network (CNN) that is trained for vowel classification. The temporal evolution of those low-level descriptors is modeled at the high-level within and across utterances via a long short-term memory (LSTM) model that takes the final depression decision. A modified version of the Local Interpretable Model-agnostic Explanations (LIME) is further used to identify the impact of the low-level spectrotemporal vowel variation on the decisions and observe the high-level temporal change of the depression likelihood. The proposed method outperforms baselines that model the spectrotemporal information in speech without integrating the vowel-based information, as well as ML models trained with conventional prosodic and spectrotemporal features. The conducted explainability analysis indicates that spectrotemporal information corresponding to non-vowel segments less important than the vowel-based information. Explainability of the high-level information capturing the segment-by-segment decisions is further inspected for participants with and without depression. The findings from this work can provide the foundation toward knowledge-driven interpretable decision-support systems that can assist clinicians to better understand fine-grain temporal changes in speech data, ultimately augmenting mental health diagnosis and care.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"33 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":"126767054","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/BHI56158.2022.9926910
A. Berti, Gianluca Carloni, S. Colantonio, M. A. Pascali, P. Manghi, P. Pagano, Rossana Buongiorno, Eva Pachetti, C. Caudai, Domenico Di Gangi, E. Carlini, Z. Falaschi, E. Ciarrocchi, E. Neri, E. Bertelli, V. Miele, R. Carpi, G. Bagnacci, Nunzia Di Meglio, M. Mazzei, A. Barucci
Researchers nowadays may take advantage of broad collections of medical data to develop personalized medicine solutions. Imaging bio-banks play a fundamental role, in this regard, by serving as organized repositories of medical images associated with imaging biomarkers. In this context, the NAVIGATOR Project aims to advance colorectal, prostate, and gastric oncology translational research by leveraging quantitative imaging and multi-omics analyses. As Project's core, an imaging bio-bank is being designed and implemented in a web-accessible Virtual Research Environment (VRE). The VRE serves to extract the imaging biomarkers and further process them within prediction algorithms. In our work, we present the realization of the data models for the three cancer use-cases of the Project. First, we carried out an extensive requirements analysis to fulfill the necessities of the clinical partners involved in the Project. Then, we designed three separate data models utilizing entity-relationship diagrams. We found diagrams' modeling for colorectal and prostate cancers to be more straightforward, while gastric cancer required a higher level of complexity. Future developments of this work would include designing a common data model following the Observational Medical Outcomes Partnership Standards. Indeed, a common data model would standardize the logical infrastructure of data models and make the bio-bank easily interonerable with other bio-banks.
{"title":"Data Models for an Imaging Bio-bank for Colorectal, Prostate and Gastric Cancer: the NAVIGATOR Project","authors":"A. Berti, Gianluca Carloni, S. Colantonio, M. A. Pascali, P. Manghi, P. Pagano, Rossana Buongiorno, Eva Pachetti, C. Caudai, Domenico Di Gangi, E. Carlini, Z. Falaschi, E. Ciarrocchi, E. Neri, E. Bertelli, V. Miele, R. Carpi, G. Bagnacci, Nunzia Di Meglio, M. Mazzei, A. Barucci","doi":"10.1109/BHI56158.2022.9926910","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926910","url":null,"abstract":"Researchers nowadays may take advantage of broad collections of medical data to develop personalized medicine solutions. Imaging bio-banks play a fundamental role, in this regard, by serving as organized repositories of medical images associated with imaging biomarkers. In this context, the NAVIGATOR Project aims to advance colorectal, prostate, and gastric oncology translational research by leveraging quantitative imaging and multi-omics analyses. As Project's core, an imaging bio-bank is being designed and implemented in a web-accessible Virtual Research Environment (VRE). The VRE serves to extract the imaging biomarkers and further process them within prediction algorithms. In our work, we present the realization of the data models for the three cancer use-cases of the Project. First, we carried out an extensive requirements analysis to fulfill the necessities of the clinical partners involved in the Project. Then, we designed three separate data models utilizing entity-relationship diagrams. We found diagrams' modeling for colorectal and prostate cancers to be more straightforward, while gastric cancer required a higher level of complexity. Future developments of this work would include designing a common data model following the Observational Medical Outcomes Partnership Standards. Indeed, a common data model would standardize the logical infrastructure of data models and make the bio-bank easily interonerable with other bio-banks.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"38 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":"128176605","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/BHI56158.2022.9926879
Boyuan Peng, Yiyang Liu, Xin Zhu, Shouhei Ikeda, S. Tsunoda
Hematological malignancies are a lethal disease that seriously endangers human lives. In addition to bone marrow biopsy, the use of MRI to analyze the bone marrow of femur is a new and efficient diagnostic method for hematological tumors. Accurate segmentation of femur plays a crucial role in screening this disease. In this paper, we compared four neural networks (PP-LiteSeg, U-Net, SegNet, and PspNet) for femur segmentation using 579 training and testing MRI images from 200 patients with HM. PP-LiteSeg demonstrated the best performance with an average Dice coefficient of 0.92.
{"title":"Femoral segmentation of MRI images using PP-LiteSeg","authors":"Boyuan Peng, Yiyang Liu, Xin Zhu, Shouhei Ikeda, S. Tsunoda","doi":"10.1109/BHI56158.2022.9926879","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926879","url":null,"abstract":"Hematological malignancies are a lethal disease that seriously endangers human lives. In addition to bone marrow biopsy, the use of MRI to analyze the bone marrow of femur is a new and efficient diagnostic method for hematological tumors. Accurate segmentation of femur plays a crucial role in screening this disease. In this paper, we compared four neural networks (PP-LiteSeg, U-Net, SegNet, and PspNet) for femur segmentation using 579 training and testing MRI images from 200 patients with HM. PP-LiteSeg demonstrated the best performance with an average Dice coefficient of 0.92.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"36 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":"127375527","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/BHI56158.2022.9926874
Beniamino Daniele, Giulio Steyde, Edoardo Spairani, G. Magenes, M. Signorini
The purpose of this study is to develop and understand whether Machine Learning models can classify Cardiotocographic (CTG) recordings of healthy fetuses or Intra Uterine Growth Restricted (IUGR) fetuses, highlighting how a large amount of data can have unexpected effects. We started from other findings in the literature to see what Machine Learning model remained consistent even with a large amount of data. The CTG records used in this study were collected at the Department of Obstetrics of the Federico II University Hospital in Naples, Italy, from 2013 to 2021. From this dataset, we chose 1548 IUGR fetuses and 1548 healthy fetuses to train our models. Each recording contained several parameters, ranging from features calculated on the entire CTG tracing, features calculated every 3 and 1 minute of recording and features related to the pregnant woman, such as age and week of gestation. We trained our machine-learning models on this dataset, checking the results obtained before and after adjusting the hyperparameters, noting that among the best models was Random Forest, which has already been present in other studies, and that the Multilayer Perceptron and the AdaBoost classifier were overall the best performing. This work can surely form a basis for future works in the fetal heart rate classification thus leading to real clinical applications.
{"title":"Discriminating Healthy and IUGR fetuses through Machine Learning models","authors":"Beniamino Daniele, Giulio Steyde, Edoardo Spairani, G. Magenes, M. Signorini","doi":"10.1109/BHI56158.2022.9926874","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926874","url":null,"abstract":"The purpose of this study is to develop and understand whether Machine Learning models can classify Cardiotocographic (CTG) recordings of healthy fetuses or Intra Uterine Growth Restricted (IUGR) fetuses, highlighting how a large amount of data can have unexpected effects. We started from other findings in the literature to see what Machine Learning model remained consistent even with a large amount of data. The CTG records used in this study were collected at the Department of Obstetrics of the Federico II University Hospital in Naples, Italy, from 2013 to 2021. From this dataset, we chose 1548 IUGR fetuses and 1548 healthy fetuses to train our models. Each recording contained several parameters, ranging from features calculated on the entire CTG tracing, features calculated every 3 and 1 minute of recording and features related to the pregnant woman, such as age and week of gestation. We trained our machine-learning models on this dataset, checking the results obtained before and after adjusting the hyperparameters, noting that among the best models was Random Forest, which has already been present in other studies, and that the Multilayer Perceptron and the AdaBoost classifier were overall the best performing. This work can surely form a basis for future works in the fetal heart rate classification thus leading to real clinical applications.","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":"130590019","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/BHI56158.2022.9926933
Hooman Rokham, Haleh Falakshahi, V. Calhoun
Among the non-invasive neuroimaging techniques, resting-state functional magnitude imaging is the most widely used method for capturing whole brain activity. Functional connectivity enables us to extract brain networks which exhibit temporal coherence from resting-state fMRI data. However, there are some limitations to fMRI which limit the questions we can ask. The latency estimated from fMRI is a mixture of the sluggish hemodynamic delay and neural latencies. Due to the large spatially varying delays related to hemodynamics, the pattern and order of activities between brain regions in a very short period will be driven by hemodynamics in this case. In this study, we proposed a method to estimate the hemodynamic delays between brain regions. We performed cross-correlation between pairs of time courses and estimated the optimal lags such that the correlation is maximized. We applied our method to a large dataset of adolescents and investigated the differences between males and females on different lag measures. In addition, we proposed short and long-time delay graphs to visualize the differences between groups more easily. Our result suggests that the female subjects had shorter hemodynamic delay compared to the male group of the same age. Significant differences were identified both within and between domain regions, including the cerebellar, somatomotor, default mode, cognitive control, and visual domain.
{"title":"Relationship of Hemodynamic Delay and Sex Differences Among Adolescents Using Resting-state fMRI Data","authors":"Hooman Rokham, Haleh Falakshahi, V. Calhoun","doi":"10.1109/BHI56158.2022.9926933","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926933","url":null,"abstract":"Among the non-invasive neuroimaging techniques, resting-state functional magnitude imaging is the most widely used method for capturing whole brain activity. Functional connectivity enables us to extract brain networks which exhibit temporal coherence from resting-state fMRI data. However, there are some limitations to fMRI which limit the questions we can ask. The latency estimated from fMRI is a mixture of the sluggish hemodynamic delay and neural latencies. Due to the large spatially varying delays related to hemodynamics, the pattern and order of activities between brain regions in a very short period will be driven by hemodynamics in this case. In this study, we proposed a method to estimate the hemodynamic delays between brain regions. We performed cross-correlation between pairs of time courses and estimated the optimal lags such that the correlation is maximized. We applied our method to a large dataset of adolescents and investigated the differences between males and females on different lag measures. In addition, we proposed short and long-time delay graphs to visualize the differences between groups more easily. Our result suggests that the female subjects had shorter hemodynamic delay compared to the male group of the same age. Significant differences were identified both within and between domain regions, including the cerebellar, somatomotor, default mode, cognitive control, and visual domain.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"27 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":"133829128","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/BHI56158.2022.9926954
Haowen Wang, Jia Huang, Guowei Wang, Hongzhou Lu, Wenjin Wang
Camera-based vital signs monitoring has been extensively researched in non-medical fields in recent years. Intensive Care Unit (ICU) typically requires continuous monitoring of patients' physiology for alarming the emergency such as patient deterioration or delirium. In this paper, we propose to use the surveillance closed-circuit television (CCTV) cameras installed in ICU for cardio-respiratory monitoring of critically-ill patients, thus created a first clinical video dataset (including 10 deteriorated patients) in ICU using CCTV cameras. Along with the dataset, a video processing framework with the latest core algorithms designed for pulse and respiratory signal extraction has been demonstrated. A joint Region-of-Interest optimization approach using pulsatile living-skin maps and respiratory maps was proposed to improve the vital signs monitoring for ICU patients. A motion intensity based quality metric was designed to reject measurement outliers induced by patient motion or nurse operation. Based on the valid measurements selected by the metric, the overall Mean Absolute Error for heart rate is 1.7 bpm, and for breathing rate is 1.6 bpm. Preliminary clinical validations show that robust cardio-respiratory monitoring is indeed feasible for CCTV cameras in ICU, and such a warding solution can be quickly integrated into current hospital information systems for large-scale deployment, by leveraging the existing hardware and infrastructures of the Internet of Medical Things.
{"title":"Surveillance Camera-based Cardio-respiratory Monitoring for Critical Patients in ICU","authors":"Haowen Wang, Jia Huang, Guowei Wang, Hongzhou Lu, Wenjin Wang","doi":"10.1109/BHI56158.2022.9926954","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926954","url":null,"abstract":"Camera-based vital signs monitoring has been extensively researched in non-medical fields in recent years. Intensive Care Unit (ICU) typically requires continuous monitoring of patients' physiology for alarming the emergency such as patient deterioration or delirium. In this paper, we propose to use the surveillance closed-circuit television (CCTV) cameras installed in ICU for cardio-respiratory monitoring of critically-ill patients, thus created a first clinical video dataset (including 10 deteriorated patients) in ICU using CCTV cameras. Along with the dataset, a video processing framework with the latest core algorithms designed for pulse and respiratory signal extraction has been demonstrated. A joint Region-of-Interest optimization approach using pulsatile living-skin maps and respiratory maps was proposed to improve the vital signs monitoring for ICU patients. A motion intensity based quality metric was designed to reject measurement outliers induced by patient motion or nurse operation. Based on the valid measurements selected by the metric, the overall Mean Absolute Error for heart rate is 1.7 bpm, and for breathing rate is 1.6 bpm. Preliminary clinical validations show that robust cardio-respiratory monitoring is indeed feasible for CCTV cameras in ICU, and such a warding solution can be quickly integrated into current hospital information systems for large-scale deployment, by leveraging the existing hardware and infrastructures of the Internet of Medical Things.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"29 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":"133405703","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/BHI56158.2022.9926931
Krittika Goyal, D. Borkholder, S. Day
Non-invasive monitoring of pulmonary health could revolutionize the care of health conditions ranging from COVID-19 to asthma to heart failure, but current technologies face challenges that limit their feasibility and adoption. Here, we introduce a novel approach to monitor respiration by measuring changes in impedance from the back of the thigh. The integration of electrodes into a toilet seat ensures patient compliance with unobtrusive daily respiration monitoring benefitting from repeatable electrode placement on the skin. In this work, the feasibility of the thigh and the sensitivity of impedance to respiration have been investigated empirically by comparing thorax and thigh-thigh bioimpedance measurements to spirometer measurements, and computationally, using finite element modeling. Empirical results show a measurable peak-peak impedance (0.022 ohm to 0.290 ohm for normal breathing across 8 subjects) with respiration across thigh-thigh and a high correlation (0.85) between lung tidal volume and impedance change due to respiration. Thigh-thigh bioimpedance measurements were found to be able to distinguish between shallow, normal, and deep breathing. Further, day-to-day variability in the relationship between impedance and tidal volume was investigated. The results suggest that the novel approach can be used to detect respiration rate and tidal volume and could provide valuable insight into disease state for conditions ranging from COVID-19 to heart failure.
{"title":"Unobtrusive In-Home Respiration Monitoring Using a Toilet Seat","authors":"Krittika Goyal, D. Borkholder, S. Day","doi":"10.1109/BHI56158.2022.9926931","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926931","url":null,"abstract":"Non-invasive monitoring of pulmonary health could revolutionize the care of health conditions ranging from COVID-19 to asthma to heart failure, but current technologies face challenges that limit their feasibility and adoption. Here, we introduce a novel approach to monitor respiration by measuring changes in impedance from the back of the thigh. The integration of electrodes into a toilet seat ensures patient compliance with unobtrusive daily respiration monitoring benefitting from repeatable electrode placement on the skin. In this work, the feasibility of the thigh and the sensitivity of impedance to respiration have been investigated empirically by comparing thorax and thigh-thigh bioimpedance measurements to spirometer measurements, and computationally, using finite element modeling. Empirical results show a measurable peak-peak impedance (0.022 ohm to 0.290 ohm for normal breathing across 8 subjects) with respiration across thigh-thigh and a high correlation (0.85) between lung tidal volume and impedance change due to respiration. Thigh-thigh bioimpedance measurements were found to be able to distinguish between shallow, normal, and deep breathing. Further, day-to-day variability in the relationship between impedance and tidal volume was investigated. The results suggest that the novel approach can be used to detect respiration rate and tidal volume and could provide valuable insight into disease state for conditions ranging from COVID-19 to heart failure.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"57 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":"132438645","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/BHI56158.2022.9926803
K. Tsarapatsani, Antonis I. Sakellarios, V. Pezoulas, V. Tsakanikas, G. Matsopoulos, W. März, M. Kleber, D. Fotiadis
The early prevention of myocardial infarction (MI), a complication of cardiovascular disease (CVD), is an urgent need for the timely provision of medical intervention and the reduction of cardiovascular mortality. The performance of machine learning (ML) has proven useful in aiding the early diagnosis of this disease. In this work, we utilize clinical cardiovascular disease risk factors and biochemical data, employing machine learning models i.e. Random Forest (RF), Extreme Grading Boosting (XGBoost) and Adaptive Boosting (AdaBoost), to predict the 10-year risk of myocardial infarction in patients with 10-years follow-up for CVD. We used the cohort of the Ludwigshafen Risk and Cardiovascular Health (LURIC) study, while 3267 patients were included in the analysis (1361 suffered from MI). We calculated the performance of machine learning models, more specifically the mean values of Accuracy (ACC), Sensitivity, Specificity and the area under the receiver operating characteristic curve (AUC) of each model. We also plotted the corresponding receiver operating characteristic curve for each model. The findings of the analysis reveal that the Extreme Gradient Boosting model detects MI with the highest accuracy (74.27 %). Moreover, explainable artificial intelligence was applied, especially the Shapley values were calculated to identify the most important features and interpret the results with XGBoost.
{"title":"Machine Learning Models to Predict Myocardial Infarction Within 10-Years Follow-up of Cardiovascular Disease Progression","authors":"K. Tsarapatsani, Antonis I. Sakellarios, V. Pezoulas, V. Tsakanikas, G. Matsopoulos, W. März, M. Kleber, D. Fotiadis","doi":"10.1109/BHI56158.2022.9926803","DOIUrl":"https://doi.org/10.1109/BHI56158.2022.9926803","url":null,"abstract":"The early prevention of myocardial infarction (MI), a complication of cardiovascular disease (CVD), is an urgent need for the timely provision of medical intervention and the reduction of cardiovascular mortality. The performance of machine learning (ML) has proven useful in aiding the early diagnosis of this disease. In this work, we utilize clinical cardiovascular disease risk factors and biochemical data, employing machine learning models i.e. Random Forest (RF), Extreme Grading Boosting (XGBoost) and Adaptive Boosting (AdaBoost), to predict the 10-year risk of myocardial infarction in patients with 10-years follow-up for CVD. We used the cohort of the Ludwigshafen Risk and Cardiovascular Health (LURIC) study, while 3267 patients were included in the analysis (1361 suffered from MI). We calculated the performance of machine learning models, more specifically the mean values of Accuracy (ACC), Sensitivity, Specificity and the area under the receiver operating characteristic curve (AUC) of each model. We also plotted the corresponding receiver operating characteristic curve for each model. The findings of the analysis reveal that the Extreme Gradient Boosting model detects MI with the highest accuracy (74.27 %). Moreover, explainable artificial intelligence was applied, especially the Shapley values were calculated to identify the most important features and interpret the results with XGBoost.","PeriodicalId":347210,"journal":{"name":"2022 IEEE-EMBS International Conference on Biomedical and Health Informatics (BHI)","volume":"40 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":"123920164","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：