Pub Date : 2022-06-22DOI: 10.1109/MeMeA54994.2022.9856508
Vaibhav Joshi, S. Vijayarangan, S. Preejith, M. Sivaprakasam
An electroencephalogram (EEG) signal is currently accepted as a standard for automatic sleep staging. Lately, Near-human accuracy in automated sleep staging has been achievable by Deep Learning (DL) based approaches, enabling multi-fold progress in this area. However, An extensive and expensive clinical setup is required for EEG based sleep staging. Additionally, the EEG setup being obtrusive in nature and requiring an expert for setup adds to the inconvenience of the subject under study, making it adverse in the point of care setting. An unobtrusive and more suitable alternative to EEG is Electrocardiogram (ECG). Unsurprisingly, compared to EEG in sleep staging, its performance remains sub-par. In order to take advantage of both the modalities, transferring knowledge from EEG to ECG is a reasonable approach, ultimately boosting the performance of ECG based sleep staging. Knowledge Distillation (KD) is a promising notion in DL that shares knowledge from a superior performing but usually more complex teacher model to an inferior but compact student model. Building upon this concept, a cross-modality KD framework assisting features learned through models trained on EEG to improve ECG-based sleep staging performance is proposed. Additionally, to better understand the distillation approach, extensive experimentation on the independent modules of the proposed model was conducted. Montreal Archive of Sleep Studies (MASS) dataset consisting of 200 subjects was utilized for this study. The results from the proposed model for weighted-F1-score in 3-class and 4-class sleep staging showed a 13.40 % and 14.30 % improvement, respectively. This study demonstrates the feasibility of KD for single-channel ECG based sleep staging's performance enhancement in 3-class (W-R-N) and 4-class (W-R-L-D) classification.
{"title":"EEG aided boosting of single-lead ECG based sleep staging with Deep Knowledge Distillation","authors":"Vaibhav Joshi, S. Vijayarangan, S. Preejith, M. Sivaprakasam","doi":"10.1109/MeMeA54994.2022.9856508","DOIUrl":"https://doi.org/10.1109/MeMeA54994.2022.9856508","url":null,"abstract":"An electroencephalogram (EEG) signal is currently accepted as a standard for automatic sleep staging. Lately, Near-human accuracy in automated sleep staging has been achievable by Deep Learning (DL) based approaches, enabling multi-fold progress in this area. However, An extensive and expensive clinical setup is required for EEG based sleep staging. Additionally, the EEG setup being obtrusive in nature and requiring an expert for setup adds to the inconvenience of the subject under study, making it adverse in the point of care setting. An unobtrusive and more suitable alternative to EEG is Electrocardiogram (ECG). Unsurprisingly, compared to EEG in sleep staging, its performance remains sub-par. In order to take advantage of both the modalities, transferring knowledge from EEG to ECG is a reasonable approach, ultimately boosting the performance of ECG based sleep staging. Knowledge Distillation (KD) is a promising notion in DL that shares knowledge from a superior performing but usually more complex teacher model to an inferior but compact student model. Building upon this concept, a cross-modality KD framework assisting features learned through models trained on EEG to improve ECG-based sleep staging performance is proposed. Additionally, to better understand the distillation approach, extensive experimentation on the independent modules of the proposed model was conducted. Montreal Archive of Sleep Studies (MASS) dataset consisting of 200 subjects was utilized for this study. The results from the proposed model for weighted-F1-score in 3-class and 4-class sleep staging showed a 13.40 % and 14.30 % improvement, respectively. This study demonstrates the feasibility of KD for single-channel ECG based sleep staging's performance enhancement in 3-class (W-R-N) and 4-class (W-R-L-D) classification.","PeriodicalId":106228,"journal":{"name":"2022 IEEE International Symposium on Medical Measurements and Applications (MeMeA)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130932606","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 : 2022-06-22DOI: 10.1109/MeMeA54994.2022.9856590
Martina Zaltieri, C. Massaroni, S. Bianchi, F. M. Cauti, E. Schena
Atrial fibrillation (AF) is the most recurrent type of cardiac arrhythmia that causes a major socio-economic burden as associated with significant morbidity and mortality. Radiofrequency catheter ablation (RFCA) is a leading clinical practice for the treatment of AF. The aim of the procedure is to induce necrosis in the ectopic foci responsible for the altered electrical pathway through temperature increments provoked by radiofrequency delivery. In this context, temperature is a key factor as determines the size of the produced thermal lesions and, in turn, the success or the failure of the treatment. As consequence, several methods have been exploited for RFCA temperature monitoring, but with several limitations. In recent times, the feasibility of using fiber Bragg grating (FBG) sensors for high-resolved and distributed temperature measurements in ex vivo myocardial swine tissues has been assessed. In this study, the heat diffusion within the tissues was investigated by producing 2D thermal maps based on multipoint FBG temperature data. Furthermore, the influence of both the delivered power and the treatment time on the dimensions of the produced thermal lesion was explored. The present research may lay the basis for the development of a model describing the spatio-temporal dynamics of the lesion formation. Such model may offer support to clinicians in selecting the proper RFCA settings.
{"title":"Analysis of the Spatio-Temporal Dynamics of Thermal Lesion Formation in Radiofrequency Cardiac Ablation","authors":"Martina Zaltieri, C. Massaroni, S. Bianchi, F. M. Cauti, E. Schena","doi":"10.1109/MeMeA54994.2022.9856590","DOIUrl":"https://doi.org/10.1109/MeMeA54994.2022.9856590","url":null,"abstract":"Atrial fibrillation (AF) is the most recurrent type of cardiac arrhythmia that causes a major socio-economic burden as associated with significant morbidity and mortality. Radiofrequency catheter ablation (RFCA) is a leading clinical practice for the treatment of AF. The aim of the procedure is to induce necrosis in the ectopic foci responsible for the altered electrical pathway through temperature increments provoked by radiofrequency delivery. In this context, temperature is a key factor as determines the size of the produced thermal lesions and, in turn, the success or the failure of the treatment. As consequence, several methods have been exploited for RFCA temperature monitoring, but with several limitations. In recent times, the feasibility of using fiber Bragg grating (FBG) sensors for high-resolved and distributed temperature measurements in ex vivo myocardial swine tissues has been assessed. In this study, the heat diffusion within the tissues was investigated by producing 2D thermal maps based on multipoint FBG temperature data. Furthermore, the influence of both the delivered power and the treatment time on the dimensions of the produced thermal lesion was explored. The present research may lay the basis for the development of a model describing the spatio-temporal dynamics of the lesion formation. Such model may offer support to clinicians in selecting the proper RFCA settings.","PeriodicalId":106228,"journal":{"name":"2022 IEEE International Symposium on Medical Measurements and Applications (MeMeA)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134459649","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 : 2022-06-22DOI: 10.1109/MeMeA54994.2022.9856546
C. Carissimo, L. Ferrigno, Giacomo Golluccio, Alessandro Marino, G. Cerro
The usage of mini-devices in medicine for continuous non-invasive monitoring of neurodegenerative diseases is rapidly increasing. Among most common diseases belonging to such category, Parkinson's is one of the main disorders, especially in aged population. It is characterized by several symptoms whose comprehensive and accurate analysis can lead to a punctual and effective diagnosis. This task is generally accomplished by an expert medical doctor but, especially in first stage, the aid of an automatic tool can help to catch even very low symptomatology. A promising solution to detect most motor issues related to Parkinson's disease is represented by Inertial Measurement Units (IMUs), typically including accelerometers, magnetometers and gyroscopes. Their metrological features, such as accuracy, sensitivity and immunity to external disturbances are critical to get a fully functional and discriminant device. Furthermore, the capability to extrapolate pathological states from measurements is a very attractive feature to automatize early warning and fast medical interventions. To accomplish for both tasks, in this paper a measuring platform containing an IMU is presented and metrologically characterized; moreover, classification tests for typical impairments due to Parkinson's disease are proposed. Although improvements in the procedure and measurement quality are on the way, the current status allows to state its suitability for the required application framework.
{"title":"Parkinson's disease aided diagnosis: online symptoms detection by a low-cost wearable Inertial Measurement Unit","authors":"C. Carissimo, L. Ferrigno, Giacomo Golluccio, Alessandro Marino, G. Cerro","doi":"10.1109/MeMeA54994.2022.9856546","DOIUrl":"https://doi.org/10.1109/MeMeA54994.2022.9856546","url":null,"abstract":"The usage of mini-devices in medicine for continuous non-invasive monitoring of neurodegenerative diseases is rapidly increasing. Among most common diseases belonging to such category, Parkinson's is one of the main disorders, especially in aged population. It is characterized by several symptoms whose comprehensive and accurate analysis can lead to a punctual and effective diagnosis. This task is generally accomplished by an expert medical doctor but, especially in first stage, the aid of an automatic tool can help to catch even very low symptomatology. A promising solution to detect most motor issues related to Parkinson's disease is represented by Inertial Measurement Units (IMUs), typically including accelerometers, magnetometers and gyroscopes. Their metrological features, such as accuracy, sensitivity and immunity to external disturbances are critical to get a fully functional and discriminant device. Furthermore, the capability to extrapolate pathological states from measurements is a very attractive feature to automatize early warning and fast medical interventions. To accomplish for both tasks, in this paper a measuring platform containing an IMU is presented and metrologically characterized; moreover, classification tests for typical impairments due to Parkinson's disease are proposed. Although improvements in the procedure and measurement quality are on the way, the current status allows to state its suitability for the required application framework.","PeriodicalId":106228,"journal":{"name":"2022 IEEE International Symposium on Medical Measurements and Applications (MeMeA)","volume":"194 ","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133323043","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 : 2022-06-22DOI: 10.1109/MeMeA54994.2022.9856543
Alexiane Pasquier, Y. Diraison, S. Serfaty, P. Joubert
The dielectric properties of tissues have been widely used to detect and monitor different pathologies. One of the remaining challenges is to timely and accurately characterize the evolution of the dielectric properties of tissues in a non-invasive and contactless way, with a simple and portable monitoring system. This paper proposes investigating the use of a loop-shaped transmission line passive resonators (TLR) to sense organic tissue changes in the radiofrequency bandwidth (in the hundreds of MHz bandwidth), through inductive coupling with the tissue. This kind of sensor can be wirelessly excited, and is able to distantly detect the dielectric modifications in the targeted tissue through the changes of the transmitted electromagnetic field. TLR-based sensors are therefore very promising for the non-invasive, wearable and continuous monitoring of tissues. In this paper, a first study is carried out to monitor the decomposition of a beef muscle sample for six consecutive days with two different TLR-based sensors featuring two investigation frequencies (160 MHz and 350 MHz). The obtained results confirmed the ability of such sensors to follow the modifications of an organic tissue through the assessment of both the conductivity and the relative permittivity of the investigated sample. Results also confirmed that the investigation frequency, for which the loss factor within the tissue is around unity, is particularly well suited to sense changes within the tissue under investigation. A second study was realized with other soft matter samples (water, cottage cheese, water/gelatin mix) to determine the ability of TLR-sensors to discriminate between soft matter of various nature. Thanks to the ability of the TLR-based sensor to assess the loss factor of the monitored samples, it was found that i) the proposed sensor is relevant to discriminate between the considered soft matter samples and ii) that this discrimination can be made particularly efficient when using the appropriate investigation frequency. Furthermore, the benefits of the use of several investigation frequencies were also demonstrated for enhanced tissue characterizations. TLR-based sensors are therefore good candidates for the non-invasive, low-cost and sensitive sensing devices dedicated to the monitoring of pathologies such as wound healing and cancer detection.
{"title":"Non-contact inductive radiofrequency monitoring of a beef muscle tissue decomposition","authors":"Alexiane Pasquier, Y. Diraison, S. Serfaty, P. Joubert","doi":"10.1109/MeMeA54994.2022.9856543","DOIUrl":"https://doi.org/10.1109/MeMeA54994.2022.9856543","url":null,"abstract":"The dielectric properties of tissues have been widely used to detect and monitor different pathologies. One of the remaining challenges is to timely and accurately characterize the evolution of the dielectric properties of tissues in a non-invasive and contactless way, with a simple and portable monitoring system. This paper proposes investigating the use of a loop-shaped transmission line passive resonators (TLR) to sense organic tissue changes in the radiofrequency bandwidth (in the hundreds of MHz bandwidth), through inductive coupling with the tissue. This kind of sensor can be wirelessly excited, and is able to distantly detect the dielectric modifications in the targeted tissue through the changes of the transmitted electromagnetic field. TLR-based sensors are therefore very promising for the non-invasive, wearable and continuous monitoring of tissues. In this paper, a first study is carried out to monitor the decomposition of a beef muscle sample for six consecutive days with two different TLR-based sensors featuring two investigation frequencies (160 MHz and 350 MHz). The obtained results confirmed the ability of such sensors to follow the modifications of an organic tissue through the assessment of both the conductivity and the relative permittivity of the investigated sample. Results also confirmed that the investigation frequency, for which the loss factor within the tissue is around unity, is particularly well suited to sense changes within the tissue under investigation. A second study was realized with other soft matter samples (water, cottage cheese, water/gelatin mix) to determine the ability of TLR-sensors to discriminate between soft matter of various nature. Thanks to the ability of the TLR-based sensor to assess the loss factor of the monitored samples, it was found that i) the proposed sensor is relevant to discriminate between the considered soft matter samples and ii) that this discrimination can be made particularly efficient when using the appropriate investigation frequency. Furthermore, the benefits of the use of several investigation frequencies were also demonstrated for enhanced tissue characterizations. TLR-based sensors are therefore good candidates for the non-invasive, low-cost and sensitive sensing devices dedicated to the monitoring of pathologies such as wound healing and cancer detection.","PeriodicalId":106228,"journal":{"name":"2022 IEEE International Symposium on Medical Measurements and Applications (MeMeA)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129462221","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 : 2022-06-22DOI: 10.1109/MeMeA54994.2022.9856477
V. D'Alessandro, Francesco Paciolla, Luisa De Palma, F. Adamo, A. Nisio, N. Giaquinto
Human errors in specimen identification and incorrect blood transfusions in hospitals cause economic losses for several million dollars and many adverse events each year, posing serious risks for patient health and carrying huge expenses for the health care system. This article presents the control of a 6 DOF (Degrees of Freedom) Mitsubishi's robot of the RV -Series to classify test tubes according to the sample's blood type (0, A or B) stored in the RFID (Radio Frequency IDentification) tag's memory. The automatization of processes using a robot, which is able to carry out repetitive and monotonous tasks, improves the standard of care and allows to reduce mortality among patients receiving transfusions with automatically classified blood. On each test tube a readable/writable MIFARE Ultralight® tag uniquely identified with a UID (Unique Identifier) has been applied. The classification is performed using the MFRC522 RFID IC (Integrated Circuit) reader connected to an Arduino UNO R3 board using a Serial Peripheral Interface bus. The execution of the task is performed only with linear trajectories and requires the development of two different levels of controllers. Moreover, additive manufacturing techniques have been used to shape both 3D printed screw cap of the test tubes and the Arduino case to hold the board.
人为错误的标本鉴定和医院不正确的输血每年造成数百万美元的经济损失和许多不良事件，给患者健康带来严重风险，并为医疗保健系统带来巨额费用。本文介绍了对三菱RV -系列6自由度机器人的控制，根据存储在RFID(射频识别)标签存储器中的样本血型(0、a或B)对试管进行分类。使用能够执行重复和单调任务的机器人实现流程自动化，提高了护理标准，并降低了接受自动分类血液输血的患者的死亡率。在每个试管上，使用UID(唯一标识符)唯一标识的可读/可写MIFARE Ultralight®标签。使用MFRC522 RFID IC(集成电路)读取器通过串行外设接口总线连接到Arduino UNO R3板，执行分类。该任务的执行仅使用线性轨迹执行，并且需要开发两个不同级别的控制器。此外，增材制造技术已被用于塑造试管的3D打印螺旋帽和Arduino外壳，以容纳电路板。
{"title":"Robotized sorter for blood classification using RFID tags","authors":"V. D'Alessandro, Francesco Paciolla, Luisa De Palma, F. Adamo, A. Nisio, N. Giaquinto","doi":"10.1109/MeMeA54994.2022.9856477","DOIUrl":"https://doi.org/10.1109/MeMeA54994.2022.9856477","url":null,"abstract":"Human errors in specimen identification and incorrect blood transfusions in hospitals cause economic losses for several million dollars and many adverse events each year, posing serious risks for patient health and carrying huge expenses for the health care system. This article presents the control of a 6 DOF (Degrees of Freedom) Mitsubishi's robot of the RV -Series to classify test tubes according to the sample's blood type (0, A or B) stored in the RFID (Radio Frequency IDentification) tag's memory. The automatization of processes using a robot, which is able to carry out repetitive and monotonous tasks, improves the standard of care and allows to reduce mortality among patients receiving transfusions with automatically classified blood. On each test tube a readable/writable MIFARE Ultralight® tag uniquely identified with a UID (Unique Identifier) has been applied. The classification is performed using the MFRC522 RFID IC (Integrated Circuit) reader connected to an Arduino UNO R3 board using a Serial Peripheral Interface bus. The execution of the task is performed only with linear trajectories and requires the development of two different levels of controllers. Moreover, additive manufacturing techniques have been used to shape both 3D printed screw cap of the test tubes and the Arduino case to hold the board.","PeriodicalId":106228,"journal":{"name":"2022 IEEE International Symposium on Medical Measurements and Applications (MeMeA)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129816024","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 : 2022-06-22DOI: 10.1109/MeMeA54994.2022.9856570
G. Campobello, C. D. Marchis, G. Gugliandolo, Alberto Giacobbe, G. Crupi, N. Donato
In this paper, a novel near-lossless compression algorithm meant for electromyography (EMG) signals is proposed and its performance is evaluated towards real EMG measurements. Differently from other near-lossless algorithms, the proposed one does not rely on either matrix decompositions or complex transformations but exploits only a straight-forward dynamic range compression and a simple encoding technique. Therefore, considering its inherent low complexity and low memory requirements, it can be easily implemented in resources constrained microcontrollers as those included in low-cost measurement instruments and e-Health Internet of Things applications. The algorithm has been tested on a dataset including dynamic EMG measurements carried out in a real-world measurement campaign on 8 different subjects, where, for each subject, the EMG signals were recorded from 8 different muscles during a pedaling session. Analytical and experimental results revealed that the proposed compression technique is able to achieve a compression ratio (CR) up to 80% with a percentage root mean square distortion (PRD) in the range between 0.34% and 13.7%. Moreover, differently from the other compression algorithms described in the literature, the proposed one allows fixing the maximum absolute error a priori thus making it possible to control and limit the desired distortion level besides the compression procedure.
{"title":"A Simple and Efficient Near-lossless Compression Algorithm for Surface ElectroMyoGraphy Signals","authors":"G. Campobello, C. D. Marchis, G. Gugliandolo, Alberto Giacobbe, G. Crupi, N. Donato","doi":"10.1109/MeMeA54994.2022.9856570","DOIUrl":"https://doi.org/10.1109/MeMeA54994.2022.9856570","url":null,"abstract":"In this paper, a novel near-lossless compression algorithm meant for electromyography (EMG) signals is proposed and its performance is evaluated towards real EMG measurements. Differently from other near-lossless algorithms, the proposed one does not rely on either matrix decompositions or complex transformations but exploits only a straight-forward dynamic range compression and a simple encoding technique. Therefore, considering its inherent low complexity and low memory requirements, it can be easily implemented in resources constrained microcontrollers as those included in low-cost measurement instruments and e-Health Internet of Things applications. The algorithm has been tested on a dataset including dynamic EMG measurements carried out in a real-world measurement campaign on 8 different subjects, where, for each subject, the EMG signals were recorded from 8 different muscles during a pedaling session. Analytical and experimental results revealed that the proposed compression technique is able to achieve a compression ratio (CR) up to 80% with a percentage root mean square distortion (PRD) in the range between 0.34% and 13.7%. Moreover, differently from the other compression algorithms described in the literature, the proposed one allows fixing the maximum absolute error a priori thus making it possible to control and limit the desired distortion level besides the compression procedure.","PeriodicalId":106228,"journal":{"name":"2022 IEEE International Symposium on Medical Measurements and Applications (MeMeA)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130258947","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 : 2022-06-22DOI: 10.1109/MeMeA54994.2022.9856406
Laura de Diego-Otón, Álvaro Hernández, Rubén Nieto, M. C. Pérez-Rubio
The common objective of techniques employed to identify the use of household appliances is related to energy efficiency and the reduction of energy consumption. In addition, through load monitoring it is possible to assess the degree of independence of tenants with minimal invasion of privacy and thus develop sustainable health systems capable of providing the required services remotely. Both approaches should initially deal with the load identification stage. For that purpose, this work presents three different solutions that take the events of the electrical current signal acquired at high frequency and process them for classification by using two different topologies of Artificial Neural Networks (ANN). The data of interest used as input for the ANN in the proposals are the normalized signal captured around the events, the images created by dividing that signal into sections and organizing them in a matrix, and the images coming from the Short Time Fourier Transform (STFT) of the signal around the event. The dataset BLUED is used to carry out the validation of the proposal, where some of the proposed architectures obtain an F1 score above 90 % for more than fifteen devices under classification.
{"title":"Comparison of Neural Networks for High-Sampling Rate NILM Scenario","authors":"Laura de Diego-Otón, Álvaro Hernández, Rubén Nieto, M. C. Pérez-Rubio","doi":"10.1109/MeMeA54994.2022.9856406","DOIUrl":"https://doi.org/10.1109/MeMeA54994.2022.9856406","url":null,"abstract":"The common objective of techniques employed to identify the use of household appliances is related to energy efficiency and the reduction of energy consumption. In addition, through load monitoring it is possible to assess the degree of independence of tenants with minimal invasion of privacy and thus develop sustainable health systems capable of providing the required services remotely. Both approaches should initially deal with the load identification stage. For that purpose, this work presents three different solutions that take the events of the electrical current signal acquired at high frequency and process them for classification by using two different topologies of Artificial Neural Networks (ANN). The data of interest used as input for the ANN in the proposals are the normalized signal captured around the events, the images created by dividing that signal into sections and organizing them in a matrix, and the images coming from the Short Time Fourier Transform (STFT) of the signal around the event. The dataset BLUED is used to carry out the validation of the proposal, where some of the proposed architectures obtain an F1 score above 90 % for more than fifteen devices under classification.","PeriodicalId":106228,"journal":{"name":"2022 IEEE International Symposium on Medical Measurements and Applications (MeMeA)","volume":"71 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114839387","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 : 2022-06-22DOI: 10.1109/MeMeA54994.2022.9856534
F. Gioia, A. L. Callara, Tobias Bruderer, Matyas Ripszam, F. Francesco, E. P. Scilingo, A. Greco
Emotional sweating occurs in response to affective stimuli like fear, anxiety, or stress and is more evident in specific parts of the body such as the palms, soles, and axillae. During emotional sweating, humans release many volatile organic compounds (VOCs) that could play a crucial role as possible com-municative signals of specific emotions. In this preliminary study, we investigated seven volatiles belonging to the chemical class of acids and released from the armpit as possible stress biomarkers. To this aim, we processed sweat VOCs and physiological stress correlates such as heart rate variability (HRV), electrodermal activity, and thermal imaging during a Stroop color-word test. Particularly, we modelled the variability of well-known stress markers extracted from the physiological signals as a function of the acid VOCs by means of LASSO regression. LASSO results revealed that the dodecanoic acid was the only selected regressor and it was able to significantly explain more than 64 % of the variance of both the mean temperature of the tip of the nose (p=0.018, R2=0.64) and of the mean HRV (p=0.011, R2=0.67). Although preliminary, our results suggest that dodecanoic acid could be a marker of the sympathetic nervous system response to stress stimuli, opening for the detection of new biomarkers of stress.
{"title":"Potential physiological stress biomarkers in human sweat","authors":"F. Gioia, A. L. Callara, Tobias Bruderer, Matyas Ripszam, F. Francesco, E. P. Scilingo, A. Greco","doi":"10.1109/MeMeA54994.2022.9856534","DOIUrl":"https://doi.org/10.1109/MeMeA54994.2022.9856534","url":null,"abstract":"Emotional sweating occurs in response to affective stimuli like fear, anxiety, or stress and is more evident in specific parts of the body such as the palms, soles, and axillae. During emotional sweating, humans release many volatile organic compounds (VOCs) that could play a crucial role as possible com-municative signals of specific emotions. In this preliminary study, we investigated seven volatiles belonging to the chemical class of acids and released from the armpit as possible stress biomarkers. To this aim, we processed sweat VOCs and physiological stress correlates such as heart rate variability (HRV), electrodermal activity, and thermal imaging during a Stroop color-word test. Particularly, we modelled the variability of well-known stress markers extracted from the physiological signals as a function of the acid VOCs by means of LASSO regression. LASSO results revealed that the dodecanoic acid was the only selected regressor and it was able to significantly explain more than 64 % of the variance of both the mean temperature of the tip of the nose (p=0.018, R2=0.64) and of the mean HRV (p=0.011, R2=0.67). Although preliminary, our results suggest that dodecanoic acid could be a marker of the sympathetic nervous system response to stress stimuli, opening for the detection of new biomarkers of stress.","PeriodicalId":106228,"journal":{"name":"2022 IEEE International Symposium on Medical Measurements and Applications (MeMeA)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129388019","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 : 2022-06-22DOI: 10.1109/MeMeA54994.2022.9856423
G. Campobello, Angelica Quercia, G. Gugliandolo, Antonino Segreto, E. Tatti, M. Ghilardi, G. Crupi, A. Quartarone, N. Donato
In this paper, we investigate performance of a re-cently proposed near-lossless compression algorithm specifically devised for multichannel electroencephalograph (EEG) signals. The algorithm exploits the fact that singular value decomposition (SVD) is usually performed on EEG signals for denoising and removing unwanted artifacts and that the same SVD can be used for compression purpose. In this paper, we derived an analytical expression for the expected compression ratio and an upper bound for the maximum distortion introduced by the algorithm after reconstruction. Moreover, performances of the algorithm have been investigated on an extended dataset containing real EEG signals related to subjects performing different sensorimotor tasks. Both analytical and experimental results reported in this paper show that the algorithm is able to attain a compression ratio proportional to the number of EEG channels by achieving a percentage root mean square distortion (PRD) in the order of 0.01 %. In particular, the achieved PRD is very low if compared with other state-of-the-art compression algorithms with similar complexity. Moreover, the algorithm allows the desired maximum absolute error to be fixed a priori. Therefore, we can consider this algorithm as an efficient tool for reducing the amount of memory necessary to record data and, at the same time, preserving actual clinical information of the signals besides compression.
{"title":"Theoretical and Experimental Investigation of an Efficient SVD-based Near-lossless Compression Algorithm for Multichannel EEG Signals","authors":"G. Campobello, Angelica Quercia, G. Gugliandolo, Antonino Segreto, E. Tatti, M. Ghilardi, G. Crupi, A. Quartarone, N. Donato","doi":"10.1109/MeMeA54994.2022.9856423","DOIUrl":"https://doi.org/10.1109/MeMeA54994.2022.9856423","url":null,"abstract":"In this paper, we investigate performance of a re-cently proposed near-lossless compression algorithm specifically devised for multichannel electroencephalograph (EEG) signals. The algorithm exploits the fact that singular value decomposition (SVD) is usually performed on EEG signals for denoising and removing unwanted artifacts and that the same SVD can be used for compression purpose. In this paper, we derived an analytical expression for the expected compression ratio and an upper bound for the maximum distortion introduced by the algorithm after reconstruction. Moreover, performances of the algorithm have been investigated on an extended dataset containing real EEG signals related to subjects performing different sensorimotor tasks. Both analytical and experimental results reported in this paper show that the algorithm is able to attain a compression ratio proportional to the number of EEG channels by achieving a percentage root mean square distortion (PRD) in the order of 0.01 %. In particular, the achieved PRD is very low if compared with other state-of-the-art compression algorithms with similar complexity. Moreover, the algorithm allows the desired maximum absolute error to be fixed a priori. Therefore, we can consider this algorithm as an efficient tool for reducing the amount of memory necessary to record data and, at the same time, preserving actual clinical information of the signals besides compression.","PeriodicalId":106228,"journal":{"name":"2022 IEEE International Symposium on Medical Measurements and Applications (MeMeA)","volume":"107 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134329510","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 : 2022-06-22DOI: 10.1109/MeMeA54994.2022.9856496
E. A. Nehary, S. Rajan
Ultrasound (US) imaging is an affordable, radiation-free screening that has been successfully used for early stage breast cancer screening. Deep learning-based classifiers are currently being used to classify breast cancer. Deep learning requires large amount of dataset for training. However, currently available databases of breast cancer US images are small and the images have tumors of different sizes. Therefore, the deep learning-based classifiers are unable to provide good generalization. To address these challenges, we propose a fusion of three models namely transfer learning, multi-scale and autoencoder. Transfer learning model is based on VGG16 and is used to overcome the issue of limited data. Convolutional autoencoders extract features that can represent even noisy images. We propose a novel multi-scale deep learning model to address learning of US images with tumors of various sizes and shapes. These three models are trained independently and then their classification outputs are fused using differential evolution (DE) algorithm to get the final classification results. The proposed novel fused ensemble of deep learning-based classifiers is evaluated using two publicly available US datasets. Transfer learning, autoencoder, and multi-scale models individually achieve an accuracy of 88%, 85%, and 89% respectively. The fusion of the outputs of the three models using DE algorithm provides a classification accuracy with an accuracy of 93%. The source code available at https://github.com/EbrahimAli1989/Breast-Cancer-classification-.git.
{"title":"Classification of Ultrasound Breast Images Using Fused Ensemble of Deep Learning Classifiers","authors":"E. A. Nehary, S. Rajan","doi":"10.1109/MeMeA54994.2022.9856496","DOIUrl":"https://doi.org/10.1109/MeMeA54994.2022.9856496","url":null,"abstract":"Ultrasound (US) imaging is an affordable, radiation-free screening that has been successfully used for early stage breast cancer screening. Deep learning-based classifiers are currently being used to classify breast cancer. Deep learning requires large amount of dataset for training. However, currently available databases of breast cancer US images are small and the images have tumors of different sizes. Therefore, the deep learning-based classifiers are unable to provide good generalization. To address these challenges, we propose a fusion of three models namely transfer learning, multi-scale and autoencoder. Transfer learning model is based on VGG16 and is used to overcome the issue of limited data. Convolutional autoencoders extract features that can represent even noisy images. We propose a novel multi-scale deep learning model to address learning of US images with tumors of various sizes and shapes. These three models are trained independently and then their classification outputs are fused using differential evolution (DE) algorithm to get the final classification results. The proposed novel fused ensemble of deep learning-based classifiers is evaluated using two publicly available US datasets. Transfer learning, autoencoder, and multi-scale models individually achieve an accuracy of 88%, 85%, and 89% respectively. The fusion of the outputs of the three models using DE algorithm provides a classification accuracy with an accuracy of 93%. The source code available at https://github.com/EbrahimAli1989/Breast-Cancer-classification-.git.","PeriodicalId":106228,"journal":{"name":"2022 IEEE International Symposium on Medical Measurements and Applications (MeMeA)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-06-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125996801","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：