Deep Learning (DL), a successful promising approach for discriminative and generative tasks, has recently proved its high potential in 2D medical imaging analysis; however, physiological data in the form of 1D signals have yet to be beneficially exploited from this novel approach to fulfil the desired medical tasks. Therefore, in this paper we survey the latest scientific research on deep learning in physiological signal data such as electromyogram (EMG), electrocardiogram (ECG), electroencephalogram (EEG), and electrooculogram (EOG). We found 147 papers published between January 2018 and October 2019 inclusive from various journals and publishers. The objective of this paper is to conduct a detailed study to comprehend, categorize, and compare the key parameters of the deep-learning approaches that have been used in physiological signal analysis for various medical applications. The key parameters of deep-learning approach that we review are the input data type, deep-learning task, deep-learning model, training architecture, and dataset sources. Those are the main key parameters that affect system performance. We taxonomize the research works using deep-learning method in physiological signal analysis based on: (1) physiological signal data perspective, such as data modality and medical application; and (2) deep-learning concept perspective such as training architecture and dataset sources.

BackgroundBlood pressure is a critical bio-signal and its importance has been increased with the aged society and the growth of cardiovascular disease population. However, most of hypertensive patients have been suffered the inconvenience in monitoring blood pressure in daily life because the measurement of the blood pressure depends on the cuff-based technique. Nowadays there are many trials to measure blood pressure without cuff, especially, photoplethysmography (PPG) based research is carried out in various ways.MethodsOur research is designed to hypothesis the relationship between vessel wall movement and pressure-flow relationship of PPG and to validate its appropriateness by experimental methods. PPG waveform is simplified by approximate model, and then it is analyzed as the velocity and the acceleration of blood flow using the derivatives of PPG. Finally, we develop pressure index (PI) as an estimation factor of blood pressure by combining of statistically significant segments of photoplethysmographic waveform.ResultsTwenty-five subjects were participated in the experiment. As a result of simulation, correlation coefficients between developed PI and blood pressure were represented with R = 0.818, R = 0.827 and R = 0.615 in systolic blood pressure, pulse pressure and mean arterial pressure, respectively, and both of result showed the meaningful statistically significance (P < 0.05).ConclusionsCurrent study can estimate only the relative variation of blood pressure but could not find the absolute pressure value. Moreover, proposed index has the limitation of diastolic pressure tracing. However, the result shows that the proposed PI is statistically significantly correlated with blood pressures, and it suggests that the proposed PI as a promising additional parameter for the cuff less blood pressure monitoring.

Post-stroke gait dysfunction occurs at a very high prevalence. A practical method to quantitatively analyze the characteristics of hemiparetic gait is needed in both clinical and community settings. This study developed a 10-channeled textile capacitive pressure sensing insole (TCPSI) with a real-time monitoring system and tested its performance through hemiparetic gait pattern analysis. Thirty-five subjects (18 hemiparetic, 17 healthy) walked down a 40-m long corridor at a comfortable speed while wearing TCPSI inside the shoe. For gait analysis, the percentage of the plantar pressure difference (PPD), the step count, the stride time, the coefficient of variation, and the phase coordination index (PCI) were used. The results of the stroke patients showed a threefold higher PPD, a higher step count (41.61 ± 10.7), a longer average stride time on the affected side, a lower mean plantar pressure on the affected side, higher plantar pressure in the toe area and the lateral side of the foot, and a threefold higher PCI (hemi: 19.50 ± 13.86%, healthy: 5.62 ± 5.05%) compared to healthy subjects. This study confirmed that TCPSI is a promising tool for distinguishing hemiparetic gait patterns and thus may be used as a wearable gait function evaluation tool, the external feedback gait training device, and a simple gait pattern analyzer for both hemiparetic patients and healthy individuals.

The electro-conductive fabric (e-textile or e-fabric) as an electrode for ECG measurement is one of the best application for ubiquitous healthcare system. However, it is difficult to measure the bio-signal due to its sensitivity variation caused by impedance change, especially by motion of the subject. In this paper, adaptive motion artifacts reduction using motion information from 3-axis accelerometer is proposed and analyzed in quantitative manner.

Spatiotemporal analysis of gait pattern is meaningful in diagnosing and prognosing foot and lower extremity musculoskeletal pathologies. Wearable smart sensors enable continuous real-time monitoring of gait, during daily life, without visiting clinics and the use of costly equipment. The purpose of this study was to develop a light-weight, durable, wireless, soft-material-based smart insole (SMSI) and examine its range of feasibility for real-time gait pattern analysis. A total of fifteen healthy adults (male: 10, female: 5, age 25.1 ± 2.64) were recruited for this study. Performance evaluation of the developed insole sensor was first executed by comparing the signal accuracy level between the SMSI and an F-scan. Gait data were simultaneously collected by two sensors for 3 min, on a treadmill, at a fixed speed. Each participant walked for four times, randomly, at the speed of 1.5 km/h (C1), 2.5 km/h (C2), 3.5 km/h (C3), and 4.5 km/h (C4). Step count from the two sensors resulted in 100% correlation in all four gait speed conditions (C1: 89 ± 7.4, C2: 113 ± 6.24, C3: 141 ± 9.74, and C4: 163 ± 7.38 steps). Stride-time was concurrently determined and R2 values showed a high correlation between the two sensors, in both feet (R2 ≥ 0.90, p < 0.05). Bilateral gait coordination analysis using phase coordination index (PCI) was performed to test clinical feasibility. PCI values of the SMSI resulted in 1.75 ± 0.80% (C1), 1.72 ± 0.81% (C2), 1.72 ± 0.79% (C3), and 1.73 ± 0.80% (C4), and those of the F-scan resulted in 1.66 ± 0.66%, 1.70 ± 0.66%, 1.67 ± 0.62%, and 1.70 ± 0.62%, respectively, showing the presence of a high correlation (R2 ≥ 0.94, p < 0.05). The insole developed in this study was found to have an equivalent performance to commercial sensors, and thus, can be used not only for future sensor-based monitoring device development studies but also in clinical setting for patient gait evaluations.

We proposed non-contacting respiration signal monitoring system for sleep apnea syndrome. Experiments were conducted by emitting 40 kHz ultrasound beam, which is set tone burst mode by 1 ms period to a subject chest. Normal respiration condition and a simulated sleep apnea syndrome condition were measured while subjects were holding breath. To obtain the actual respiration signal from the raw signal, ultrasound attenuation characteristics were considered. The Doppler ultrasound signal was detectable once the received signal obtained by demodulation circuits passed through a low pass filter (LPF). The signal's ripples were eliminated by moving average method and the signal's peaks were detected by phase portrait reconstruction method to measure the respiration rate.

No abstract

In the Korean construction industry, legal and institutional safety management improvements are continually being pursued. However, there was a 4.5% increase in the number of workers’ deaths at construction sites in 2017 compared to the previous year. Failure to wear safety helmets seems to be one of the major causes of the increase in accidents, and so it is necessary to develop technology to monitor whether or not safety helmets are being used. However, the approaches employed in existing technical studies on this issue have mainly involved the use of chinstrap sensors and have been limited to the problem of whether or not safety helmets are being worn. Meanwhile, improper wearing, such as when the chinstrap and harness fixing of the safety helmet are not properly tightened, has not been monitored. To remedy this shortcoming, a sensing safety helmet with a three-axis accelerometer sensor attached was developed in this study. Experiments were performed in which the sensing data were classified whether the safety helmet was being worn properly, not worn, or worn improperly during construction workers’ activities. The results verified that it is possible to differentiate among wearing status of the proposed safety helmet with a high accuracy of 97.0%.