Manne Hannula

Measurement of respiratory rate with high-resolution accelerometer and emfit pressure sensor

Respiratory rate is an essential parameter in the clinical monitoring of hospital patients. It can be measured in various ways, such as by recording chest movements, breathing flow or heart rate variations. Current sensor technology allows the development of new kinds of convenient and portable respiratory rate recorders, including smart shirts, which enable more efficient healthcare processes in hospitals. This study carried out respiratory rate measurements using a sensor belt with a high-resolution accelerometer (capacitive MEMS) and an EMFit (electret film) pressure sensor. Results obtained from tests on 10 subjects showed that both sensors are feasible for respiratory rate measurement; the reliability of the MEMS was 90%, while that of the EMFit was 90- 100%. In addition, the results showed that the location of the sensor module on the chest is important.

One-repetition maximum bench press performance estimated with a new accelerometer method.

Rontu, J-P, Hannula, MI, Leskinen, S, Linnamo, V, Salmi, JA. One-repetition maximum bench press performance estimated with a new accelerometer method. J Strength Cond Res 24(8): 2018-2025, 2010-The one repetition maximum (1RM) is an important method to measure muscular strength. The purpose of this study was to evaluate a new method to predict 1RM bench press performance from a submaximal lift. The developed method was evaluated by using different load levels (50, 60, 70, 80, and 90% of 1RM). The subjects were active floorball players (n = 22). The new method is based on the assumption that the estimation of 1RM can be calculated from the submaximal weight and the maximum acceleration of the submaximal weight during the lift. The submaximal bench press lift was recorded with a 3-axis accelerometer integrated to a wrist equipment and a data acquisition card. The maximum acceleration was calculated from the measurement data of the sensor and analyzed in personal computer with LabView-based software. The estimated 1RM results were compared with traditionally measured 1RM results of the subjects. An own estimation equation was developed for each load level, that is, 5 different estimation equations have been used based on the measured 1RM values of the subjects. The mean (±SD) of measured 1RM result was 69.86 (±15.72) kg. The mean of estimated 1RM values were 69.85-69.97 kg. The correlations between measured and estimated 1RM results were high (0.89-0.97; p < 0.001). The differences between the methods were very small (−0.11 to 0.01 kg) and were not significantly different from each other. The results of this study showed promising prediction accuracy for estimating bench press performance by performing just a single submaximal bench press lift. The estimation accuracy is competitive with other known estimation methods, at least with the current study population.

Comparison between artificial neural network and multilinear regression models in an evaluation of cognitive workload in a flight simulator

In this study, the performances of artificial neural network (ANN) analysis and multilinear regression (MLR) model-based estimation of heart rate were compared in an evaluation of individual cognitive workload. The data comprised electrocardiography (ECG) measurements and an evaluation of cognitive load that induces psychophysiological stress (PPS), collected from 14 interceptor fighter pilots during complex simulated F/A-18 Hornet air battles. In our data, the mean absolute error of the ANN estimate was 11.4 as a visual analog scale score, being 13-23% better than the mean absolute error of the MLR model in the estimation of cognitive workload.

Development and Evaluation of One Arm Electrode Based ECG Measurement System

In this work we have developed an electrocardiogram (ECG) measurement system, which measures ECG by two electrodes attached to the left arm of the subject. The measurement system consists of wrist and upper arm electrodes connected to an amplifier, which is further connected via data acquisition card to a laptop-computer. The computer included Labview® and Matlab® based integrated real-time analysis software, which reads, filters and analyzes the ECG signal. In the implementation of the system specific solutions to the amplifier and analysis software were built to be able to measure and interpret the very weak and noisy ECG signal originating from the electrodes. The measurement system was evaluated by comparing the heart rate interval data calculated from the new ECG measurement system to Polar S810i heart rate meter and Biopac M35 measurement system. The results showed that the new ECG measurement system yielded equal results compared to the reference measurements for subjects at rest. For moving subjects the reliability of the new measurement system reduced due to artefacts. The application experiences about the system were especially encouraging. The usability of the system was excellent due to simple one-arm electrode solution.

Neural network analysis of musculoskeletal responses to electrical AC-stimulus

Analyzing the human body by the application of alternating electrical currents is not a widely known method in medicine. In our research, we stimulated test persons by exposing them to different low frequencies and measured their responses to them. This method, known as FAM (Frequency Analysis Method), can be used to estimate the physiological condition of patients. In this study, we present a method of processing the results using neural networks. By producing user-friendly visual data, the processing method aids a physiotherapist in the interpretation of the results, resulting in a more reliable diagnosis.

Classification accuracy of a frequency analysis method: comparison between supervised SOM and KNN

In FAM measurement an alternative current stimulus at several frequencies is fed to the human body resulting in physiological responses whose thresholds are recorded. The idea of the FAM is to investigate the physiological properties of the human body by analyzing those thresholds. The basic objective is to make diagnostic classification on the basis of the measured threshold values. In this study properties of two methods, supervised SOM and kNN, are applied to the diagnostic classification task. The classification accuracy of those methods in FAM data analysis is defined and properties of the methods and the data are discussed. The classification accuracy of both methods was about 70% in classification to two classes and this result shows the supervised SOM has about the same performance in accuracy as the kNN has in the classification of the FAM data.

Validation of Individual Calibration Procedure in Prediction of One Repetition Maximum in Bench Press

In literature various studies about application of accelerometer sensors in analysis of weight lifting performance can be found. One specific application is prediction of one repetition maximum (1-RM) from submaximal lift, where the 1-RM is estimated on the basis of submaximal weight and acceleration values occuring during the lift. The prediction accuracy has been found to be good at a large subject group level. However, differences between individual subjects may result in inaccurate prediction in occasional cases. Therefore utilization of anthropometric data or information of weight lifting experience of the individual could improve the accuracy of the prediction. To evaluate this, in this study individual calibration of the prediction algorithm was evaluated in case of accelerometer based 1-RM prediction in bench press. The calibration procedure was tested with a total of 30 subjects (15 females, 15 males). The data analysis showed excellent correlation coefficient R of 0,99 (p<0,001) between the predicted 1-RM and the actual 1-RM in the validation phase of the study. Relative absolute error of 1-RM was 4,6 % for bench press. Without use individual calibration the correlation coefficient would have been 0,99 (p<0,001) and the relative absolute error would have been 9,2 %. The results clearly show the individual calibration increase significantly the 1-RM prediction accuracy.

Applicability of function-based analysis in retrospective data analysis of noise-induced hearing loss in the Finnish Defence Forces

In retrospective studies, analysis of statistical cross-sectional and longitudinal data is always a challenge. To obtain statistically reliable results, the quantity and quality of the data have to be carefully evaluated. In this study, application of specific functions that establish versatile evaluation of retrospective data in the time domain was studied in a case of retrospective analysis of noise-induced hearing loss (NIHL) among 1337 soldiers in the Finnish Defence Forces (FDF). The data included a number of audiogram (AG) variables that were recorded from subjects at varying intervals. Specific functions were used to present AG results according to one common continuous time scale, the age of the subjects. The aim of this study was to evaluate how much inaccuracy this kind of presentation would produce in analysis of the AG results. The inaccuracy was evaluated by comparing the median values of function-based values and the original data values according to the age of the subjects. The results of the study showed that the resulting error was less than 3 dB between the ages of 24–48 for all the AG variables for all the subjects in this data. This level of error is significantly less than the typical measurement resolution (5 dB) of AG, indicating that the proposed function-based analysis is appropriate for analysis of NIHL data.

Artificial Neural Network Analysis in the Evaluation of Ear Canal and Tympanic Membrane Properties from Acoustic Reflectometry Data

Acoustic reflectometry (AR) can be applied to detect middle ear effusion (MEE) in order to diagnose otitis media with effusion (OME). Natural variation in the anatomy of the ear canal and tympanic membrane affects the result of AR. In the present study the effect of the length of the ear canal and the tension of the tympanic membrane on the results of AR was modelled. Six plastic models of the ear canal and tympanic membrane were constructed, with unique canal lengths and unique tympanic membrane tensions. The plastic models were measured with AR and the resulting data were analyzed with an artificial neural network (ANN) method. The results indicate that, with help of the ANN, the length of the ear canal and the tension of the tympanic membrane can be identified from AR data; in the validation phase the ANN classified the different ear canal lengths correctly in all six cases, and in five of the six cases it correctly classified the tympanic membrane tenseness. These test results may be useful when developing the AR method for more accurate diagnostics of OME.

Prediction of One Repetition Maximum in Dumbbell Concentration Curl and Shoulder Press

In this study the prediction of maximum performance in dumbbell concentration curl and shoulder press was examined. The idea of the study was to predict the one repetition maximum (1-RM) with help of analysis of accelerations during a submaximal lift. In the study 30 gym trainees (all males, age 26±7 years, height 179±6 cm, weight 77±10 kg) performed dumbbell concentration curl and shoulder press exercises which all were accurately measured with help of a specific three-dimensional acceleration sensor based analysis system. The characteristics of the acceleration as a function of submaximal weight was analyzed and the prediction of one repetition maximum were evaluated. The data analysis resulted in 1-RM prediction mean absolute error of 2,8 kg for the dumbbell concentration curl and 12,3 kg for the shoulder press when using a simple linear prediction model. The detailed analysis showed that prediction mean absolute errors of 1,9 kg and 3,3 kg for dumbbell concentration curl and shoulder press would be achieved by adjusting the prediction model appropriately. The results show that it is possible predict the 1-RM with help of accelerometer from the submaximal performance in studied weight lifting exercises. However, the detailed prediction algorithm for 1-RM prediction requires further development.