فصلنامه مهندسی پزشکی زیستی
سال چهاردهم شماره 3 (پاییز 1399)

Back pain is a common medical problem. There is no clear cause for the back pain problem so far, but in most cases, spinal instability can be noted. Lumbar spine fixation is performed to treat the problems of low back pain. Spinal fixation can be done with or without surgery. One of the surgical methods is the use of spinal screws in which the strength and stability of the screw are of great importance. The strength and stability of the screw in the bone reduces the time and cost of treatment, reduces the amount of bleeding and accelerates the patient's treatment. In this study, screws were inserted using a digital torque meter. An impact was applied using an impact hammer and resonated sound was recorded using a microphone. The vibration mode of the screw was obtained by processing the signal generated by MATLAB R2017 software and plotting the fast Fourier transform. Finally, tensile test was performed to obtain the ultimate pull-out force. The innovation of this study was to use modal analysis method and to correlate its results with that of the ultimate pull-out force and peak insertion torque. In this study, five screws with different screw depth, and screw thread crest thickness were examined. Also, the effect of self-tapping was investigated. The peak insertion torque, ultimate pull-out strength and natural frequency occurred at 182 Nm, 992 N and 1916 Hz, respectively, for the cylindrical pedicle screw. By comparing the obtained data, results showed a linear relationship between insertion torque and pull-out force of the screws. Due to the lack of significant difference between natural frequency and pull-out force of the self-drilling and non-self-drilling tip screws (comparing between screws number 3 and 4 and between 1 and 5), the use of self-tapping screws can be advantageous. The trend of the dependent parameters in all three methods i.e. insertion torque, pull-out force and natural frequency are the same, indicating the non-destructive advantage of modal analysis in in-vivo surgical application.

In recent years, driver fatigue has become one of the major causes of road accidents, and many studies have been conducted to analyze driver fatigue. EEG signals are considered the most reliable method for measuring driver fatigue because of the non-invasive nature. Manual interpretation of EEG signals for detection of driver fatigue is impossible, so an automatic detection of driver fatigue from EEG signals should be provided. One of the problems regarding the automatic detection of driver fatigue is extraction and selection of discriminative features witch generally leads to computational complexity. This paper prepares a new approach to automatic classifying 2 stages of driver fatigue from 6 active regions of EEG signals. In the proposed method, directly apply the raw EEG signal to convolutional neural network-long short time memory (CNN-LSTM) network, without involving feature extraction/selection. This is a challenging process in previous literature. The proposed network architecture includes 7 convolutional layers with 3 LSTM layers followed by 2 fully connected layers. The LSTM network in a fusion with the CNN network has been used to increase stability and reduce oscillation. The simulation results of the proposed method for classifying 2 stages of driver fatigue for 6 active regions A, B, C, D, E (based single-channel) and F show the accuracy of 99.23%, 97.55%, 98.00%, 97.26%, 98.78%, 93.77% and Cohen’s Kappa coefficient of 0.98, 0.96, 0.97, 0.96, 0.98 and 0.92 respectively. Furthermore, comparing the obtained results with the previous methods reveals the performance improvement of the proposed driver fatigue detection in terms of accuracy. According to the high accuracy of the proposed single-channel (region E) method, it can be used for the design of automatic detection of driver fatigue systems with high speed and accuracy.

In modern prostheses, accurate processing of surface electromyogram (sEMG) signals has a significant effect on optimal muscle control. Although these signals are useful for diagnosing neuromuscular diseases, controlling prosthetic devices and detecting hand movements, non-robustness of EMG signal-based recognition will give rise to various movement disorders. In this paper, we present an optimal approach to classify EMG signals for hand gesture and movement recognition, whose purpose is to be used as an efficient method of diagnosing neuromuscular diseases, determining the type of treatment and physiotherapy. The main assumption of this study is to improve the accuracy of recognition and therefore, we proposed a novel hand gesture and movement recognition model consists of three steps: (1) EMG signal features extraction based on time-frequency domain and fractal dimension features; (2) feature selection by soft ensembling of three procedures in which includes two sample T-tests, entropy and common wrapper feature reduction, and (3) classification based on kernel parameters optimization of SVM classifier by using Gases Brownian Motion Optimization (GBMO) algorithm. Two UC2018 DualMyo and UCI datasets have been considered to evaluate the proposed model. The first dataset is used to classify eight hand gestures and the second dataset is employed for the classification of six types of movement. The experiment results and statistical tests reveal that the designed approach has desirable performance with an average accuracy of above 98% in both datasets. Contrary to similar methods that perform classifications in finite classes with high error rates, the integrated method has satisfactory accuracy, robustness and reliability. Not only the proposed method contributes to the design of prostheses, but also provides effective outcomes for rehabilitation applications and clinical diagnosis processes.

Using a smart method to automatically detect different stages of epilepsy in medical applications, to reduce the workload of physicians in analyzing epilepsy data by visual inspection is one of the major challenges in recent years. One of the problems of automatic identification of different stages of epilepsy is extraction of desirable features which can make the most distinction between different stages of epilepsy. The process of finding the proper features is generally time consuming. This study presents a new approach for the automatic identification of different epileptic stages. In this paper, a sparse represantion-based classification (SRC) with proposed dictionary learning is used to automatically identify the different stages of epilepsy using the EEG signal. The proposed method achieves 100% accuracy, sensitivity and specificity in 8 out of 9 scenarios. Also the proposed algorithm is resistant to Gaussian noise up to 0 decibels. The results show that using the proposed algorithm to identify different epileptic stages has a higher success rate than other similar methods.

The brain electrical signal has been widely used in clinical and academic research, due to its ease of recording, non-invasiveness, and precision. One of the applications can be emotion recognition from the brain's electrical signal. Generally, two types of parameters (Valence and Arousal) are used to determine the type of emotion, which, in turn, indicate "positive or negative" and "level of extroversion or excitement" for a specific emotion. The significance of emotion is determined by the effects of this phenomenon on daily tasks, especially in cases where the person is confronted with activities that require careful attention and concentration. In the emotion recognition problem, firstly, using proper emotion stimuli, different emotions are created for the subjects under study and the brain signals corresponding to each stimulus are recorded. The two main steps for solving the emotion recognition problem are extracting suitable features and using appropriate classification or regression methods. In previous studies, different visual and auditory have been used, and various linear and nonlinear features and classifiers have been investigated. In this paper, the main goal was the improvement of linear regression algorithms to estimate the criteria for recognizing human emotions more efficiently. For this purpose, we proposed a new algorithm that uses the sparseness of the mixing vector along with the linear regression cost function. The effectiveness of the proposed algorithm on simulated data has been investigated and its superiority to linear regression algorithms such as PLS, LASSO,, SOPLS and Ridge was shown. Also, to apply the proposed algorithm on EEG data corresponding to emotion recognition, the DEAP dataset was used, and the AR coefficients were extracted from the EEG signals. The results obtained from the proposed algorithm were compared with those of the other linear regression algorithms, which in total showed the relative superiority of the proposed method.

Functional Magnetic Resonance Imaging (fMRI) is a non-invasive neuroimaging technique for analyzing the brain functions through low-frequency fluctuations called the Blood-Oxygen-Level-Dependent (BOLD) signals. Measurement of the functional connectivity in brain networks is usually done by the fMRI time-series through Pearson Correlation Coefficients (PCC). As the PCC shows linear dependencies, in this study, non-linear relationships in the fMRI signals of the patients with Alzheimer's Disease (AD) were investigated using the kernel trick method. Kernel trick approach maps the input information into a higher dimension space and implementing the linear calculations in a new space that is proportionate to the non-linear relationships in the primary space. After generating the weighted undirected brain graphs based on the Automated Anatomical Labeling (AAL) atlas, different kernel functions with different parameters were applied. Then the graph global measures including degree, strength, small-worldness, modularity, and efficiencies features were computed and the non-parametric permutation test was performed. According to the results, the kernel trick method showed more significant differences with AD and healthy subjects in comparison with the simple PCC and it could be because of the non-linear correlations that are not captured by the PCC. Among different kernel functions, the Polynomial function had the best performance. Applying this kernel, the classification was done by the Support Vector Machine (SVM) classifier. The achieved accuracy was equal to 98.68±0.79%. The Occipital and Temporal lobes and also the Default Mode Network (DMN) were analyzed and the kernel trick method showed more significant differences in all of them. It is worthwhile to mention that the right and left Angular area of DMN showed no significant changes in none of the methods and it could be concluded that the AD does not affect this area effectively.

Transcranial direct current stimulation (tDCS) is the most-used non-invasive brain stimulation method. However, the main challenge in tDCS studies is its heterogeneity and large inter-individual variability in response. Brain anatomy, that varies from person to person, can change electric field distribution patterns in the brain and should be considered as a source of variation. Previous findings support that tDCS-induced EFs affect brain activity and ultimately change behavioral outcomes. Nonetheless, the exact relationship between EFs and brain activity alterations has not yet been investigated. In this randomized double-blinded sham-controlled crossover study, 14 subjects with methamphetamine use disorders were recruited and tDCS with 2 mA current intensity was applied over the dorsolateral prefrontal cortex. Each subject participated in two sessions for sham or real stimulation with at least a 1-week washout period. In each session, structural and functional MRI during a cue-induced craving task were collected immediately before and after tDCS. Individualized computational head models were simulated based on structural MR images and finite element methods. Group-level analysis of the models showed inter-individual variability across the subjects with maximum electric field intensity in frontal pole (0.3424±0.07). Furthermore, functional data, based on a drug minus neutral contrast, showed that real versus sham stimulation decreased brain activity in superior temporal gyrus and posterior cingulate cortex (p < 0.001). However, we did not find a significant correlation between induced EFs and brain activity alterations. In sum, in this study, we suggested a pipeline for integrating electric fields with functional neuroimaging data to bring new insights into the tDCS mechanism of action and future studies are required to establish, or to refute, this conclusion.