مجله دانشکده پزشکی اصفهان
پیاپی 174 (1390)

Mathematical models can provide insights into cancerous cell's growth and their interaction with healthy cells, immune cells and chemotherapeutic drugs that are used in cancer therapy. Moreover, mathematical models have been developed to aid in describing the mechanisms of cytotoxic drug availability and effects on healthy cell populations. Finding a desirable treatment protocol for patients is one of the most important objectives of mathematical modeling. The convention in designing the optimal chemotherapy strategies is making use of classical optimal control theory. In this study, a new mathematical model is developed for analysis of dynamics of cancerous cells in different phases of cell cycle, immune cells, chemotherapeutic drug concentration and toxicity. Finally, the theorem of Lyapunov stability is applied to design an optimal treatment protocol. The results of simulation shows that after seven times of chemotherapy during fifty days, all the cancerous cells have been diminished, and the disease will remain in this desirable state up to six months. In this research, a new mathematical model for describing the dynamics of a cancerous system have been proposed, and an optimal treatment protocol has been designed applying Lyapunov stability theorem. The population of cancerous cells has been decreased to zero and will be remained in this state for six months. By Applying vaccine therapy cancerous cells’ growth could be prevented. Vaccine therapy changes the parameters of the system and stabilizes tumor free equilibrium point.

Myocardial ischemia is caused by a lack of sufficient blood flow to the contractile cells and may lead to myocardial infarction with its severe sequel of heart failure, arrhythmias, and death. Therefore, its early diagnosis and treatment is of great importance. During past several years, several strategies have been proposed for automatic detection of ischemic cardiac beats using the ST-T complex of the electrocardiogram (ECG) which is recorded during long-term Holter monitoring. However, automatic detection of ischemic subjects using short-term analysis of the ECG is an open problem. In this paper, we presented a new method for automated detection of ischemic patients. The study was conducted based on short-term analysis of the ECG, time-frequency analysis of heart-rate variability (HRV) and recurrent log linearized Gaussian mixture neural network (RLLGMN) as the classifier. For this purpose the feature vector was extracted from the energy of HRV was obtained from 98 subjects (50 healthy and 48 patients) at different frequency bands during different autonomic tests (i.e. controlled normal breathing, controlled deep breathing and active transition tests) and classified using a probabilistic neural network. The results showed that a correct classification rate of 82% for the healthy subjects and 86% for the ischemic subjects was achieved using the proposed method. The results indicated that HRV signals and recurrent probabilistic neural network can be used as a noninvasive method for identifying ischemia. The proposed method is simple, fast and noninvasive, and does not require the long-term recording of ECG signals as well as exercise treadmill test.

Chronic obstructive pulmonary disease (COPD) is one of the most prevalent and dangerous pulmonary diseases in the world. It is forecasted that COPD will be the third deadly disease in the future. Therefore, developing non-invasive methods for diagnosis of the disease would be helpful for physicians and patients. Based on clinical investigations and spirometry tests, ten adult patients with COPD (6 male and 4 female) with mean age of 49.8 years were enrolled as the case group. In addition, ten age and sex-matched healthy, non-COPD individuals (6 male and 4 female) with mean age of 45.4 years were recruited as the controls. Lung CT-scan images of the subjects were processed and analyzed by a computer to find a relationship. The elasticity of lung parenchyma variation was obtained with digital image processing. The normalized average of this pattern was found to be 21.6% in patients and 40.7% in controls. In addition, normalized mean value of Hounsfield unit variations in square 10 pixel × 10 pixel windows in the expiratory images were calculated as a parameter of air-trapping in COPD. Differences between the groups were shown by student t-test (P < 0.05). This study showed that the variation of lung parenchyma elasticity and Hounsfield units are found by processing and analysis of the full inspiration and expiration images. These factors can be used as criteria in diagnosis of COPD. Moreover, the severity of the disease can be presented by the proposed method.

Since muscle force depends on various factors, presenting explicit equations for predicting it is difficult. Due to the importance of these equations in different musculoskeletal problems and analysis of motion, proposing solutions for estimating muscle force is indisputable. Although numerous relations have been suggested, the complexity of relations between the caused musculotendon force and the effective factors leads to a lot of problems in suggesting efficient comprehensive computational equations. Thus, the previous proposed relations either do not have universality for estimating muscle force, or are too nonlinear, massive and inefficient. In this study, after reviewing previous mathematical relations for predicting muscle force, an artificial neural network was implemented to predict force in musculotendon actuators. The present study provided an appropriate and computationally efficient mathematical model with limited parameters. The model can be used to determine the forces generated by musculotendon actuators in various lengths and contraction velocities. Utilizing a neural network lets us estimate muscle force quicker and more accurately. This could fundamentally increase the computational efficiency in different musculoskeletal problems.

Designing brain-computer interface (BCI) systems is one of the concerns of people today. These systems operate by brain signals and so far, much research has been done in this regard. The most conventional systems are based on mental task signals. In the design of BCI systems based on mental activity, selecting a feature space with higher resolution and less processing time is important. In this study, Anderson mental task signals, a known and available database in such systems, were used. According to the nonlinear and non-stationary properties of electroencephalogram (EEG) signals, this study tried to review and analyze new empirical mode decomposition (EMD) algorithms, as well as conventional and successful methods such as autoregressive (AR) spectrum and entropy, for discrimination of mental task signals. EMD algorithm is compatible with nonlinear and non-stationary properties of EEG signals. Therefore, using an EMD algorithm along with the concept of entropy for modeling complexity values and AR spectrum, as a significant function in the frequency domain would provide great discrimination. Application of EMD algorithm and its parallel schemes (EMD entropy) would result in a feature vector with less dimensions requiring less than 2 seconds to extract features. Thus, such combination would require a maximum of 0.1 seconds to separate 10-second signals which can be beneficial in real-time BCI systems.

Dynamic behaviors of probe tip and nanoparticles have been investigated by modeling the manipulation of nanoparticles in the air using atomic force microscopy (AFM) as a nano manipulator. This study evaluated the manipulation of submerged nanoparticles in liquid environment. The artificial nanoparticle manipulation was analyzed by theoretical analysis of forces in liquids and dynamics of spherical nanoparticle pushing. Hydrodynamic drag force and surface tension of the liquid were then calculated and used in the new dynamic modeling. The problem was simulated for a gold nanoparticle on a silicon substrate in water. The results showed that the required manipulation force and time for nanoparticle sliding and rolling increased by respectively 7% and 3% in water as compared to the air. Moreover, for various submerged lengths of the cantilever in water, the critical values related to sliding and rolling were delayed for 9% and 10.5%, respectively. The required critical force and time for the manipulation in water showed a little increase over the existing values for air.

Sliding mode control is an effective method for controlling the neuromusculoskeletal systems. A major problem of conventional sliding mode control is exponential convergence of the tracking errors. To solve this problem, researchers proposed a robust control strategy called terminal sliding mode control. The main advantages of the terminal sliding mode control are not only robustness against uncertainties and external disturbances but also finite time convergence of tracking errors.

In this paper, we propose a decentralized control strategy which is based on terminal sliding mode control, for control of the ankle joint in paraplegic subjects using functional electrical stimulation. Agonist-antagonist co-activation is used to control the ankle movement.

The proposed control strategy was employed for control of ankle joint in three paraplegic subjects. The control task was to determine the stimulation pattern in order to converge the ankle movement trajectory to the desired trajectory. The experimental results on three paraplegic subjects showed that the proposed controller provided an excellent tracking control of reference trajectories. It could also generate control signals to compensate the effects of muscle fatigue.

The results of this study showed the proposed control strategy as an effective approach for controlling movements in paraplegic subjects using functional electrical stimulation.

In this study, we tried to choose the proper material to use on a magnetic resonance imaging (MRI)-compatible robot. The robot is to take soft tissue samples in an image guided process. Since this robot is made of MRI-compatible material, its links are invisible in the MR images. Finding such markers would enable us to distinguish the position of links in the images and as a result in the real world. Four materials compatible with magnetic environments, including liquid paraffin, nitroglycerin, vitamin E pills, and silicon oil, were tested to find the most appropriate one to be used as the marker. In addition to compatibility with MRI environment, the material had to have suitable contrast to avoid blurry images. The tests were performed in a Philips MRI system at 1 and 1.5 Tesla. Liquid paraffin, nitroglycerin, vitamin E pills, and silicon oil scored 50, 62.35, 45.6, and 53.15, respectively. Most of the evaluated factors were satisfied by nitroglycerin. After all tests and considerations, nitroglycerin was selected as the best material.

The purpose of this study was to investigate the performance of SVMs in predicting the drug resistance of HIV, using amino acids sequence analysis. support vector machines were used in this study. LIBSVM software was used to train and test SVMs. This paper presents the performance of SVMs in predicting the drug resistance of HIV, using amino acids sequence analysis. The results of four algorithms (HIVDB, ANRS, REGA and VGI) have been developed to interpret results of HIV genotypic resistance tests, are explored. The most efficient method for each drug is determined. Also, it is indicated that SVMs are a highly successful classifier with accuracy between 86.27- 98.77, for predicting of HIV drug resistance. Using SVMs for HIVDB method results, has best performance for APV, NFV, ABC, AZT, D4T, DDI, TDF and DLV drugs. Using SVMs for ANRS method results has best performance for IDV, 3TC, TDF, EFV and NVP drugs. Using SVMs for REGA method results has best performance for LPV and AZT drugs and using SVMs for VGI method results has best performance for IDV, LPV,RTV, SQV and DDI. The results show that SVMs are a highly successful classifier for predicting of HIV drug resistance. Before starting treatment with each drug, one can determine HIV drug resistance with machine learning methods.