|
|
|
 |
Search published articles |
 |
|
Behnam Solatinia, Amirhesan Yahyapour,
Volume 0, Issue 0 (2-2025)
Abstract
Introduction: Accurate and early detection of brain tumors from MRI images plays a vital role in improving prognosis and treatment planning for patients. However, manual interpretation of MRI images is time-consuming and prone to human error. This study aims to present and evaluate an automated and efficient method based on the YOLOv8 algorithm and deep learning for detecting brain tumors in MRI images. Materials and Methods: In this study, the YOLOv8 model was trained using a dataset of 500 labeled brain tumor MRI images. The model's performance on the evaluation and test datasets was assessed using the metrics of Precision, Recall, and AP50. The model was trained in the Google Colab environment using a Tesla T4 GPU and over 100 epochs. Results: The results showed that the trained YOLOv8 algorithm achieved high accuracy and speed in detecting brain tumors. The obtained AP50 values on the evaluation (94.5) and test (94.6) datasets, along with high Precision and Recall values, were evidence of the model's strong and stable performance in tumor detection. The qualitative examination of sample images also confirmed the high accuracy of the algorithm in locating and detecting tumors. Furthermore, the short training time and high detection speed were other advantages of the YOLOv8 algorithm in this study. Conclusion: The present study demonstrates that the YOLOv8 algorithm has significant potential for the automatic and efficient detection of brain tumors in MRI images. The appropriate balance between accuracy and speed, as well as the good generalizability of the model, makes YOLOv8 a valuable auxiliary tool for radiologists achieving faster and more accurate detection of brain tumors. Future studies focus on increasing the volume of training data, improving the model architecture, and clinically evaluating the proposed approach.
Zahrasadat Hashemi, Arezou Eshaghabadi, Fatemeh Alipour, Maryam Jafarian, Sayed Mostafa Modarres Mousavi,
Volume 4, Issue 3 (9-2016)
Abstract
Introduction: An effort to establish phylogenetic values for the major gamma-Aminobutyric acid A (GABAA) receptor subunit mRNAs α, β, γ, δ, ε, θ, π could be improved our knowledge on their classification and function. In addition, the similarities and divergences between different species can be important to determine the function of these receptors. Materials and Methods: After alignment of mRNA complete gene sequences of GABAA subunits in homosapinse and rattus norvegicus species, the phylogenetic tree were constructed with CLC Main Workbench 5.5 software. Results: The results revealed highly similarities between GABAA subunits of homosapience and rattus norvegicus. In addition, these findings illustrated some divergences between β1, β2, β3, γ1, γ2, ε and θ subunits with other subunits. Conclusion: The similarities and divergences among various GABAA subunits may be an important cause of different distribution and function of GABAA subunits in different region of the central nervous system.
Maryam Tohidi-Moghaddam, Sajjad Zabbah, Reza Ebrahimpour,
Volume 4, Issue 4 (12-2016)
Abstract
Introduction: Most decisions are based on the accumulation of discrete pieces of evidence. This evidence has usually been separated with the various intervals. Indeed, how the brain gathers and combines distinct pieces of information received at different times is need to be clarified. In order to investigate the kinship between brain function and human behavior, the behavioral experimental studies could be designed. Previous studies demonstrated that subjects gather and effectively combine discrete evidence to improve their accuracy. In addition, it has been shown that the latest information has a larger influence on decisions. However, it remains unclear that why this larger influence of the later pulses occurs and what can affect this influence. Materials and Methods: Dealing with these questions a perceptual decision-making task has been implemented by the psychophysics’ toolbox in MATLAB. Subjects, during the task, were instructed to report the direction of motion in a noisy random dot stimulus with certain keys. Stimuli were presented in continuous (one pulse) or discrete (two continuousness pulses separated with different intervals) form. Each of these two forms of stimuli was presented randomly during each session. Each session has been included 300 trials. Each subject has done 3600 trials. Data have been analyzed by regression models. Results: We observed that in double-pulse trials, the strength of the second pulse was more crucial in the accuracy of responses compared to the first pulse. In addition, this accuracy was dependent on the differences between the strength of the first and the last pulses. Conclusion: These findings suggest that a key factor which affects the importance of pulses is the strength of the previous pulse. As the difference between the motion strength increases, the effect of the second pulse on choice accuracy enhanced.
Arash Alaeddini, Fatemeh Alipour, Zahrasadat Hashemi, Sayed Mostafa Modarres Mousavi,
Volume 5, Issue 2 (4-2017)
Abstract
Introduction: Neurons secreting gamma aminobutyric acid (GABA), an inhibitory neurotransmitter, as their primary neurotransmitter are named GABAergic neurons. Phylogenetics based on sequence data provides more accurate descriptions of patterns of relatedness. Materials and Methods: After alignment of mRNA complete gene sequences of alpha GABAA subunits in homosapiens and rattus norvegicus species, the phylogenetic tree were constructed with CLC Main Workbench 5.5 software. Results: The findings revealed 100 percent similarities between alpha 1, alpha 3, and alpha 6 subunits of GABAA receptor in homosapience and rattus norvegicus. Furthermore, the highest rate of divergences observed between alpha 1 subunit with alpha 3, alpha 4, alpha 5, and alpha 6 subunits in both species. Conclusion: The highest similarities among alpha subunits of GABAA in human and rat suggest the accuracy of rat models for experimental studies on inhibitory neurotransmitters in the central nervous system.
Fatemeh Fallah, Reza Ebrahimpour,
Volume 6, Issue 2 (4-2018)
Abstract
Introduction: Human visual system is able to recognize the objects relevant information in natural images rapidly and efficiently. In the recognition process, an object can belong to different levels of abs traction (superordinate, basic, and subordinate) in a hierarchical s tructure. However, it remains unclear whether different ques tions at levels of object categorization for identical s timulus create different activation responses in the brain or not. Materials and Methods: In order to inves tigate the relation between brain function and human behavior, three behavioral experimental s tudies have been designed with help of psychophysics’ toolbox in MATLAB R2015a. During these experiments, the participants asked to record animate, face, and animal face images as target images respect to the superordinate, basic, and subordinate levels, respectively. The experiments include seven blocks of 96 trials in superordinate (four blocks), basic (two blocks), and subordinate (one block) levels. Totally each subject has done 672 trials. Results: We observed that subjects’ reaction time were task dependent for the same images in contras t to previous s tudies. That is, images in the superordinate level were observed in the early component of reaction time whereas basic and subordinate levels emerged relatively late. In all levels, only a set of 48 target images (animal face) was analyzed. These target images were randomly mixed with other ones. Conclusion: The results showed that superordinate level is well separated from the other two levels. In other words, this level needs more general information for object recognition process than other levels. These findings sugges t that categorization of objects at different levels has done by three dis tinct neuronal circuits. Moreover, these results indicate that there are some top-down signals which change the information processing path respect to the ques tions.
Seyedeh Mahboobe Seyed Abbasi, Seyed Salman Zakariaee, Abbas Rahimiforoushani,
Volume 6, Issue 3 (7-2018)
Abstract
Introduction: The hemodynamic response function (HRF), reflecting cerebral blood flow in response to neural activity, plays a crucial role in the analysis of the brain data obtained by functional magnetic resonance imaging (fMRI). In this study, a comparison of two statistical models was performed to evaluate HRF for block design. Materials and Methods: fMRI data from 3 patients with brain tumor were taken using a 3 Tesla scanner. Analysis of fMRI data was performed by the SPM12 toolbox in MATLAB software. The AIC, SBC and MSE indices were used to select the most convenient HRF mode. Results: Based on the simulation data, HRF estimated by canonical HRF model plus time derivations (TD) model was more consistent with simulated HRF. These models were evaluated on real data. The MSE, AIC and SBC indices were obtained for TD-logistic model (IL) models (for TD and logistic IL models; 0.052 /, 1235.1, 1223.9 and 0.068 / -1091.5 / - 1049.2, respectively). Based on the average values of T, W, H and model selection indicators, IL model for estimating HRF in healthy regions of the brain and brain tumor is a more appropriate approach. Conclusion: The results of the present study can be helpful for the evaluation and diagnosis of HRF in high-metabolism points. Using the IL model to estimate HRF in the block design may lead to a better estimation of HRF and thus maintaining patient health and quality of life after surgical treatment and non-surgical medical procedures.
Mahdieh Ghasemi, Ali Foroutannia,
Volume 7, Issue 1 (1-2019)
Abstract
Introduction: In the recent years, neuroimaging research on functional Magnetic Resonance Imaging (fMRI) is used in many pathological and mental conditions. The analysis of alterations in the resting state networks (RSN) is an important method for the scrupulous understanding of the function and connectivity changes of the disease in order to provide new diagnostic and therapeutic approaches. In this paper, we studied the resting-state functional MRI (Rs-fMRI) data in Parkinson’s disease (PD) to explore the complex disruption in the RSNs and the functional interactions between them. Materials and Methods: A total Rs-fMRI data of 10 Parkinsonism and 10 healthy people in the 3T-MRI system were considered. Probabilistic independent component analysis (PICA) was used to extract network components. RSNs were identified using spatial correlation with a rest reference template network. Dual regression and randomize technique calculated individual differences between the groups. Results: Group component maps resulted in some main clusters of RSN that significantly overlapped with the reference network, such as the visual cortex, salience network, and supplementary motor area. Individual differences between RSN maps identified temporal, salience and cingulate networks as the main clusters. Conclusion: Most of the previous studies investigated the functional connectivity alterations in PD by seed-based analysis. Here, we employed the data-driven approach based on group PICA to extract and evaluate RSN changes in all related neural networks. Our finding indicates that changes of the functional architecture of the RSNs are associated with PD.
Elham Pourafrouz, Saeed Setayeshi, Iman Allah Bigdeli, Mir Mohsen Pedram,
Volume 9, Issue 2 (3-2021)
Abstract
Introduction: Artificial intelligence researchers are trying to implement human intelligence on the machine. This study aimed to develop an appropriate predictive computer model to evaluate the effectiveness of mindfulness-based cognitive therapy on irritability. Materials and Methods: The design of the present study is quasi-experimental with a pre-test and post-test method. 135 individuals who referred to Khane Mehr counseling center in Mashhad and participated in an 8-session mindfulness-based cognitive therapy (MBCT) course were included in this study. Totally, 11 MBCT courses were held and 10 to 14 people participated in each course. Participants completed the irritability questionnaire (Pourafrouz & et al.) at two stages (before treatment and after treatment). In order to examine the differences from pre-test to post-test in this research, the variance analysis of repeated measures was used. Results: There was a significant difference between pre-test and post-test irritability scores. The effect of mindfulness was 83%. To develop the prediction model, three Bayesian, regression, and neural network models were compared. The Bayesian model, with 93% accuracy test data, was considered the most appropriate model. Moreover, the Bayesian models with input and output clustering (85.7%), the Bayesian with classification (71.49%), and the sequential neural network (64.29%) were identified as suitable models to predict the effectiveness of 8-session mindfulness courses on reducing irritability. The Bayesian model with output clustering, one-output regression, and the Convulsions Neural Network did not have sufficient predictive accuracy for the effectiveness of mindfulness. Conclusion: Using cognitive modeling, we can predict the efficacy of mindfulness-based cognitive therapy on irritability.
Mojtaba Mohammadpoor, Atefe Alizadeh,
Volume 9, Issue 2 (3-2021)
Abstract
Introduction: Electroencephalography (EEG) is the most commonly used method to study the function of the brain. This study represents a computerized model for distinguishing between epileptic and healthy subjects using EEG signals with relatively high accuracy. Materials and Methods: The EEG database used in this study was obtained from the data available in Andrzejak. This dataset consists of 5 EEG sets (designated as A to E), each containing 100 EEG sections. Collections A and B comprised EEG signals that have been taken from 5 healthy volunteers. The C and D sets referred to EEGs from patients with focal epilepsy (without ictal recordings) and the E set was derived from a patient with ictal recording. Support vector machines were used after applying principal components analysis or linear discriminant analysis over the features of the signals. MATLAB has been used to implement and test the proposed classification algorithm. To evaluate the proposed method, the confusion matrix, overall success rate, ROC, and the AUC of each class were extracted. K-fold cross-validation technique was used to validate the results. Results: The overall success rate achieved in this study was above 82%. Dimension reduction algorithms can improve its accuracy and speed. Conclusion: It is helpful to be able to predict the occurrence of a seizure early and accurately. Using the computerized model represented in this study could accomplish this goal.
|
|
