mojtaba mohammadpoor

In this research, support vector machine (SVM) as a supervised classification method has been used to explore the relationship between the geochemical anomaly and the surface alterations quantitatively in the Tanurcheh mineralization area. The Tanurcheh area has been located in the Khorasan Razavi province, Iran. This area has been considered as a high potential region for Cu and Au mineralization. The different mineralization processes of Au and Cu have unclearly been intertwined in this area and have created extreme surface alterations.Determination of the major origin of mineralization that has created strong alterations in this area is an important issue that can be addressed using a new proposed scenario. The relationship between the geochemical distribution map and the alteration zone was mathematically calculated using the proposed approach and then the geochemical anomaly map was predicted based on the alteration zones as an innovative achievement.In this paper, the Au and Cu geochemical data were divided into three classes, namely background, regional anomaly and local anomaly using the probability plot method. Two threshold values for Cu (70 and 300 PPM) and Au (0.13 and 0.4 PPM) were obtained by the probability plot method. Then the SVM was utilized to classify the geochemical samples using the ASTER images based on these obtained thresholds. The ASTER 14-band images were used as features in this classification. Using this novel scenario, the relationships between the Au and Cu mineralization processes with the intensity of alterations were determined and therefore the origin of these alteration zones was clarified. The SVM classification indices of correct classification rate (CCR) and confusion matrix demonstrate the main origin of alterations is related to the Cu mineralization process in this area. The CCR indices obtained based on the Au and Cu thresholds are 0.66 and 0.85 respectively. It demonstrates the intensity of alterations has more been affected by the Cu mineralization process and there is a relatively good relationship between the alteration zone and the Cu geochemical distribution map. Finally, the geochemical anomaly and background maps were properly predicted using the SVM and the ASTER bands. This paper shows the new application of SVM as a powerful tool for the interpretation of geochemical anomaly and the intensity of alteration.

Most people experience low back pain at least once in their lifetime. Lumbar disc herniation is one of the major causes of low back pain. Treatment methods for disc herniation are very diverse. So, diagnose the exact size of herniation and it`s location can greatly helps specialists in choosing the best treatment methods. In this research, an automated method for diagnosing lumbar disc herniation using MR images is proposed. To achieve this goal, 130 MR images was collected . In the proposed method, using three algorithms, namely region growing, OTSU and active contour, the intervertebral discs and their boundary were precisely separated from the background of the image. In the next step, after extracting the most significant features of the image, images were divided into healthy and unhealthy classes by SVM classifier with 89.9% accuracy. Classification accuracy also compared with other classifiers such as KNN, ensemble, decision trees, and finally determined, SVM classifier has the highest accuracy in classification.

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.

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.

The overall success rate achieved in this study was above 82%. Dimension reduction algorithms can improve its accuracy and speed.

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.

Magnetic resonance imaging (MRI) plays an important role in clinical diagnosis. The ability of fuzzy c-mean (FCM) algorithm in segmenting MR images has been proven. Some MR images are contaminated with noise. FCM performance is degraded in noisy images. Several efforts are done to overcome this weakness.

The aim of this study was to propose a new method for MR image segmentation which is more resistant than other methods when noisy MR images are confronted.

In this study, simulated brain database prepared by BrainWeb was be used for analysis. First FCM and its improvements were analysed and their ability in segmenting noisy MR images were evaluated. Next, knowing that applying genetic algorithm on improver fuzzy c-mean (IFCM) could improve its performance, a new segmentation method was proposed by applying particle swarm optimization on IFCM.

The proposed algorithm was applied on some intentionally noise-added MR images. Similarity between the segmented image and the original one was measured using Dice index. Other off-the-shelf algorithms were also tested in the same conditions. The indices were presented together. In order to compare the algorithms’ performances, the experiments were repeated using different noisy images.

The obtained results show that the proposed algorithms have better performance in segmenting noisy MR images than existing methods.

Handwritten digit recognition has got a special role in different applications in the field of digital recognition including; handwritten address detection, check, and document. Persian handwritten digits classification has been facing difficulties due to different handwritten styles, inter-class similarities, and intra-class differences. In this paper, a novel method for detecting Persian handwritten digits is presented. In the proposed method, a combination of Histogram of Oriented Gradients (HOG), 4-side profiles of the digit image, and some horizontal and vertical samples was used and the dimension of the feature vector was reduced using Principal Component Analysis (PCA). The proposed method applied to the HODA database, and Support Vector Machine (SVM) was used in the classification step. Results revealed that the detection accuracy of such method has 99% accuracy with an adequate rate due to existing unacceptable samples in the database, therefore, the proposed method could improve the outcomes compared to other existing methods.

Breast cancer is the second cause of mortality among women. Early detection of it can enhance the chance of survival. Screening systems such as mammography cannot perfectly differentiate between patients and healthy individuals. Computer-aided diagnosis can help physicians make a more accurate diagnosis.
Regarding the importance of separating normal and abnormal cases in screening systems, a hierarchical classification system is defined in this paper. The proposed system is including two Adaptive Boosting (AdaBoost) classifiers, the first classifier separates the candidate images into two groups of normal and abnormal. The second classifier is applied on the abnormal group of the previous stage and divides them into benign and malignant categories. The proposed algorithm is evaluated by applying it on publicly available Mammographic Image Analysis Society (MIAS) dataset. 288 images of the database are used, including 208 normal and 80 abnormal images. 47 images of the abnormal images showed benign lesion and 33 of them had malignant lesion.
Applying the proposed algorithm on MIAS database indicates its advantage compared to previous methods. A major improvement occurred in the first classification stage. Specificity, sensitivity, and accuracy of the first classifier are obtained as 100%, 95.83%, and 97.91%, respectively. These values are calculated as 75% in the second stage
A hierarchical classification method for breast cancer detection is developed in this paper. Regarding the importance of separating normal and abnormal cases in screening systems, the first classifier is devoted to separate normal and tumorous cases. Experimental results on available database shown that the performance of this step is adequately high (100% specificity). The second layer is designed to detect tumor type. The accuracy in the second layer is obtained 75%.