Niyazi Kilic

Evaluation of face recognition techniques using PCA, wavelets and SVM

In this study, we present an evaluation of using various methods for face recognition. As feature extracting techniques we benefit from wavelet decomposition and Eigenfaces method which is based on Principal Component Analysis (PCA). After generating feature vectors, distance classifier and Support Vector Machines (SVMs) are used for classification step. We examined the classification accuracy according to increasing dimension of training set, chosen feature extractor-classifier pairs and chosen kernel function for SVM classifier. As test set we used ORL face database which is known as a standard face database for face recognition applications including 400 images of 40 people. At the end of the overall separation task, we obtained the classification accuracy 98.1% with Wavelet-SVM approach for 240 image training set.

Classification of Pulmonary Nodules by Using Hybrid Features

Early detection of pulmonary nodules is extremely important for the diagnosis and treatment of lung cancer. In this study, a new classification approach for pulmonary nodules from CT imagery is presented by using hybrid features. Four different methods are introduced for the proposed system. The overall detection performance is evaluated using various classifiers. The results are compared to similar techniques in the literature by using standard measures. The proposed approach with the hybrid features results in 90.7% classification accuracy (89.6% sensitivity and 87.5% specificity).

Breast Cancer Detection with Reduced Feature Set

This paper explores feature reduction properties of independent component analysis (ICA) on breast cancer decision support system. Wisconsin diagnostic breast cancer (WDBC) dataset is reduced to one-dimensional feature vector computing an independent component (IC). The original data with 30 features and reduced one feature (IC) are used to evaluate diagnostic accuracy of the classifiers such as k-nearest neighbor (k-NN), artificial neural network (ANN), radial basis function neural network (RBFNN), and support vector machine (SVM). The comparison of the proposed classification using the IC with original feature set is also tested on different validation (5/10-fold cross-validations) and partitioning (20%–40%) methods. These classifiers are evaluated how to effectively categorize tumors as benign and malignant in terms of specificity, sensitivity, accuracy, F-score, Youdens index, discriminant power, and the receiver operating characteristic (ROC) curve with its criterion values including area under curve (AUC) and 95% confidential interval (CI). This represents an improvement in diagnostic decision support system, while reducing computational complexity.

A new method for pulmonary nodule detection using decision trees

A computer-aided detection (CAD) can help radiologists in diagnosing of lung diseases at an early level. In this study, a new CAD system for pulmonary nodule detection from CT imagery is presented by using morphological features and patient information properties. Decision trees are utilized for classification and overall detection performance is evaluated. Results are compared to similar techniques in the literature by using standard measures. Proposed CAD system with random forest classifier result in 90.5 % sensitivity and 87.6 % specificity of detection performance.

Efficient Estimation of Osteoporosis using Artificial Neural Networks

In this communication, Artificial Neural Network (ANN) is applied to discriminate osteoporotic fracture and control cases in a group of 304 patients. ANN is one of the popular methods in optimization of complex engineering problems compared to the classical statistical methods. In our study group, we consider some parameters as inputs: three bone densitometry parameters (BMD) (Femoral neck BMD, total body BMD and L2L4 spine BMD), three fractal parameters [1,5] (Hmin, Hmean, Hmax), and age of the patient. We studied three ANN structures with various inputs and hidden neurons. We have reached up to 81.66% correct classification. In comparison we have tested a classical discriminant analysis (Mahalanobis-Fisher) and we only obtained 72% of correct classification. We can conclude that ANN is one of the promising methods in the diagnosis of osteoporosis.

Mammographic Mass Detection using Wavelets as Input to Neural Networks

The objective of this paper is to demonstrate the utility of artificial neural networks, in combination with wavelet transforms for the detection of mammogram masses as malign or benign. A total of 45 patients who had breast masses in their mammography were enrolled in the study. The neural network was trained on the wavelet based feature vectors extracted from the mammogram masses for both benign and malign data. Therefore, in this study, Multilayer ANN was trained with the Backpropagation, Conjugate Gradient and Levenberg–Marquardt algorithms and ten-fold cross validation procedure was used. A satisfying sensitivity percentage of 89.2% was achieved with Levenberg–Marquardt algorithm. Since, this algorithm combines the best features of the Gauss–Newton technique and the other steepest-descent algorithms and thus it reaches desired results very fast.

Diagnosis of Renal Failure Disease Using Adaptive Neuro-Fuzzy Inference System

Adaptive Neuro-Fuzzy Inference System (ANFIS) is one of the useful and powerful neural network approaches for the solution of function approximation and pattern recognition problems in the last decades. In this paper, the diagnosis of renal failure disease is investigated using ANFIS approach. Totally the raw data of 112 patients is obtained from Istanbul and Cerrahpasa Medical Faculties of Istanbul University, Turkey. Sixty-four of them are related to renal failures and the rest data belong to healthy persons. In ANFIS model, three rules and Gaussian membership functions are chosen, where rules are determined by the subtractive clustering method. Seven parameters of the patients are considered for the input of the system. These are: Blood Urea Nitrogen (BUN), Creatinine, Uric Acid, Potassium (K), Calcium (Ca), Phosphorus (P) and age. We try to decide whether the patient is ill or not. We have reached 100% success in ANFIS and have better results compared to Support Vector Machine (SVM) and Artificial Neural Networks (ANN).

Colonic Polyp Detection in CT Colonography with Fuzzy Rule Based 3D Template Matching

In this paper, we introduced a computer aided detection (CAD) system to facilitate colonic polyp detection in computer tomography (CT) data using cellular neural network, genetic algorithm and three dimensional (3D) template matching with fuzzy rule based tresholding. The CAD system extracts colon region from CT images using cellular neural network (CNN) having A, B and I templates that are optimized by genetic algorithm in order to improve the segmentation performance. Then, the system performs a 3D template matching within four layers with three different cell of 8 × 8, 12 × 12 and 20 × 20 to detect polyps. The CAD system is evaluated with 1043 CT colonography images from 16 patients containing 15 marked polyps. All colon regions are segmented properly. The overall sensitivity of proposed CAD system is 100% with the level of 0.53 false positives (FPs) per slice and 11.75 FPs per patient for the 8 × 8 cell template. For the 12 × 12 cell templates, detection sensitivity is 100% at 0.494 FPs per slice and 8.75 FPs per patient and for the 20 × 20 cell templates, detection sensitivity is 86.66% with the level of 0.452 FPs per slice and 6.25 FPs per patient.

A Novel Approach to Malignant-Benign Classification of Pulmonary Nodules by Using Ensemble Learning Classifiers

Computer-aided detection systems can help radiologists to detect pulmonary nodules at an early stage. In this paper, a novel Computer-Aided Diagnosis system (CAD) is proposed for the classification of pulmonary nodules as malignant and benign. The proposed CAD system using ensemble learning classifiers, provides an important support to radiologists at the diagnosis process of the disease, achieves high classification performance. The proposed approach with bagging classifier results in 94.7 %, 90.0 % and 77.8 % classification sensitivities for benign, malignant and undetermined classes (89.5 % accuracy), respectively.

Multifont Ottoman character recognition using support vector machine

In this study, an optical character recognition (OCR) system, which implements segmentation, normalization, edge detection and recognition of the Ottoman script, is proposed. Each multifont Ottoman character is written with four different shapes according to its position in the word being at beginning, middle, at the end and in isolated form. We have used printed type of Ottoman scripts in image acquisition. Then image segmentation, normalization and finally edge detection are performed for feature extraction, where edge detection is achieved by cellular neural network (CNN) approach. After these pre-proces steps, we recognize these multifont Ottoman characters using support vector machine (SVM) technique. In SVM training, polynomial (linear and quadratic) and Gaussian radial basis function kernels are chosen. The proposed recognition system has succeeded in classification up to 87.32% with quadratic kernel.