Muhammed Cinsdikici

Detection of blood vessels in ophthalmoscope images using MF/ant (matched filter/ant colony) algorithm

Blood vessels in ophthalmoscope images play an important role in diagnosis of some serious pathologies on retinal images. Hence, accurate extraction of vessels is becoming a main topic of this research area. Matched filter (MF) implementation for blood vessel detection is one of the methods giving more accurate results. Using this filter alone might not recover all the vessels (especially the capillaries). In this paper, a novel approach (MF/ant algorithm) is proposed to overcome the deficiency of the MF. The proposed method is a hybrid model of matched filter and ant colony algorithm. In this work, the accuracy and parameters of the hybrid algorithm are also discussed. The proposed method shows its success using the well known reference ophthalmoscope images of DRIVE database.

Vehicle-Classification Algorithm Based on Component Analysis for Single-Loop Inductive Detector

This paper presents a novel vehicle-classification algorithm that uses the time-variable signal generated by a single inductive loop detector. In earlier studies, the noisy raw signal was fed into the algorithm by reducing its size with rough sampling. However, this approach loses the original signal form and cannot be the best exemplar vector. The developed algorithm suggests three contributions to cope with these problems. The first contribution is to clear the noise with discrete Fourier transform (DFT). The second contribution is to transfer the noiseless pattern into the Principal Component Analysis (PCA) domain. PCA is exploited not only for decorrelation but for explicit dimensionality reduction as well. This goal cannot be achieved by simple raw data sampling. The last contribution is to expand the principal components with a local maximum (Lmax) parameter. It strengthens the classification accuracy by emphasizing the undercarriage height variation of the vehicle. These parameters are fed into the three-layered backpropagation neural network (BPNN). BPNN classifies the vehicles into five groups, and the recognition rate is 94.21%. This recognition rate has performed best, compared with the methods presented in published works.

Automatic number plate information extraction and recognition for intelligent transportation system

Abstract In this paper, a new license plate information retrieval system is designed and developed. The system has two main modules: segmentation and recognition. In segmentation, interested information on the image is extracted through the processes of Kaiser resizing, morphological filtering, artificial shifting and bi-directional vertical thresholding. In recognition module, a novel approach for principal component analysis (PCA) and fast backpropagation neural net composition is used as a recognizer. The novel approach is about the construction of Eigen space through the PCA that is used for feature extraction. Our approach is more tolerable to the problems of classical application PCA such as rotation, scaling and character width dependence. The outputs of the new feature extractor used as inputs to the fastbackpropagation neural net recognizer module. This neural network trained with scaled conjugate gradient function. For each module, alternative available methods are mentioned and proper sequence of operations is developed. Finally, overall performance of the system is exported.

Three techniques for automatic extraction of corpus callosum in structural midsagittal brain MR images: Valley Matching, Evolutionary Corpus Callosum Detection and Hybrid method

Abstract Corpus callosum (CC) is an important structure for medical image registration. We propose three novel fully automated for the extraction of CC. Our first algorithm, Valley matching (VM), is based on fixed searched range in histogram processing and uses prior anatomical information for locating CC. The second one, Evolutionary CC Detection (ECD), based on genetic algorithm presents a new fitness function that uses anatomical ratios, instead of fixed prior knowledge without the need for preprocessing. The final one, called Evolutionary Valley Matching (EVM), takes advantages of the strong points of the first and second algorithms. The search space defined for ECD is reduced by VM which uses crowding method to find the peaks in the multi-modal histogram. Another important contribution of this study is that there is no existing method using genetic algorithm for extracting CC. Our proposed algorithms perform with the success rates up to 95.5%.

License plate segmentation for intelligent transportation systems

A license plate segmentation system is designed and developed. The system has preprocessing, approximate region finding, plate extraction and character segmentation modules. For each module, alternative available methods are examined and proper sequence of operations is developed. In character segmentation module, a novel method is devised. Finally, overall performance of the system is reported.

Vascular landmark classification in retinal images using fuzzy RBF

This paper suggests the use of radial based neural networks for classification of the landmark points from retina vessels in the retinal vascular images to diagnose the disease in the diabetic retinopathy patients and to track the periodic differences in retinal vessel images. In the suggested method, Gold Standard images from DRIVE database are used. The performance of landmark detection by the suggested method shows that the method can be used as an algorithm for registration of retinal images.

Two population-based optimization algorithms for minimum weight connected dominating set problem

Abstract Minimum weight connected dominating set (MWCDS) is a very important NP-Hard problem used in many applications such as backbone formation, data aggregation, routing and scheduling in wireless ad hoc and sensor networks. Population-based approaches are very useful to solve NP-Hard optimization problems. In this study, a hybrid genetic algorithm (HGA) and a population-based iterated greedy (PBIG) algorithm for MWCDS problem are proposed. To the best of our knowledge, the proposed algorithms are the first population-based algorithms to solve MWCDS problem on undirected graphs. HGA is a steady-state procedure which incorporates a greedy heuristic with a genetic search. PBIG algorithm refines the population by partially destroying and greedily reconstructing individual solutions. We compare the performance of the proposed algorithms with other greedy heuristics and brute force methods through extensive simulations. We show that our proposed algorithms perform very well in terms of MWCDS solution quality and CPU time.

Automatic registration of structural brain MR images to MNI image space

Disease diagnosis has been made by experts examining the images obtained by magnetic resonance imaging (MRI) technique, the disease process is observed using images taken at different times. Brain MR images are registered to the standard brain atlases because the human brain has a complex structure and varies from person to person. Corpus Callosum (CC) has a big importance for medical image registration because it can be easily distinguished on T1-weighted structural brain MR images and does not vary prominently between individuals. In this study, from the midsagittal brain MR image that belongs to the patient CC is detected fully automatically via Valley Matching (VM) Algorithm. The contribution of this study is registration of patients MR image onto the Montreal Neurological Institute (MNI) image space by using automatically detected reference points.

A simple and improvable method for face region extraction

A new face region extraction method is proposed in this study. The leading property of this method is its application simplicity. Performance success of 65% is achieved with this method and in order to improve this method for images which include multi-faces, we aim to incorporate Zernike moments as a future work.

Robust activation detection methods for real-time and offline fMRI analysis

We propose two contributions with novel approaches to fMRI activation analysis. The first is to apply confidence intervals to locate activations in real-time, and second is a new metric based on robust regression of fMRI signals. These contributions are implemented in our four proposed methods; Instantaneous Activation Method (IAM), Instantaneous Activation Method with Past Blocks (IAMP) for real-time analysis, Task Robust Regression Distance Method (TRRD) for the new metric with robust regression and Instantaneous Robust Regression Distance Method (IRRD) for both contributions. For comparison, a statistical offline method called Task Activation Method (TAM) and a correlation analysis method are also implemented. The methods are initially evaluated with synthetic data generated using two different approaches; first using varying hemodynamic response function signals to simulate a wide range of stimuli responses, along with a Gaussian white noise, and second using no activity state data of a real fMRI experiment, which removes the need to generate noise. The methods are also tested with real fMRI experiments and compared with the results obtained by the widely used SPM tool. The results show that instantaneous methods reveal activations that are lost statistically in an offline analysis. They also reveal further improvements by robust fitting application, which minimizes the outlier effect. TRRD has an area under the ROC curve of 0,7127 for very noisy synthetic images, is reaching up to 0,9608 as the noise decreases, while the instantaneous score is in the range of 0,6124 to 0,8019 in the same noise levels.