Optimal query-based relevance feedback in medical image retrieval using score fusion-based classification.

In this paper, a new content-based medical image retrieval (CBMIR) framework using an effective classification method and a novel relevance feedback (RF) approach are proposed. For a large-scale database with diverse collection of different modalities, query image classification is inevitable due to firstly, reducing the computational complexity and secondly, increasing influence of data fusion by removing unimportant data and focus on the more valuable information. Hence, we find probability distribution of classes in the database using Gaussian mixture model (GMM) for each feature descriptor and then using the fusion of obtained scores from the dependency probabilities, the most relevant clusters are identified for a given query. Afterwards, visual similarity of query image and images in relevant clusters are calculated. This method is performed separately on all feature descriptors, and then the results are fused together using feature similarity ranking level fusion algorithm. In the RF level, we propose a new approach to find the optimal queries based on relevant images. The main idea is based on density function estimation of positive images and strategy of moving toward the aggregation of estimated density function. The proposed framework has been evaluated on ImageCLEF 2005 database consisting of 10,000 medical X-ray images of 57 semantic classes. The experimental results show that compared with the existing CBMIR systems, our framework obtains the acceptable performance both in the image classification and in the image retrieval by RF.

Multiresolution image registration in digital x-ray angiography with intensity variation modeling.

Digital subtraction angiography (DSA) is a widely used technique for visualization of vessel anatomy in diagnosis and treatment. However, due to unavoidable patient motions, both externally and internally, the subtracted angiography images often suffer from motion artifacts that adversely affect the quality of the medical diagnosis. To cope with this problem and improve the quality of DSA images, registration algorithms are often employed before subtraction. In this paper, a novel elastic registration algorithm for registration of digital X-ray angiography images, particularly for the coronary location, is proposed. This algorithm includes a multiresolution search strategy in which a global transformation is calculated iteratively based on local search in coarse and fine sub-image blocks. The local searches are accomplished in a differential multiscale framework which allows us to capture both large and small scale transformations. The local registration transformation also explicitly accounts for local variations in the image intensities which incorporated into our model as a change of local contrast and brightness. These local transformations are then smoothly interpolated using thin-plate spline interpolation function to obtain the global model. Experimental results with several clinical datasets demonstrate the effectiveness of our algorithm in motion artifact reduction.

An automatic fuzzy-based multi-temporal brain digital subtraction angiography image fusion algorithm using curvelet transform and content selection strategy.

Recently image fusion has prominent role in medical image processing and is useful to diagnose and treat many diseases. Digital subtraction angiography is one of the most applicable imaging to diagnose brain vascular diseases and radiosurgery of brain. This paper proposes an automatic fuzzy-based multi-temporal fusion algorithm for 2-D digital subtraction angiography images. In this algorithm, for blood vessel map extraction, the valuable frames of brain angiography video are automatically determined to form the digital subtraction angiography images based on a novel definition of vessel dispersion generated by injected contrast material. Our proposed fusion scheme contains different fusion methods for high and low frequency contents based on the coefficient characteristic of wrapping second generation of curvelet transform and a novel content selection strategy. Our proposed content selection strategy is defined based on sample correlation of the curvelet transform coefficients. In our proposed fuzzy-based fusion scheme, the selection of curvelet coefficients are optimized by applying weighted averaging and maximum selection rules for the high frequency coefficients. For low frequency coefficients, the maximum selection rule based on local energy criterion is applied to better visual perception. Our proposed fusion algorithm is evaluated on a perfect brain angiography image dataset consisting of one hundred 2-D internal carotid rotational angiography videos. The obtained results demonstrate the effectiveness and efficiency of our proposed fusion algorithm in comparison with common and basic fusion algorithms.

Fast and automatic algorithm for optic disc extraction in retinal images using principle-component-analysis-based preprocessing and curvelet transform.

Optic disc or optic nerve (ON) head extraction in retinal images has widespread applications in retinal disease diagnosis and human identification in biometric systems. This paper introduces a fast and automatic algorithm for detecting and extracting the ON region accurately from the retinal images without the use of the blood-vessel information. In this algorithm, to compensate for the destructive changes of the illumination and also enhance the contrast of the retinal images, we estimate the illumination of background and apply an adaptive correction function on the curvelet transform coefficients of retinal images. In other words, we eliminate the fault factors and pave the way to extract the ON region exactly. Then, we detect the ON region from retinal images using the morphology operators based on geodesic conversions, by applying a proper adaptive correction function on the reconstructed image's curvelet transform coefficients and a novel powerful criterion. Finally, using a local thresholding on the detected area of the retinal images, we extract the ON region. The proposed algorithm is evaluated on available images of DRIVE and STARE databases. The experimental results indicate that the proposed algorithm obtains an accuracy rate of 100% and 97.53% for the ON extractions on DRIVE and STARE databases, respectively.

A novel material detection algorithm based on 2D GMM-based power density function and image detail addition scheme in dual energy X-ray images.

Material detection is a vital need in dual energy X-ray luggage inspection systems at security of airport and strategic places. In this paper, a novel material detection algorithm based on statistical trainable models using 2-Dimensional power density function (PDF) of three material categories in dual energy X-ray images is proposed. In this algorithm, the PDF of each material category as a statistical model is estimated from transmission measurement values of low and high energy X-ray images by Gaussian Mixture Models (GMM). Material label of each pixel of object is determined based on dependency probability of its transmission measurement values in the low and high energy to PDF of three material categories (metallic, organic and mixed materials). The performance of material detection algorithm is improved by a maximum voting scheme in a neighborhood of image as a post-processing stage. Using two background removing and denoising stages, high and low energy X-ray images are enhanced as a pre-processing procedure. For improving the discrimination capability of the proposed material detection algorithm, the details of the low and high energy X-ray images are added to constructed color image which includes three colors (orange, blue and green) for representing the organic, metallic and mixed materials. The proposed algorithm is evaluated on real images that had been captured from a commercial dual energy X-ray luggage inspection system. The obtained results show that the proposed algorithm is effective and operative in detection of the metallic, organic and mixed materials with acceptable accuracy.

Seizure-specific wavelet (Seizlet) design for epileptic seizure detection using CorrEntropy ellipse features based on seizure modulus maximas patterns.

BACKGROUND: EEG signal analysis of pediatric patients plays vital role for making a decision to intervene in presurgical stages. NEW METHOD: In this paper, an offline seizure detection algorithm based on definition of a seizure-specific wavelet (Seizlet) is presented. After designing the Seizlet, by forming cone of influence map of the EEG signal, four types of layouts are analytically designed that are called Seizure Modulus Maximas Patterns (SMMP). By mapping CorrEntropy Induced Metric (CIM) series, four structural features based on least square estimation of fitted non-tilt conic ellipse are extracted that are called CorrEntropy Ellipse Features (CEF). The parameters of the SMMP and CEF are tuned by employing a hybrid optimization algorithm based on honeybee hive optimization in combination with Las Vegas randomized algorithm and Elman recurrent classifier. Eventually, the optimal features by AdaBoost classifiers in a cascade structure are classified into the seizure and non-seizure signals. RESULTS: The proposed algorithm is evaluated on 844h signals with 163 seizure events recorded from 23 patients with intractable seizure disorder and accuracy rate of 91.44% and false detection rate of 0.014 per hour are obtained by 7-channel EEG signals. COMPARISON WITH EXISTING METHOD(S): To overcome the restrictions of general kernels and wavelet coefficient-based features, we designed the Seizlet as an exclusive kernel of seizure signal for first time. Also, the Seizlet-based patterns of EEG signals have been modeled to extract the seizure. CONCLUSIONS: The reported results demonstrate that our proposed Seizlet is effectiveness to extract the patterns of the epileptic seizure.

A novel retina-based human identification algorithm based on geometrical shape features using a hierarchical matching structure.

BACKGROUND AND OBJECTIVES: Retinal image is one of the most secure biometrics which is widely used in human identification application. This paper represents a rotation and translation-invariant human identification algorithm based on a new definition of geometrical shape features of the retinal image using a hierarchical matching structure. METHODS: In this algorithm, the retinal images are represented by regions which are surrounded by blood vessels that are named Surrounded Regions (SRs). A perfect set of region-based and boundary-based features are defined on the SRs. In the boundary-based features, by defining corner points of the SR, novel features such as angle of SR corner, centroid distance and weighted corner angle are defined which they can describe well the variation rate of boundary and geometry of the SR. To match the SR of a query with enrolled SR in database, the extracted features in a hierarchical structure from simpler features through more complex features are applied to filter the enrolled SRs in the database for search space reduction. At last, the matched candidate SRs with the query SRs determine the identification or rejection of query image by proposed decision making scenario. In this scenario, the identification is carried out when at least two SRs of the query are matched with two SRs of an individual in the database. RESULTS: The proposed algorithm is evaluated on STARE and DRIVE retinal image databases in six different experiments and is achieved an accuracy rate of 100% and an average processing time of 3.216sec and 3.225sec, respectively. The reported results demonstrate the efficiency of our proposed algorithm in the eye-movement condition. CONCLUSION: In our work, by defining the SR-based features and employing a hierarchical matching structure, the computational complexity of matching step is reduced and also the identification performance is improved. Moreover, the proposed algorithm overcomes the problem of natural movements of the head and eye during the capturing process.

Real-time seizure prediction using RLS filtering and interpolated histogram feature based on hybrid optimization algorithm of Bayesian classifier and Hunting search.

BACKGROUND AND OBJECTIVES: Epileptic seizure prediction using EEG signal analysis is an important application for drug therapy and pediatric patient monitoring. Time series estimation to obtain the future samples of EEG signal has vital role for detecting seizure attack. In this paper, a novel density-based real-time seizure prediction algorithm based on a trained offline seizure detection algorithm is proposed. METHODS: In the offline seizure detection procedure, after signal preprocessing, histogram-based statistical features are extracted from signal probability distribution. By defining a deterministic polynomial model on the normalized histogram, a novel syntactic feature that is named Interpolated Histogram Feature (IHF) is proposed. Moreover, with this feature, Seizure Distribution Model (SDM) as a descriptor of the seizure and non-seizure signals is presented. By using a novel hybrid optimization algorithm based on Bayesian classifier and Hunting Search (HuS) algorithm, the optimal features are selected. To detect the seizure attacks in the online mode, a Multi-Layer Perceptron (MLP) classifier is trained with the optimal features in the offline procedure. For online prediction, the enhanced Recursive Least Square (RLS) filter is applied to estimate sample-by-sample of the EEG signal. Also, a density-based signal tracking scenario is introduced to update and tune the parameters of RLS filtering algorithm. RESULTS: Our prediction algorithm is evaluated on 104 hours of EEG signals recorded from 23 pediatric patients. Our online signal prediction algorithm provides the accuracy rate of 86.56% and precision rate of 86.53% simultaneously using the trained MLP classifier from the offline mode. The recall rate of seizure prediction is 97.27% and the false prediction rate of 0.00215 per hour is achieved as well. Ultimately, the future samples of EEG signal are estimated, and the time of seizure signal prediction is also converged to 6.64 seconds. CONCLUSION: In our proposed real-time algorithm, by implementing a density-based signal tracking scenario, the future samples of signal with suitable time is predicted and the seizure is detected based on the optimal features from the IHF and histogram-based statistical features with acceptable performance.

Clavulanic acid production estimation based on color and structural features of Streptomyces clavuligerus bacteria using self-organizing map and genetic algorithm.

The utilization of antibiotics produced by Clavulanic acid (CA) is an increasing need in medicine and industry. Usually, the CA is created from the fermentation of Streptomycen Clavuligerus (SC) bacteria. Analysis of visual and morphological features of SC bacteria is an appropriate measure to estimate the growth of CA. In this paper, an automatic and fast CA production level estimation algorithm based on visual and structural features of SC bacteria instead of statistical methods and experimental evaluation by microbiologist is proposed. In this algorithm, structural features such as the number of newborn branches, thickness of hyphal and bacterial density and also color features such as acceptance color levels are extracted from the SC bacteria. Moreover, PH and biomass of the medium provided by microbiologists are considered as specified features. The level of CA production is estimated by using a new application of Self-Organizing Map (SOM), and a hybrid model of genetic algorithm with back propagation network (GA-BPN). The proposed algorithm is evaluated on four carbonic resources including malt, starch, wheat flour and glycerol that had used as different mediums of bacterial growth. Then, the obtained results are compared and evaluated with observation of specialist. Finally, the Relative Error (RE) for the SOM and GA-BPN are achieved 14.97% and 16.63%, respectively.

A novel fast epileptic seizure onset detection algorithm using general tensor discriminant analysis.

Seizure onset detection with minimum latency has a key role in improving the therapy studies of epilepsy. In this article, an epileptic seizure onset detection algorithm based on general tensor discriminant analysis is proposed to detect the seizure through EEG signals with smallest delay before the development of clinical symptoms. In this algorithm, seizure and nonseizure EEG signal epochs are exhibited by spectral, spatial, and temporal domains (third-order tensors) in wavelet decomposition. Then, to reduce feature space, projection matrices are extracted from tensor-represented EEG signal by general tensor discriminant analysis. In this strategy, the discriminative information in the training tensors is preserved that it is a benefit in comparison with common feature space reduction algorithms such as principal component analysis and multilinear subspace analysis. The proposed seizure onset detection algorithm is evaluated on 44 epileptic patients from 2 standard datasets and recognizes 98% of seizures with average delay of 4.5 seconds. The obtained results show efficiency and effectiveness of our proposed algorithm in comparison with other algorithms.

Medical X-ray Image Hierarchical Classification Using a Merging and Splitting Scheme in Feature Space.

Due to the daily mass production and the widespread variation of medical X-ray images, it is necessary to classify these for searching and retrieving proposes, especially for content-based medical image retrieval systems. In this paper, a medical X-ray image hierarchical classification structure based on a novel merging and splitting scheme and using shape and texture features is proposed. In the first level of the proposed structure, to improve the classification performance, similar classes with regard to shape contents are grouped based on merging measures and shape features into the general overlapped classes. In the next levels of this structure, the overlapped classes split in smaller classes based on the classification performance of combination of shape and texture features or texture features only. Ultimately, in the last levels, this procedure is also continued forming all the classes, separately. Moreover, to optimize the feature vector in the proposed structure, we use orthogonal forward selection algorithm according to Mahalanobis class separability measure as a feature selection and reduction algorithm. In other words, according to the complexity and inter-class distance of each class, a sub-space of the feature space is selected in each level and then a supervised merging and splitting scheme is applied to form the hierarchical classification. The proposed structure is evaluated on a database consisting of 2158 medical X-ray images of 18 classes (IMAGECLEF 2005 database) and accuracy rate of 93.6% in the last level of the hierarchical structure for an 18-class classification problem is obtained.

A novel real-time patient-specific seizure diagnosis algorithm based on analysis of EEG and ECG signals using spectral and spatial features and improved particle swarm optimization classifier.

This paper proposes a novel real-time patient-specific seizure diagnosis algorithm based on analysis of electroencephalogram (EEG) and electrocardiogram (ECG) signals to detect seizure onset. In this algorithm, spectral and spatial features are selected from seizure and non-seizure EEG signals by Gabor functions and principal component analysis (PCA). Furthermore, four features based on heart rate acceleration are extracted from ECG signals to form feature vector. Then a neural network classifier based on improved particle swarm optimization (IPSO) learning algorithm is developed to determine an optimal nonlinear decision boundary. This classifier allows to adjust the parameters of the neural network classifier, efficiently. This algorithm can automatically detect the presence of seizures with minimum delay which is an important factor from a clinical viewpoint. The performance of the proposed algorithm is evaluated on a dataset consisting of 154 h records and 633 seizures from 12 patients. The results indicate that the algorithm can recognize the seizures with the smallest latency and higher good detection rate (GDR) than other presented algorithms in the literature.

Content-based medical image classification using a new hierarchical merging scheme.

Automatic medical image classification is a technique for assigning a medical image to a class among a number of image categories. Due to computational complexity, it is an important task in the content-based image retrieval (CBIR). In this paper, we propose a hierarchical medical image classification method including two levels using a perfect set of various shape and texture features. Furthermore, a tessellation-based spectral feature as well as a directional histogram has been proposed. In each level of the hierarchical classifier, using a new merging scheme and multilayer perceptron (MLP) classifiers (merging-based classification), homogenous (semantic) classes are created from overlapping classes in the database. The proposed merging scheme employs three measures to detect the overlapping classes: accuracy, miss-classified ratio, and dissimilarity. The first two measures realize a supervised classification method and the last one realizes an unsupervised clustering technique. In each level, the merging-based classification is applied to a merged class of the previous level and splits it to several classes. This procedure is progressive to achieve more classes. The proposed algorithm is evaluated on a database consisting of 9100 medical X-ray images of 40 classes. It provides accuracy rate of 90.83% on 25 merged classes in the first level. If the correct class is considered within the best three matches, this value will increase to 97.9%.