An improved saliency model of visual attention dependent on image content.

Many visual attention models have been presented to obtain the saliency of a scene, i.e., the visually significant parts of a scene. However, some mechanisms are still not taken into account in these models, and the models do not fit the human data accurately. These mechanisms include which visual features are informative enough to be incorporated into the model, how the conspicuity of different features and scales of an image may integrate to obtain the saliency map of the image, and how the structure of an image affects the strategy of our attention system. We integrate such mechanisms in the presented model more efficiently compared to previous models. First, besides low-level features commonly employed in state-of-the-art models, we also apply medium-level features as the combination of orientations and colors based on the visual system behavior. Second, we use a variable number of center-surround difference maps instead of the fixed number used in the other models, suggesting that human visual attention operates differently for diverse images with different structures. Third, we integrate the information of different scales and different features based on their weighted sum, defining the weights according to each component's contribution, and presenting both the local and global saliency of the image. To test the model's performance in fitting human data, we compared it to other models using the CAT2000 dataset and the Area Under Curve (AUC) metric. Our results show that the model has high performance compared to the other models (AUC = 0.79 and sAUC = 0.58) and suggest that the proposed mechanisms can be applied to the existing models to improve them.

A neuro-computational model of visual attention with multiple attentional control sets.

In numerous activities, humans need to attend to multiple sources of visual information at the same time. Although several recent studies support the evidence of this ability, the mechanism of multi-item attentional processing is still a matter of debate and has not been investigated much by previous computational models. Here, we present a neuro-computational model aiming to address specifically the question of how subjects attend to two items that deviate defined by feature and location. We simulate the experiment of Adamo et al. (2010) which required subjects to use two different attentional control sets, each a combination of color and location. The structure of our model is composed of two components "attention" and "decision-making". The important aspect of our model is its dynamic equations that allow us to simulate the time course of processes at a neural level that occur during different stages until a decision is made. We analyze in detail the conditions under which our model matches the behavioral and EEG data from human subjects. Consistent with experimental findings, our model supports the hypothesis of attending to two control settings concurrently. In particular, our model proposes that initially, feature-based attention operates in parallel across the scene, and only in ongoing processing, a selection by the location takes place.

Extracting features from wrist vein images using fractional fourier transform for person verification.

One of the major concerns is the security and protection of individuals' privacy in society. Biometric methods have been developed in recent years and they are widely used in many places and devices to protect information and assets. Wrist veins are inside the body and their pattern is unique for each person. In this paper, the PUT wrist vein dataset is used that comprises of palm and wrist vein images and each section has 1200 images of right and left hand. Wrist vein images are analyzed in the time-frequency domain by applying Fractional Fourier transform (FrFT), and the extracted features include phase, magnitude, real, and imaginary parts of FrFT coefficients. Since the number of features is very large by implementing FrFT, receiver operating characteristic (ROC) is applied for feature scoring and the best features are selected by this tool. Support Vector Machine (SVM) is used to classify real and impostor samples. The results of various features extracted by FrFT are compared, and according to the obtained results, we deduced that the phase feature is stronger than other features for person authentication based on wrist vein images, and this feature achieved 100% accuracy.

Machine learning distinguishes neurosurgical skill levels in a virtual reality tumor resection task.

This study outlines the first investigation of application of machine learning to distinguish "skilled" and "novice" psychomotor performance during a virtual reality (VR) brain tumor resection task. Tumor resection task participants included 23 neurosurgeons and senior neurosurgery residents as the "skilled" group and 92 junior neurosurgery residents and medical students as the "novice" group. The task involved removing a series of virtual brain tumors without causing injury to surrounding tissue. Originally, 150 features were extracted followed by statistical and forward feature selection. The selected features were provided to 4 classifiers, namely, K-Nearest Neighbors, Parzen Window, Support Vector Machine, and Fuzzy K-Nearest Neighbors. Sets of 5 to 30 selected features were provided to the classifiers. A working point of 15 premium features resulted in accuracy values as high as 90% using the Supprt Vector Machine. The obtained results highlight the potentials of machine learning, applied to VR simulation data, to help realign the traditional apprenticeship educational paradigm to a more objective model, based on proven performance standards. Graphical abstract Using several scenarios of virtual reality neurosurgical tumor resection together with machine learning classifiers to distinguish skill level.

Segmentation of cardiac fats based on Gabor filters and relationship of adipose volume with coronary artery disease using FP-Growth algorithm in CT scans.

Heart mediastinal and epicardial fat tissues are related to several adverse metabolic effects and cardiovascular risk factors, especially coronary artery disease (CAD). The manual segmentation of those fats is that the high dependence on user intervention and time-consuming analyzes. As a result, the automated measurement of cardiac fats could be considered as one of the most important biomarkers for cardiovascular risks in imaging and medical visualization by physicians. In this paper, we validate an automatic approach for the cardiac fat segmentation in non-contrast CT images then investigate the correlation between cardiac fat volume and CAD using the association rule mining algorithm. The pre-processing step includes threshold and contrast enhancement, the feature extraction step includes Gabor filter bank based on GLCM, the cardiac fat segmentation step is predicated on pattern recognition classification algorithms, and eventually, the step of investigating the relationship between cardiac fat volume and CAD is using FP-Growth algorithm. Experimental validation using CT images of two databases points to a good performance in cardiac fat segmentation. Experiments showed that the accuracy of the designed algorithm using the ensemble classifier with the best performance over other classifiers for the cardiac fat segmentation was 99.2%, with a sensitivity of 96.3% and a specificity of 99.8%. The results of using the FP-Growth algorithm showed that the low volume of epicardial (Confidence = 0.6818, Lift = 1.0626) and mediastinal (Confidence = 0.6696, Lift = 1.0436) fat are associated with healthy individuals and the high volume of epicardial (Confidence = 0.8, Lift = 2.2326) and mediastinal (Confidence = 0.75, Lift = 2.093) fat are related to individuals of CAD. As a result, cardiac fats can be used as a reliable biomarker tool in predicting the extent of CAD stenosis.

Phonocardiography-based mitral valve prolapse detection with using fractional fourier transform.

Mitral Valve Prolapse (MVP) is a common condition among people, which is often benign and does not need any serious treatment. However, this doesn't mean that MVP can't cause any problems. In malignant conditions, MVP can cause mitral failure and also heart failure. Early diagnosis of MVP is significantly important to control and reduce its complications. Since the phonocardiogram signal provides useful information about heart valves function, it can be used for MVP detection. To detect MVP, the signal was denoised and segmented into heart cycles and constant three-second pieces in the first and second approaches, respectively. Next, based on the Fractional Fourier Transform (FrFT), the desired features were extracted. Then, the extracted features were windowed by a Moving Logarithmic Median Window (MLMW) and optimum features were selected using Mahalanobis, Bhattacharyya, Canberra, and Minkowski distance criteria. Finally, using the selected features, classification was performed by using the K-Nearest Neighbor (KNN) and the Suppor Vector Machine (SVM) classifiers to find out whether a segment is prolapsed. The best results of the experiment on the collected database contain 15 prolapsed and 6 non-prolapsed subjects using the A-test method show 96.25 ± 2.43 accuracy, 98.5 ± 3.37 sensitivity, 94.0 ± 5.16 specificity, 96.0 ± 3.44 precision, 92.5 ± 4.86 kappa, and 96.6 ± 2.34 f-score with the SVM classifier.

Tremor Assessment during Virtual Reality Brain Tumor Resection.

OBJECTIVE: Assessment of physiological tremor during neurosurgical procedures may provide further insights into the composites of surgical expertise. Virtual reality platforms may provide a mechanism for the quantitative assessment of physiological tremor. In this study, a virtual reality simulator providing haptic feedback was used to study physiological tremor in a simulated tumor resection task with participants from a "skilled" group and a "novice" group. DESIGN: The task involved using a virtual ultrasonic aspirator to remove a series of virtual brain tumors with different visual and tactile characteristics without causing injury to surrounding tissue. Power spectral density analysis was employed to quantitate hand tremor during tumor resection. Statistical t test was used to determine tremor differences between the skilled and novice groups obtained from the instrument tip x, y, z coordinates, the instrument roll, pitch, yaw angles, and the instrument haptic force applied during tumor resection. SETTING: The study was conducted at the Neurosurgical Simulation and Artificial Intelligence Learning Centre, Montreal Neurological Institute and Hospital, McGill University, Montreal, Canada. PARTICIPANTS: The skilled group comprised 23 neurosurgeons and senior residents and the novice group comprised 92 junior residents and medical students. RESULTS: The spectral analysis allowed quantitation of physiological tremor during virtual reality tumor resection. The skilled group displayed smaller physiological tremor than the novice group in all cases. In 3 out of 7 cases the difference was statistically significant. CONCLUSIONS: The first investigation of the application of a virtual reality platform is presented for the quantitation of physiological tremor during a virtual reality tumor resection task. The goal of introducing such methodology to assess tremor is to highlight its potential educational application in neurosurgical resident training and in helping to further define the psychomotor skill set of surgeons.

An Iterative Method for Estimating Nonlinear Elastic Constants of Tumor in Soft Tissue from Approximate Displacement Measurements.

Objectives: Various elastography techniques have been proffered based on linear or nonlinear constitutive models with the aim of detecting and classifying pathologies in soft tissues accurately and noninvasively. Biological soft tissues demonstrate behaviors which conform to nonlinear constitutive models, in particular the hyperelastic ones. In this paper, we represent the results of our steps towards implementing ultrasound elastography to extract hyperelastic constants of a tumor inside soft tissue. Methods: Hyperelastic parameters of the unknown tissue have been estimated by applying the iterative method founded on the relation between stress, strain, and the parameters of a hyperelastic model after (a) simulating the medium's response to a sinusoidal load and extracting the tissue displacement fields in some instants and (b) estimating the tissue displacement fields from the recorded/simulated ultrasound radio frequency signals and images using the cross correlation-based technique. Results: Our results indicate that hyperelastic parameters of an unidentified tissue could be precisely estimated even in the conditions where there is no prior knowledge of the tissue, or the displacement fields have been approximately calculated using the data recorded by a clinical ultrasound system. Conclusions: The accurate estimation of nonlinear elastic constants yields to the correct cognizance of pathologies in soft tissues.

Efficient Sensitivity Based Reconstruction Technique to Accomplish Breast Hyperelastic Elastography.

Hyperelastic models have been acknowledged as constitutive equations which reliably model the nonlinear behaviors observed from soft tissues under various loading conditions. Among them, the Mooney-Rivlin, Yeoh, and polynomial models have been proved capable of accurately modeling responses of breast tissues to applied compressions. Hyperelastic elastography technique takes advantage of the disparities between hyperelastic parameters of varied tissues and the change in hyperelastic parameters in pathological processes. The precise reconstruction of hyperelastic parameters of a completely unknown pathology in the breast in a noninvasive and nondestructive way using the ultrasound elastography has been scrutinized in this paper. In the ultrasound elastography, tissue displacement field is extracted from radio frequency signals or images recorded using the ultrasound medical imaging system; hence the exact displacement field might not be obtained. Our results indicate that the parameters estimated by manipulating the iterative sensitivity-matrix based method converge to tissue's real hyperelastic parameters providing appropriate parameters are assigned to the hypothetical hyperelastic and regularization parameters. Iterative methods have therefore been proposed to compute proper hypothetical hyperelastic and regularization parameters. Accurate estimates of hyperelastic parameters of obscure breast pathology have been achieved even from imprecise measurements of displacements induced in the tissue by the ramp excitation.

Authentication Based on Pole-zero Models of Signature Velocity.

With the increase of communication and financial transaction through internet, on-line signature verification is an accepted biometric technology for access control and plays a significant role in authenticity and authorization in modernized society. Therefore, fast and precise algorithms for the signature verification are very attractive. The goal of this paper is modeling of velocity signal that pattern and properties is stable for persons. With using pole-zero models based on discrete cosine transform, precise method is proposed for modeling and then features is founded from strokes. With using linear, parzen window and support vector machine classifiers, the signature verification technique was tested with a large number of authentic and forgery signatures and has demonstrated the good potential of this technique. The signatures are collected from three different database include a proprietary database, the SVC2004 and the Sabanci University signature database benchmark databases. Experimental results based on Persian, SVC2004 and SUSIG databases show that our method achieves an equal error rate of 5.91%, 5.62% and 3.91% in the skilled forgeries, respectively.