Search

In this paper, we introduce a modification of the popular U-Net neural network architecture for medical image segmentation. Our proposed architecture replaces the concatenation operations in the traditional U-Net's skip connections with content-adaptive convolution, thereby significantly reducing the number of parameters of the network. Our experiments on two segmentation tasks - cell nuclei segmentation, and pneumo-thorax segmentation - demonstrated that the modified architecture achieves higher segmentation accuracy compared to the original U-net.

It is possible to improve neuronavigation during image-guided surgery by warping the high-quality preoperative brain images so that they correspond with the current intraoperative configuration of the brain. In this paper, the accuracy of registration results obtained using comprehensive biomechanical models is compared with the accuracy of rigid registration, the technology currently available to patients. This comparison allows investigation into whether biomechanical modeling provides good-quality image data for neuronavigation for a larger proportion of patients than rigid registration. Preoperative images for 33 neurosurgery cases were warped onto their respective intraoperative configurations using both the biomechanics-based method and rigid registration. The Hausdorff distance–based evaluation process, which measures the difference between images, was used to quantify the performance of both registration methods. A statistical test for difference in proportions was conducted to evaluate the null hypothesis that the proportion of patients for whom improved neuronavigation can be achieved is the same for rigid and biomechanics-based registration. The null hypothesis was confidently rejected (p < 10−4). Even the modified hypothesis that fewer than 25% of patients would benefit from the use of biomechanics-based registration was rejected at a significance level of 5% (p = 0.02). The biomechanics-based method proved particularly effective in cases demonstrating large craniotomy-induced brain deformations. The outcome of this analysis suggests that nonlinear biomechanics-based methods are beneficial to a large proportion of patients and can be considered for use in the operating theater as a possible means of improving neuronavigation and surgical outcomes.

It is possible to improve neuronavigation during image-guided surgery by warping the high-quality preoperative brain images so that they correspond with the current intraoperative configuration of the brain. In this work, the accuracy of registration results obtained using comprehensive biomechanical models is compared to the accuracy of rigid registration, the technology currently available to patients. This comparison allows us to investigate whether biomechanical modeling provides good quality image data for neuronavigation for a larger proportion of patients than rigid registration. Preoperative images for 33 cases of neurosurgery were warped onto their respective intraoperative configurations using both biomechanics-based method and rigid registration. We used a Hausdorff distance-based evaluation process that measures the difference between images to quantify the performance of both methods of registration. A statistical test for difference in proportions was conducted to evaluate the null hypothesis that the proportion of patients for whom improved neuronavigation can be achieved, is the same for rigid and biomechanics-based registration. The null hypothesis was confidently rejected (p-value

Knowledge of the mechanical properties of the brain-skull interface is important for surgery simulation and injury biomechanics. These properties are known only to a limited extent. In this study we conducted in situ indentation of the sheep brain, and proposed to derive the macroscopic mechanical properties of the brain-skull interface from the results of these experiments. To the best of our knowledge, this is the first ever analysis of this kind. When conducting in situ indentation of the brain, the reaction force on the indentor was measured. After the indentation, a cylindrical sample of the brain tissue was extracted and subjected to uniaxial compression test. A model of the brain indentation experiment was built in the Finite Element (FE) solver ABAQUS™. In the model, the mechanical properties of the brain tissue were assigned as obtained from the uniaxial compression test and the brain-skull interface was modeled as linear springs. The interface stiffness (defined as sum of stiffnesses of the springs divided by the interface area) was varied to obtain good agreement between the calculated and experimentally measured indentor force-displacement relationship. Such agreement was found to occur for the brain-skull interface stiffness of 11.45 Nmm⁻¹/mm². This allowed identification of the overall mechanical properties of the brain-skull interface.

In this paper we evaluate the accuracy of warping of neuro-images using brain deformation predicted by means of a patient-specific biomechanical model against registration using a BSpline-based free form deformation algorithm. Unlike the BSpline algorithm, biomechanics-based registration does not require an intra-operative MR image which is very expensive and cumbersome to acquire. Only sparse intra-operative data on the brain surface is sufficient to compute deformation for the whole brain. In this contribution the deformation fields obtained from both methods are qualitatively compared and overlaps of Canny edges extracted from the images are examined. We define an edge based Hausdorff distance metric to quantitatively evaluate the accuracy of registration for these two algorithms. The qualitative and quantitative evaluations indicate that our biomechanics-based registration algorithm, despite using much less input data, has at least as high registration accuracy as that of the BSpline algorithm.

This paper compares the warping of neuro-images using brain deformation predicted by means of patient-specific biomechanical model with the neuro-image registration using BSpline-based free form deformation algorithm. Deformation fields obtained from both algorithms are qualitatively compared and overlaps of edges extracted from the images are examined. Finally, an edge-based Hausdorff distance metric is defined to quantitatively evaluate the accuracy of registration for these two algorithms. From the results it is concluded that the patient-specific biomechanical model ensures higher registration accuracy than the BSpline registration algorithm.

The hand shape authentication has been widely used on the biometrics field with an extensive range of potential applications since it is the well-suited with respect to security facility. The aim of this paper is to improve the accuracy and robustness of the hand based authentication system. In this paper, a hand shape authentication with a fixed angle radon transformation and a centroid scheme is proposed to achieve high accuracy and robustness with translation and rotation invariancy. Furthermore, the hand shape feature vector is developed using the radon transform with centroid to represent hand geometry, avoiding the more complicated and liable to errors of hand landmark extraction. This scheme can handle variability of hand's position, translation and rotation, especially in a peg free system with the help of centroid and radon transformation. The experimental results are promising and confirm the usefulness of the proposed approach for personal authentication by implementing the real hand geometry verification/identification system and it has proven to work effectively and competitively with low false acceptance and false rejection rates.

Estimation of the symbol rate has important applications in receiver synchronization for symbol time recovery. In this paper the problem is investigated using Smoothen Non-Linear Energy Operator (SNEO). Unlike wavelet based methods in [2], [3], [4] the proposed algorithm is completely blind because it does not require any priory information regarding the modulation type or carrier frequency. Moreover, the proposed algorithm is computationally efficient. Simulation results also proof the effectiveness of the proposed algorithm.

Biometric authentication refers to the automatic verification of a person's identity from physiological or behavioral characteristics presented by him or her. In this paper an authentication scheme from hand images is presented. Instead of dealing with hand measurements, typically termed as ‘hand geometry’, this method verifies with entire hand shape. Peg free and position invariant features are calculated using Radon Transform. Low resolution hand images captured by a document scanner are processed to extract feature vectors. The proposed scheme is tested on a data set of 136 images with simple Euclidian norm based match score. The method attained an Equal Error Rate (EER) of 5.1%.

A new parameter estimation method for linear chirp signal is proposed. This method utilizes the Discrete Fractional Fourier Transform (DFRFT) along with Bowtie Chirplet Transform to estimate the amplitude and phase parameters of multi-component chirp signals. DFRFT is used for estimating amplitude parameter whereas the chirp rate and initial frequency are estimated using Chirplet Transform. Performance of the proposed method in noise is analyzed by Monte- Carlo simulation for different Signal to Noise Ratio (SNR). Good performance was achieved at SNR as low as -10 dB.

Detection of gait characteristics has found considerable interest in fields of biomechanics and rehabilitation sciences. In this paper an approach for abnormal gait detection employing Discrete Fourier Transform (DFT) followed by Discrete Wavelet transform (DWT) analysis has been presented. The joint angle characteristics in frequency domain have been analyzed and using the harmonic coefficient recognition for abnormal gait has been performed. A classification of gaits has been attempted using k-means clustering based on the data acquired from DFT and DWT. Future work will be the expansion of the detection introduced in this system to include abnormality detection instead of just an abnormal or normal detection that would prove to be a valuable addition for use in a variety of applications including unobtrusive clinical gait analysis, automated surveillance etc.

Research in person identification and authentication has attracted significant attention from the researchers and scientists. This paper presents a biometric user authentication based on a person's foot step. The advantage of this recognition method over other types of biometrics is that it enables unobtrusive user authentication where other types of biometrics are not available. Firstly the ground reaction force data was extracted using force plate to gather ground reaction force for individuals. Later we utilized the discrete wavelet transform to de-noise the experimental data and in the final step, histograms were used to identify different person's foot step. The experimental results show improvements in identification accuracies compared to previously reported work.

A new parameter estimation method for linear amplitude modulated chirp signal is proposed. This method utilizes the discrete fractional Fourier transform (DFRFT) along with Radon Wigner transform (RWT) to estimate the amplitude and phase parameters of multi-component chirp signals. Expressions of DFRFT for estimating amplitude and initial phase parameters are derived. The chirp rate and initial frequency is estimated using RWT. Performance of the proposed method in noise is analyzed by Monte-Carlo simulation for different signal to noise ratio (SNR). Better performance was achieved at SNR as low as -10 dB.

Recently face recognition has attracted significant attention from the researchers and scientists in various fields of research, such as biomedical informatics, pattern recognition, vision, etc due its applications in commercially available systems, defense and security purpose Face recognition presents a very challenging problem in real application in computer vision and pattern recognition due to variation of face. A large number of face recognition algorithms, along with their modifications are available over the past three decades. In this paper a practical method for face reorganization utilizing head cross section data based on Procrustes analysis is proposed. Firstly, a number of head cross section data were extracted from 3D head scanner along sagittal plane for eight different subjects. After extracting 3D head cross section data a comparison analysis were performed utilizing Procrustes distance to differentiate their face pattern from each other. The performance analysis of face recognition was analyzed based on K nearest neighbor classifier. The experimental results presented here verify that the proposed method is considerable effective.