Arunava De

Grammatical Swarm Based-Adaptable Velocity Update Equations in Particle Swarm Optimizer

In this work, a new method for creating diversity in Particle Swarm Optimization is devised. The key feature of this method is to derive velocity update equation for each particle in Particle Swarm Optimizer using Grammatical Swarm algorithm. Grammatical Swarm is a Grammatical Evolution algorithm based on Particle Swarm Optimizer. Each particle updates its position by updating velocity. In classical Particle Swarm Optimizer, same velocity update equation for all particles is responsible for creating diversity in the population. Particle Swarm Optimizer has quick convergence but suffers from premature convergence in local optima due to lack in diversity. In the proposed method, different velocity update equations are evolved using Grammatical Swarm for each particles to create the diversity in the population. The proposed method is applied on 8 well-known benchmark unconstrained optimization problems and compared with Comprehensive Learning Particle Swarm Optimizer. Experimental results show that the proposed method performed better than Comprehensive Learning Particle Swarm Optimizer.

Masking Based Segmentation of Diseased MRI Images

We have devised a new technique to segment an diseased MRI image wherein the diseased part is segregated using a masking based thresholding technique together with entropy maximization. The particle swarm optimization technique (PSO) is used to get the region of interest (ROI) of the MRI image. The mask used is a variable mask. The rectangular mask is grown using an algorithm provided in the subsequent sections using similarity of neighbourhood pixels. Tests on various diseased MRI images show that small diseased objects are successfully extracted irrespective of the complexity of the background and difference in intensity levels and class sizes. Previous works are based on bimodal images whereas our work is based on multimodal images.

Entropy maximization based segmentation, transmission and Wavelet Fusion of MRI images

A method of progressive transmission of Magnetic Resonance Image with lesions over long distances is proposed. The Magnetic Resonance Images at the transmitter end are segregated on the basis of presence of lesions. Entropy Maximization using Hybrid Particle Swarm Optimization algorithm that incorporates a Wavelet theory based mutation operation is used for segmentation of Magnetic Resonance Images. It applies the Multi-resolution Wavelet theory to overcome the stagnation phenomena of the Particle Swarm Optimization. Thus the segmentation algorithm using Hybrid Particle Swarm Algorithm explores the solution space more effectively for a better solution. Varying percentages of Discrete Cosine Transform coefficients of segmented Magnetic Resonance Images are used for progressive image transmission. For a particular image data, the progressively received images are of different resolutions. At the receivers end the progressively received images of different resolutions are fused using Multi-resolution wavelet analysis to get a visually suitable image for diagnosis. The doctor or a radiologist identifies a particular class of image with lesions and may ask for the entire un-segmented Magnetic Resonance Image dataset of a particular patient for further diagnosis. The proposed system helps to reduce the load on the system by choosing not to transmit the Magnetic Resonance Images without lesions.

MRI segmentation using Entropy maximization and Hybrid Particle Swarm Optimization with Wavelet Mutation

A Hybrid Particle Swarm Optimization algorithm that incorporates a Wavelet theory based mutation operation is used for segmentation of Magnetic Resonance Images. We use Entropy maximization using Hybrid Particle Swarm algorithm with Wavelet based mutation operation to get the region of interest of the Magnetic Resonance Image. It applies the Multi-resolution Wavelet theory to enhance the Particle Swarm Optimization Algorithm in exploring the solution space more effectively for a better solution. Tests on various MRI images with lesions show that lesions are successfully extracted.

Brain MRI segmentation for tumor detection via entropy maximization using Grammatical Swarm

This paper presents a new method for the segmentation of Magnetic Resonance Imaging (MRI) of brain tumor. First, discrete wavelet transform (DWT)-based soft-thresholding technique is used for removing noise in the MRI. Second, intensity inhomogeneity (IIH) independent of noise is removed from the MRI image. Third, again DWT is used to sharpen the de-noised and IIH corrected image. In this method, the image is decomposed into first level using wavelet decomposition and approximate values are assigned to zero and reconstruct the image results in detailed image. The detailed image is added with the pre-processed image to produce sharpened image. Entropy maximization using Grammatical Swarm (GS) algorithm is used to obtain a set of threshold values and a threshold value is selected with the expert knowledge to separate the lesion part from the other non-diseased cells in the image.

Detection of Progression of Lesions in MRI Using Change Detection

Change detection is a process of identifying the changes in a state of an object over time. We use the phenomena of change detection to detect the changes occurring in MRI of brain having cancerous and non cancerous lesions. A Hybrid Particle Swarm Optimization algorithm that incorporates a Wavelet theory based mutation operation is used for segmentation of lesions in Magnetic Resonance Images. The segmented lesions are the Region of Interest. This method of using change detection algorithm would be helpful in detecting changes in Region of Interests of MRI with lesions and also to view the progress of treatment for cancerous lesions.

MRI brain lesion segmentation using generalized opposition-based glowworm swarm optimization

An improved glowworm swarm optimization algorithm with generalized opposition-based learning is proposed in this paper and is used in segmentation for magnetic resonance images. Noises are removed and intensity inhomogeneities are corrected in the MR images. Next, a clustering technique with glowworm swarm optimization algorithm with generalized opposition based learning is used. Finally, lesions are separated from the normal tissues of the brain in the post-processing step. The performance of the proposed methodology based on both numerical and visual results are compared with K-means and particle swarm optimization based methodologies over two sets of MR images. The experimental results demonstrate that the proposed methodology statistically outperforms other methodologies.

MRI Skull Bone Lesion Segmentation Using Distance Based Watershed Segmentation

The objective of separating touching objects in an image is a very difficult task. The task is all the more difficult when the touching objects are healthy tissues and unhealthy tissues of lesions in human brain.

Registration of progressively transmitted MR with lesions in brain

A method of registration of progressively transmitted MR images with lesions are proposed. The MR image is segmented using Entropy Maximization together with Hybrid Particle Swarm Optimization that incorporates Wavelet Mutation operation. The segmented MR image is progressively received at the receivers end. Registration is used to ensure correct transmission and fusion of segmented MR image with lesions.

Entropy Maximization, Stationary Wavelet and DCT Based Segmentation, De-Noising and Progressive Transmission Technique for Diseased MRI Images

We have devised a way of segmentation and progressive transmission of diseased MRI images. We use Particle Swarm Optimization (PSO) to get the region of interest (ROI) of the diseased MRI image. We use the concept of Multi-resolution Wavelet analysis to de-noise the ROI. We use Stationary Wavelet Transform together with Soft Thresholding Technique for de-noising purpose. A variable mask is used to get the segmented image. Varying percentages of DCT coefficients are used for progressive transmission of the diseased MRI image. Clustering of the images using K-Means algorithm result in predominantly two cluster namely that of diseased cells and background. Test on various MRI images show that the small diseased objects are successfully extracted irrespective of the complexity of the background and difference in intensity levels and class sizes. The proposed method only transmits the diseased MRI for further diagnosis of the disease and treatment.