Syed Muhammad Monir

Denoising functional magnetic resonance imaging time-series using anisotropic spatial averaging

Abstract We propose a novel iterative scheme for adaptive smoothing of functional MR images. The method estimates a signal model at every voxel in the time-series, which is subsequently used in determining the weights of the smoothing kernel. The method does not require any information about the test hypothesis and is well-suited as a preprocessing step for both hypothesis-driven and data-driven analysis techniques. We demonstrate the performance of the proposed method by applying it to preprocess both simulated and real fMRI data. The method is found to effectively suppress the noise while preserving the shapes of the active brain regions.

Granger causality: Comparative analysis of implementations for Gene Regulatory Networks

Granger Causality (GC) is an effective tool for determining functional connectivity in time-series data. However, application of GC is limited by the curse of dimensionality in many applications, e.g. Gene Regularity Networks (GRN). Various methods have been proposed to overcome this limitation. To the best of our knowledge, there is no detailed comparative study of such methods. We aim to perform a detailed comparative study of a few of such methods using different statistical measures under various constraints.

Iterative adaptive spatial filtering for noise-suppression in functional magnetic resonance imaging time-series

We present an iterative scheme for adaptive smoothing of functional magnetic resonance images. We propose a novel similarity measure to estimate the weights of the smoothing filter based on the functional similarity of the voxels under the smoothing kernel with the voxel under consideration as well as their similarity with a reference time‐course representing the expected BOLD response. We demonstrate the performance of the proposed method by applying the method to preprocess both simulated and real fMRI data. The method improves the functional SNR of the data while preserving the shapes of the functionally active region and its performance is not compromised when structured noise is the dominant noise source.

Performance analysis of Viterbi decoder using a DSP technique

Increasing the speed of the wireless communication requires a reliable solution for data transfer. The signal to noise ratio (SNR) of the channel in digital wireless communication is one of the major limitations on the operating performance. To enhance the performance, solution in terms of coded data and error-correcting code has been introduced. Viterbi decoder is one of the techniques used for this purpose and Viterbi algorithm is used for decoding. This algorithm is an extremely fast and efficient method of decoding the coded data from the channel. In this paper, the structure of baseband processing unit and implementation of convolutional encoder and Viterbi decoder is described. The convolutional encoder of rate 1/2 and constraint length 3 and Viterbi decoder of rate 1/2 and constraint length 3 using a TMS320C54 DSP chip is designed.

A Framework for Interactive Content-Based Image Retrieval

With the exponential increase in the size of digital image databases in the past few years, traditional way of manually annotating the images with text and then using text-based queries for image retrieval has been giving way to content-based image retrieval (CBIR) systems that use the visual contents of the images to automatically index and retrieve digital images. However, there is always a gap between high-level human perception of an image and the low-level image features used to describe its contents. This gap between low-level image features and semantic image content is the major bottleneck faced by traditional CBIR systems. Modern CBIR systems overcome this problem by using interactive learning, bringing the user in the loop. Such systems learn from feedbacks given by the user about the relevance or irrelevance of the current retrieval results. This paper presents a framework for interactive content-based image retrieval. Considering relevance feedback as a learning problem, a learning machine based on radial basis functions (RBF) neural networks (NN) is implemented and the system has been tested for its effectiveness using a database of 10,000 images

Implementation of Viterbi Decoder for WCDMA System

This paper describes the implementation of soft-decision Viterbi decoder on TMS320C62timesDSP. In this work a soft-decision Viterbi decoder is implemented on code rate 1/3 and constraint length 9. Originally the aim is to achieve a decoding rate of 32Kbps as per WCDMA IS-95 standard recommendation, which also employs a convolutional encoder with 8 trellis states and Viterbi decoder with 256 states in the trellis. Presently there are numerous standards for digital cellular systems, especially when considering the third generation system. Each of these contains operation modes utilizing convolutional coding which is usually decoded via Viterbi decoder. In each standard there are different formulations of this coding with each form imparting requirements upon the decoder. This gives rise the need for Viterbi decoders with highly flexible capabilities. Realizing such flexibilities in DSP software is straightforward. This paper presents a flexible Viterbi decoder constructed to operate as a loosely coupled coprocessor for the DSP. Single shift register convolutional codes with 1/n code rates are targeted. Some application of the implemented decoder and the bottlenecks in the implementation have also been discussed

Overcoming the ill-balanced data problem in functional MRI clustering analysis

In functional magnetic resonance imaging (fMRI) data, activated voxels are usually very small in number and are embedded in a mass of inactive voxels. For clustering analysis, this situation generates an ill-balanced data problem among different classes of voxels. In this paper we propose a novel method to overcome the ill-balanced data problem, by reducing the number of voxels to be processed by the clustering algorithm. We divide the functional data into small overlapping regions and decide the presence or absence of functional activity in a region, on the basis of condition number of a matrix constructed from the feature vectors of the voxel in that region. Only the regions that are potentially active are retained for clustering analysis. Results are presented for both simulated and real data and advocate that the proposed method effectively solves the ill-balanced data problem.

Data-driven analysis of functional MRI time-series using a region-growing approach

We present a data-driven method to analyze functional magnetic resonance imaging (fMRI) time-series where multiple hypotheses are generated for inferential methods from the data itself without any assumptions on the time-series. The method does not require the number of clusters to be defined a priori. Activation detection is based on region growing which specifically suits the spatiotemporal characteristics of fMRI data. Results presented for simulated as well as real fMRI data show that the proposed method efficiently segments fMRI data into regions of distinct functional activity.

Noise suppression in functional MRI data using anisotropic spatial averaging

We present an adaptive smoothing scheme for denoising functional magnetic resonance imaging (fMRI) data using weighted average filtering. A novel metric is proposed that assigns the weights of the smoothing kernel on the basis of similarity of the voxels under the smoothing kernel with the voxel under consideration as well as a reference time course. Pearsons coefficient of correlation is used as the similarity measure. The performance of this simple yet effective smoothing scheme is tested by applying it on both synthetic and real fMRI data. The method is found to be effective in suppressing random noise while preserving the shapes of the activated regions.

Iterative adaptive filtering for random noise reduction in functional MRI time-series

This paper presents a novel method for adaptive filtering of functional magnetic resonance imaging (fMRI) time-series. The method progressively reduces noise from the fMRI time courses based on selective spatial averaging of the underlying voxels. A new similarity measure is proposed to assign the weights of the averaging kernel. The performance of the proposed method is verified by its application on synthetic as well as real fMRI data. The results show that pre-processing the data with the proposed method results in an increased sensitivity along with an excellent specificity of fMRI analysis.