Dale H. Mugler

Ballistocardiogram Artifact Removal in EEG-fMRI Signals Using Discrete Hermite Transforms

Simultaneously recorded electroencephalogram (EEG) and functional magnetic resonance imaging (fMRI) is rapidly emerging as a powerful neurophysiological research and clinical tool. However, the quality of the EEG, recorded in the MRI scanner, is affected by the ballistocardiogram (BCG), which is an artifact related to the cardiac cycle. The BCG has a complete spectral overlap with the EEG and is nonstationary over time, making its suppression a signal processing challenge. We propose a novel method for the identification and suppression of this artifact using shape basis functions of the new dilated discrete Hermite transform. The BCG artifacts are modeled continuously, using these discrete Hermite basis functions and are subsequently subtracted from the ongoing EEG. Experimental EEG data was recorded within and outside a 3 Tesla MRI scanner, from a total of 6 subjects under a variety of experimental conditions. The efficiency of this algorithm was quantitatively assessed by adding known BCG templates, at varying Signal to Noise Ratios (SNRs), to EEG recorded outside the scanner. Significant suppression of the BCG artifact (p<0.05) was achieved without distorting the underlying EEG. Using EEG data recorded inside the MR scanner, this method was compared with existing BCG artifact removal techniques and its performance was found to be superior to the Average Artifact Subtraction (AAS) method and comparable to the Independent Component Analysis (ICA) based methods. The computational simplicity of this technique allows for real time implementation.

A robust DSP integrator for accelerometer signals

The need for an accepted method for determining position data from digital accelerometer readings with known frequency range is very important. The method of this paper uses spectral information and provides more stability and accuracy than classic methods for the DSP case. It even reduces to a classic method for the nonoscillatory case. Examples show the robustness of the method, overcoming the instability apparent with other integration methods.

Computationally efficient linear prediction from past samples of a band-limited signal and its derivative

Formulas for linear prediction of a band-limited signal are developed, where the signal may be either deterministic or wide-sense stationary. The prediction formulas are based on finite differences modified by two or more parameters, and finite differences allow the formulas to be easily adapted to changes in order of the prediction. It is shown that a formula to predict the next signal value from a set of past, equally spaced values is a formula that can be extended to provide prediction at points even beyond that. In addition, the formula is extended to a difference scheme involving an arbitrary number of parameters as well as to a formula that includes samples of the derivative of the signal. This approach differs from that of solving the normal (or Yule-Walker) equations, but it has the advantage that the (suboptimal) prediction coefficients are independent of the particular signal spectrum or autocorrelation function. >

VLSI Architectures for the 4-Tap and 6-Tap 2-D Daubechies Wavelet Filters Using Algebraic Integers

This paper proposes a novel algebraic integer (AI) based multi-encoding of Daubechies-4 and -6 2-D wavelet filters having error-free integer-based computation. Digital VLSI architectures employing parallel channels are proposed, physically realized and tested. The multi-encoded AI framework allows a multiplication-free and computationally accurate architecture. It also guarantees a noise-free computation throughput the multi-level multi-rate 2-D filtering operation. A single final reconstruction step (FRS) furnishes filtered and down-sampled image outputs in fixed-point, resulting in low levels of quantization noise. Comparisons are provided between Daubechies-4 and -6 designs in terms of SNR, PSNR, hardware structure, and power consumptions, for different word lengths. SNR and PSNR improvements of approximately 30% were observed in favour of AI-based systems, when compared to 8-bit fixed-point schemes (six fractional bits). Further, FRS designs based on canonical signed digit representation and on expansion factors are proposed. The Daubechies-4 and -6 4-level VLSI architectures are prototyped on a Xilinx Virtex-6 vcx240t-1ff1156 FPGA device at 282 MHz and 146 MHz, respectively, with dynamic power consumption of 164 mW and 339 mW, respectively, and verified on FPGA chip using an ML605 platform.

A fast adaptive filter for electrocardiography

A new adaptive filter structure is proposed to reduce power line interference from ECGs. It is based on the principle of continuously tracking the frequency, amplitude and phase of the noise, using a modified form of the short time Fourier transform. This information is used to reconstruct the noise signal, which is then subtracted from the noisy ECG. The method was tested using ECGs from the MIT-BIH database with superimposed noise of varying amplitude and frequency (60 Hz +/- 5 Hz). It was found to be efficient even at low sampling rates. The process is computationally simple enough for real time implementation. It also obviates the need for a reference input. This technique can be applied to many other bipotential signals, and can be adapted for hardware implementation.

Real time monitoring of ischemic changes in electrocardiograms using discrete Hermite functions

A novel scheme for real time detection of ischemic features from long term electrocardiograms (ECG), based on the dilated discrete Hermite expansion is proposed. The discrete Hermite functions used for the expansion are eigenvectors of a symmetric tridiagonal matrix that commutes with the centered Fourier matrix. The ECG signals were expanded in terms of Hermite functions using a simple dot product. The resulting coefficients were found to have details about the shape of the ECG signal. The first 50 coefficients had all sufficient information to reconstruct the ECG signal with acceptable percentage RMS difference (PRD). A committee neural network classifier with these 50 input parameters was trained to identify ischemic features, namely ST segment and T wave changes. A sensitivity of 98% and a specificity of 97.3% were achieved. A comparison of these figures with other contemporary classification schemes revealed a superior performance.

A weighted k-nearest neighbor method for gene ontology based protein function prediction

Numerous genome projects have produced a large and ever increasing amount of genomic sequence data. However, the biological functions of many proteins encoded by the sequences remain unknown. Protein function annotation and prediction become an essential and challenging task of post-genomic research. In this paper, we present an automated protein function prediction system based on a set of proteins of known biological functions. The functions of the proteins are characterized with gene ontology (GO) annotations. The prediction system uses a novel measure to calculate the pair-wise overall similarity between protein sequences. The protein function prediction is performed based on the GO annotations of similar sequences using a weighted k-nearest neighbor method. We show the prediction accuracies obtained using the model organism yeast (Sacchyromyces cerevisiae). The results indicate that the weighted k-nearest neighbor method significantly outperforms the regular k-nearest neighbor method for protein molecular function prediction.In tennis competition, there are some reasons leading to the competition results inaccurately. So I put forward a real-time competition simulation system to solve the tennis problem. It can overcome the limitations and the blind spots that occur in human observation. The system includes four parts: image collection, moving object detection and tracking, static scene simulation, competition simulation. This paper presents an algorithm for the detection and tracking of moving objects in sequence images that can be applied in simulations of tennis competition. The approach differs from most other methods that solve the problems of object occlusion and object deformation. The experimental results proved that our method is feasibility and usefulness.

Linear prediction from samples of a function and its derivatives

Some formulas for the prediction of the values of a band-limited function based on its samples from the past are generalized by including past samples of its first derivative. The new sums, developed by an approach based on Newton series, make it possible to double the distance between the sample points. The resulting formulas are shown to apply to the prediction problem for a large class of entire functions of exponential type. In addition, a related prediction formula which uses past samples of successively higher derivatives is shown to behave similarly to the Taylor series approximation, again for a class of functions that includes the band-limited functions.

The Centered Discrete Fourier Transform and a parallel implementation of the FFT

This paper describes a novel method for the computation of the Discrete Fourier Transform (DFT). The development of a truly centered DFT is coupled with a method for computing the Centered DFT to provide an FFT that requires no complex multiplications and which allows a highly parallel implementation.

Adaptive filtering of ballistocardiogram artifact from EEG signals using the dilated discrete hermite transform

Electroencephalogram (EEG) signals, when recorded within the strong magnetic field of an MRI scanner are subject to various artifacts, of which the ballistocardiogram (BCG) is one of the prominent ones affecting the quality of the EEG. The BCG artifact varies slightly in shape and amplitude for every cardiac cycle making it difficult to identify and remove. This paper proposes a novel method for the identification and elimination of this artifact using the shape basis functions of the new dilated discrete Hermite transform. In this study, EEG data within and outside the scanner was recorded. On removal of the BCG artifact for the EEG data recorded within the scanner, a significant reduction in amplitude at the frequencies associated with the BCG artifact was observed. In order to quantitatively assess the efficacy of this method, BCG artifact templates were added to segments of EEG signals recorded outside the scanner. These signals, when filtered using the proposed method, had no significant difference (p<0.05) from the original signals, indicating that the technique satisfactorily eliminates the BCG artifact and does not introduce any distortions in the original signal. The method is computationally efficient for real-time implementation.