| Summary: | The use of parametric modelling approach for Magnetic Resonance Imaging (MRI) reconstruction has been shown to produce images with higher resolution compared to the use of Fast Fourier Transform (FFT) technique. Despite this success, two problems lessen the use of this technique, these are: non availability of optimal method of estimating model order and the model coefficients determination.
In this research work, a new method of Autoregressive Moving Average (ARMA) coefficients using three layer complex valued neural network ARMA techniques (CVNN-CARMA) with split complex-value weight and adaptive linear activation functions is hereby proposed. The proposed model coefficients determination in conjunction with various methods of optimal model order determination were then applied on MRI data using both Transient Error Reconstruction Algorithm (TERA) and modified Transient Error Reconstruction Algorithm to obtain images with improved resolution.
Future work include extending this modelling method to two dimensional domain, evaluating the performance of the proposed CVNN-CARMA and using a trained artificial neural network to automatically obtain the model order of a complex valued data.
Keywords: Autoregressive Model Algorithm (ARMA), Magnetic Resonance Imaging (MRI),
Reconstruction
|
|---|