Edge detection at low resolution using Padé approximants.

Many applications of MRI involve image segmentation for quantitation and other purposes. Extraction of salient structures in images often relies on use of image thresholds or direct edge detection both of which benefit from high-resolution data to achieve reliable and accurate results. Threshold based measurements are also often highly dependant on the threshold level chosen and this dependence is accentuated at lower resolution. In conventional MRI, acquiring higher resolution data increases the scan time due to the serial nature of the phase encoding process. This reduces the achievable temporal resolution and increases the risk of motion artefact. Here we present a method for automatic edge detection based on Pade approximants, which does not require any thresholding and, more importantly, can be preformed with reduced data sets. This method was tested using cardiac data since delineation of the boundary between the myocardium and the blood pool is often required in the assessment of cardiovascular function and since cardiac MR has added time constraints that can constitute a critical limiting factor.