Potential errors due to aliasing in digital video analysis of quantitative autoradiography

Abstract

Digital image analysis of quantitative autoradiographic (QAR) films is widely used in neuroscience applications. Unless proper precautions are taken when autoradiographic images are converted to digital form they can be inadvertently modified by improper application of the sampling process. This type of modification is termed aliasing error and can cause nonexistent structures to appear in the reconstructed digital image, changing the apparent optical density values of the data. The theoretical basis of aliasing error is presented, along with examples of aliasing from optical resolution test patterns and 2-deoxy[ 14C]glucose (2-DG) experimental QAR images. We show that aliasing can change the apparent shape of structures, as well as the derived values obtained from QAR experiments. In an example with experimental 2-DG images, aliasing in the cerebellar cortex consistently underestimates tissue radioactivity levels in gray matter ( P < 0.001) and overestimates levels in adjacent white matter ( P < 0.001). Additional data transformations, such as the equations used for blood flow or glucose utilization, can, somewhat unpredictably, accentuate the errors introduced by aliasing. We present a discussion of the autoradiographic image features and electronics design that play a role in introducing aliasing errors and means by which aliasing can be recognized and minimized.