Distributional error, image analysis

Slouf et al. [37] reported a novel automated quantitative analysis technique, SEMq, which was based on a light scattering theory. In this method automated image analysis of electron micrographs is used to determine the total amount of wear particles released into the tissues around total joint replacements. This technique was shown to provide rapid, reliable, and reproducible results, which indicated that the distribution of wear particles around TJRs is nonhomogeneous [37]. However, this technique does not calculate absolute numbers of wear particles but rather determines relative numbers of wear particles in one sample compared to another. The technique may also be sensitive to errors introduced by particle cliunping and protein contamination of filters used for SEM analysis. 

We have described the construction of continuous images from the data of a coded aperture instrument with mask minus anti-mask background subtraction and a continuous detector. We have assumed that the mask used is an HURA, but simular results should hold for all masks with good imaging properties. The error analysis presented shows the the importance of correctly matching mask cell size to detector position resolution, and reveals that a trade-off must be made between flux sensitivity and source localization ability. Our results on the distribution of noise peaks in images can be used in setting realistic thresholds for source detection. 

The localization of structural features in the electron probe is aided by scanning techniques " that produce electron microprobe images of small sectors of the sample surface. It is also possible to obtain scanning images of element distribution. The most attractive feature of electron probe microanalysis is the fact that a quantitative analysis is possible, with errors of less than 3 % relative in most cases. Data evaluation requires the use of a computer, but the... 

---