Microvessel diameter estimation: Error bias correction of serial measurements

Abstract

The assessment of vessel patency can be substantially improved by serial microvessel diameter measurements taken successively along an extensive length of the vessel. It is possible to avoid making the a priori assumptions about the existence or location of local constriction sites implicit in single diameter measurements. The problem then becomes one of making sense of tens or hundreds of measurements for each vessel. Equivalent diameter is defined here as as the diameter of a uniform circular cylinder of the same length as the original vessel, and having the same total resistance. Direct computation of the equivalent diameter, without taking measurement errors into account, leads to an underestimation of the true equivalent diameter even if the individual diameter measurements were not biased. We have developed a method for effectively eliminating this bias. It has been applied to serial microvessel diameter measurements of the guinea pig cochlea, automatically measured using an image analysis system. In this report, the results were developed for diameter estimates with an approximate gaussian distribution; however the method is readily extended to other error distributions. Convergence of the bias compensation was rapid. Use of the new method is advisable with as few as three diameter estimates per vessel.