Page 40 - Advanced concepts in orbital wall fractures

38
Chapter 2
method. The anterior plane was used to separate excess volume outside the bony orbit in all segmentations in order to be able to compare them. This eliminated any doubts about volume differences at the anterior boundary.
Gold standard
In the literature, two methods are suggested to be the gold standard for measuring orbital volume: slice-by-slice manual segmentation of CT scans and the use of different kinds of filling materials (beads, silicone, and water) for the enucleated orbits of cadavers20. Both methods have their advantages and disadvantages. The advantage of the manual segmentation is that only a CT scan is required to measure the volume. The disadvantage is that it still is an observer-dependent process, and therefore, it is subject to discrepancies in assessment between observers. The filling method has the advantage that a real volume is measured. The disadvantage is that it is difficult to contour the anterior border of the orbit, which means that it is practically impossible to measure the exact orbital volume. Apart from this, the method can only be used in anatomical specimen and is useless in a clinical situation. In this study, the manual segmentation method was used as the gold standard. The reproducibility of the method investigated was validated and demonstrated sufficient high correlation for both interobserver and intraobserver measurements. Trauma scans were used on purpose to mimic the clinical preoperative setting. This means that the patient was not always scanned in a well- aligned position. The agreement may have been even higher when scan data of properly aligned patients had been used.
(Semi-)automatic methods
As mentioned in the introduction, 2 cc increase of orbital volume leads to 2 mm of enophthalmos on average, which is considered to be clinically significant. Accuracy and reproducibility should be well within those limits to prevent measuring errors from contributing to poor surgical outcome due to planning. In the past two decades, several (semi-)automatic software methods have been tested with varying results5,21-23. This is partly due to the differences in choice of gold standard, approach, and study design. It is difficult to compare results from these studies. A study by Deveci et al.