Page 42 - Advanced concepts in orbital wall fractures

40
Chapter 2
Differences in volume between the (semi-)automatic methods and the gold standard are greatly influenced by differences in defining the border of the inferior orbital fissure (Fig. 6). To overcome this repeating error within the automatic method, a manual adjustment was introduced in method SAA. It was thought that this would correct overestimation and prevent large errors. Unfortunately, it consistently produced an underestimation and had poorer accuracy with a mean difference of 0.86 cc (SD 0.27) compared to the gold standard. The reproducibility of this method was acceptable, but worse than the other two (semi-) automatic methods. Furthermore, method SAA is more time-consuming with average time of 327 s (SD 36.2). The semi-automatic method without manual adjustments proved to be accurate with an average difference of 0.24 cc (SD 0.27) compared to the gold standard.
The distance map results of the mean distance measure between method SA versus gold standard and method SAA versus gold standard illustrated that the dataset of method SA had a better general fit compared to the gold standard. The 95th percentile of the absolute distance measure of both datasets showed that method SAA had less outliers than method SA compared to the gold standard. This concludes that method SA has the best fit in comparison with the gold standard. However, it is more susceptible to large differences in specific areas than method SAA. Inspection of the distance maps showed that most outliers were situated near the inferior orbital fissure.
The results of method SAA are surprising, as one would expect that slight manual adjustments would improve method SA. The results of the distance maps and Fig. 6 for method SA show an accurate resemblance to the gold standard in most regions. The only region that is different is the region around the inferior orbital fissure, where a volume increment is seen in the SA model. The distance map of method SAA (Fig. 6) corresponds to that of method SA, except for the inferior orbital fissure, which now shows a volume decrement compared to the gold standard. An underestimation of the total volume was seen in method SAA for both observers, probably due to overcorrection of the orbital contour by the use of the built-in smart shaper tool and difficulty to find the border of the orbital volume