Requirements exist clinically for quantitative analysis of facial swel
ling. A quick and non-invasive measurement of facial volume is desirab
le. The acquisition of three-dimensional images, corresponding to spat
ial models of patients' facial surfaces, has been facilitated using a
Spatial Vision System (SVS) developed by Thorn EMI Central Research La
boratories (UK). The SVS acquires four video images (two stereo pairs)
and constructs a three-dimensional surface representation using stere
o matching of spatial features acquired using 'speckled' illumination.
Pre-and post-operative binary volumetric images are then spatially al
igned using surface matching algorithms. Accuracy of registration is o
ptimized using Match Volume Excavation. Segmentation of post-operative
swelling is achieved by volume differencing in combination with morph
ological erosion. Volumetric quantification is achieved using a method
of three-dimensional region growing, giving the number of connected v
oxels in the image volume from a user selected seed point. Accuracy of
image acquisition has been quantified by imaging a range of hemispher
ical phantoms of known volumes. The mean error in repeated measurement
of volume was found to be within 3.5%. The perception of temporal cha
nges in facial swelling is highly subjective. Further work is being un
dertaken to develop an Enhanced Reality visualization tool in which gr
aphical swelling annotations are combined with real-time patient image
s.