An investigation into the use of stereophotogrammetry for the analysis of craniofacial dysmorphology in schizophrenia

• Main Author: Cousins, Michael
• Other Authors: Douglas, Tania S
• Format: Dissertation
• Language: English
• Published: University of Cape Town 2015
• Subjects: Biomedical Engineering
• Online Access: http://hdl.handle.net/11427/15467

Studies of craniofacial dysmorphology in schizophrenia, carried out since the 1960s, have reported minor physical anomalies in those with schizophrenia, prominently in the craniofacial region. Indirect methods, most notably 3D laser imaging, have been used previously for investigating craniofacial dysmorphology in schizophrenia. This project aimed to investigate the ability of a stereophotogrammetry system to detect craniofacial dysmorphology in individuals diagnosed with schizophrenia. Furthermore, observed dysmorphology was characterised and compared with that found in previous studies. Three-dimensional craniofacial landmark coordinates were obtained from images collected using a bespoke design stereophotogrammetry system. The system includes a camera rig and a calibration rig. On the camera rig is mounted three digital single-lens reflex cameras hardwired to a trigger for simultaneous image capture. The calibration rig consists of a frame with strategically positioned retro-reflective calibration markers of known 3D orientation. The precision and reliability of the stereophotogrammetry system was tested using a human subject. Measurements were taken using the system and directly using callipers by two operators on two separate occasions. Intra- and inter-operator precision and inter-modality reliability were calculated and scored. All intra- and inter-operator precision scores were at least below a 7% error, and considered "good". Inter -modality reliability scores had at least a "good" score in 72% of all measurements. Excluding one soft landmark and one landmark with small measurement value, all inter-modality reliability scores were at least "good". The study cohort consisted of 17 African (8 control, 9 schizophrenia) and 13 Caucasian ( 8 control, 5 schizophrenia) males. A set of 18 landmarks focused about the eyes, nose, mouth and chin was identified for each subject and collated in 3D coordinate space. Geometric morphometric analysis - particularly generalised Procrustes analysis and principal component analysis - was carried out on these landmark sets. Discriminant Function Analysis was applied to identify discriminating features in the data set, and classification techniques, aided by feature selection, were applied to separate affected and control subjects. In the African cohort, the results showed wider inward slanting (cat-like) eyes, a wider upturned nose and narrower downturned mouth. In the Caucasian cohort, narrower and wide set eyes, a narrower downturned nose with anteriorly displaced alare, a wider downturned mouth and posteriorly set chin were shown. The Caucasian cohort demonstrates similar dysmorphology as described in the literature. Published data for the African cohort is lacking. The nearest mean and k- nearest neighbour classifiers had the highest accuracy in the African and Caucasian groups respectively, with 71% and 77% correct classification. The efficacy of the stereophotogrammetry system introduced in this study has been shown, with craniofacial dysmorphology in schizophrenia successfully detected. Further studies with larger cohorts are recommended to attempt improved classification accuracy, but a platform now exists to pursue dysmorphology studies in other psychoses, such as bipolar disorder.