Facial recognition techniques applied to the automated registration of patients in the emergency treatment of head injuries

This paper describes the development of a registration framework for image-guided solutions to the automation of certain routine neurosurgical procedures. The registration process aligns the pose of the patient in the preoperative space to that of the intra-operative space. CT images are used in the pre-operative (planning) stage, whilst white light (TV camera) images are used to capture the intra-operative pose. Craniofacial landmarks, rather than artificial markers, are used as the registration basis for the alignment. To further synergy between the user and the image-guided system, automated methods for extraction of these landmarks have been developed. The results obtained from the application of a Polynomial Neural Network (PNN) classifier based on Gabor features for the detection and localisation of the selected craniofacial landmarks, namely the ear tragus and eye corners in the white light modality are presented. The robustness of the classifier to variations in intensity and noise is analysed. The results show that such a classifier gives good performance for the extraction of craniofacial landmarks.