A Non-Contact Hand-Actuated Facial Scanning System to Aid Analysis and Planning for Reconstructive Surgery

Padraig Butler and David Vernon, Department of Computer Science, Maynooth College, Co. Kildare, Ireland.

The information required in reconstructive surgery contrasts with the information required in traditional medicine in that it is primarily concerned with the body's structure rather than its physiology. The acquisition of this structural information poses a number of problems. For example, the faces of children, who represent a ]arge proportion of the potential patients, alter with age so data must be acquired on an on-going basis.

Unfortunately, current systems are very large, incorporating bulky motorised fixtures and optical jigs. Often, they require that the patient's head to be fixed during the scanning process. In addition, such intrusive systems are intimidating to children. Ultimately, plastic surgeons require a portable system which can be used to gather the required information in remote locations, outside the laboratory, either in the ward or at the patient's home or school.

The Department of Computer Science at Maynooth, in co-operation with the Children's Hospital, Temple Street, Dublin, has developed a highly portable 3-D computer vision system which can be used without the aid of bulky jigs and motors. The system works by projecting a stripe of light on to the patient's face. This stripe deforms as it is reflected from the face and the distorted pattern is picked up by a video camera. The resultant image is then passed to a computer system so that the shape of the part of the face illuminated by the stripe can be reconstructed. The shape of the entire face can be recovered by scanning over the face with this light-stripe projecton In our system, the face is scanned by rotating the light-stripe projector and the camera about the head of the patient. This rotation is done simply by moving the device by hand. However, in order to reconstruct the entire surface of the face from the individual profiles, we need to know the angles at which each profile is acquired. This is done automatically by analysing registration patterns embedded in each image. Current work is concerned with maximising the fidelity and accuracy of the system and with effecting automatic face registration to enable quantitative assessment of facial features to assist the surgeon in planning for and in evaluating surgical procedures.