Precision and accuracy of tridimensional localization in Statscan digital medical radiology

The emergence of computerized medical imaging in early 1970s, which merged with digital technology in the 1980s, was celebrated as a major breakthrough in three-dimensional (3D) medicine. However, a recent South African innovation, the high speed scanning Lodox Stat scanCritical Digital Radiology modality, has posed challenges in X-ray photogrammetry. This is due to the system's imaging geometry. This research investigates the applicability of Direct Linear Transformation (DLT) method in Lodox Statscan 3D point localization. Static modelsdesigned from metal frames bearing targets of different contrast have been imaged on the Statscan system to generate experimental data. These models were used to eliminate distortions that arise from involuntary human body movements. A control frame for the 3D models has been generated at an accuracy of ± 0.5mm. Point positioning accuracy has been computed by comparing the acquired Statscan 3D point positions to the established control. Two different reference frames were used, showing that point positions could be established with RMS values in the mm range in the middle axis of the Statscan X-ray patient platform. This range of acceptable mm accuracies extends about 15 to 20 cm sideways towards the edge of the table and to about 20 cm above the table surface. Beyond this range, accuracy deteriorated significantly. The experiments further showed that the inclusion of control points close to the table edges and more than 20cm above the table resulted in lower accuracies for the L - parameters ofthe DLT solution than those derived from points close to the center axis only. As the accuracy of the L - parameters propagates into accuracy of the final coordinates of newly determined points, it becomes essential to restrict the space of the control points to the above described limits.

Grantor dc:publisher.institution

Division of Geomatics

Year dc:date.issued

2006

Author dc:creator

• Kimuyu, Jacinta Syokau

Advisor dc:contributor.advisor

• Rüther, Heinz

Language dc:language.iso

eng

Handle dc:identifier.uri

http://hdl.handle.net/11427/10923

OAI identifier oai:identifier

oai:open.uct.ac.za:11427/10923