| Current medical treatment for comminuted bone fractures, i.e., traumatic bone fractures that result in many bone fragments, is based upon fracture severity classifications that physicians determine subjectively. Due to the subjectivity in the interpretation of the available information, the severity of a single fracture case may be classified differently by physicians. Accurate, reliable, and repeatable classification of fracture severity is an important factor in planning effective treatment and an overall positive prognosis for difficult fracture cases. Recent work has placed an emphasis on developing computational tools to analyze CT image data and estimate fracture severity. This research explores the statistical relationships between fracture severity and quantities derived from a new virtual bone fracture reconstruction system. Many of these quantities have not been previously available due to the lack of a system capable of virtually reconstructing highly-fragmented bone fractures. The existence of a new bone reconstruction system makes available a new set of measurable values whose relationships to fracture severity have been discussed but never been quantitatively examined. The relation between fracture severity and these quantities is heretofore unknown and this thesis provides an initial analysis of utility of these features for fracture severity prediction via automated CT image analysis. This thesis discusses the predictive capabilities of features for fracture severity for seven clinical cases, ranked by three orthopaedic surgeons. |
