High energy impacts at joint locations often generate highly fragmented, or comminuted bone fractures. A leading current approach for treatment requires physicians qualitatively to classify the fracture to one of four possible fracture severity cases. Each case then has a sequence of best-practices for obtaining the best possible prognosis for the patient. It has been observed that qualitative evaluation of fracture severity by physicians can vary significantlywhich can lead to potential misclassication and mis-treatment of these fracture cases. Major indicators of fracture severity are (i) fracture surface area, i.e., how much surface area was generated when the bone broke apart and (ii) dispersion, i.e., how far the fragments have rotated and translated from their original anatomic positions. Work in this dissertation develops computational tools that solve the bone puzzle-solving problem automatically or semi-automatically and extract previously unavailable quantitative information for these indicators from each bone fragment that are intended to assist physicians in making a more accurate and reliable fracture severity classification. The system applies novel three-dimensional (3D) puzzle-solving algorithms to identify the fracture fragments in the CT image data and piece them back together in a virtual environment. Doing so provides quantitative values for both fracture surface area and dispersion that reduce variability in fracture severity classifications and prevent mis-diagnosis for fracture cases that may be difficult to qualitatively classify using traditional approaches. This dissertation describes the system, the underlying algorithms and demonstrates the virtual reconstruction results and quantitative analysis of comminuted bone fractures from six clinical cases.