Patient-specific three-dimensional composite bone models for teaching and operation planning

Abstract

Background: Orthopedic trauma care relies on two-dimensional radiograms both before and during the operation. Understanding the three-dimensional nature of complex fractures on plain radiograms is challenging. Modern fluoroscopes can acquire three-dimensional volume datasets even during an operation, but the device limitations constrain the acquired volume to a cube of only 12-cm edge. However, viewing the surrounding intact structures is important to comprehend the fracture in its context. We suggest merging a fluoroscope's volume scan into a generic bone model to form a composite full-length 3D bone model.

Methods: Materials consisted of one cadaver bone and 20 three-dimensional surface models of human femora. Radiograms and computed tomography scans were taken before and after applying a controlled fracture to the bone. A 3D scan of the fracture was acquired using a mobile fluoroscope (Siemens Siremobil). The fracture was fitted into the generic bone models by rigid registration using a modified least-squares algorithm. Registration precision was determined and a clinical appraisal of the composite models obtained.

Results: Twenty composite bone models were generated. Average registration precision was 2.0 mm (range 1.6 to 2.6). Average processing time on a laptop computer was 35 s (range 20 to 55). Comparing synthesized radiograms with the actual radiograms of the fractured bone yielded clinically satisfactory results.

Conclusion: A three-dimensional full-length representation of a fractured bone can reliably be synthesized from a short scan of the patient's fracture and a generic bone model. This patient-specific model can subsequently be used for teaching, surgical operation planning, and intraoperative visualization purposes.

Fig 1

The volume acquired by a 3D capable fluoroscope is only approximately 12 × 12 × 12 cm (gray cube) and does not cover the entire bone. Assembling the fluoroscope’s 3D scan of the fracture with a generic bone model synthesizes a full-length composite bone model.

Fig 2

Concept of composite bone model synthesis. (a) The bone morphology and gross fracture localization are determined on a plain radiogram. (b) The most appropriate 3D model is selected from a set of generic bone models. (c) A 3D scan of the fracture is acquired on a mobile 3D-capable fluoroscope. Restricting the scan to just the fracture site limits radiation exposure. (d) The generic 3D bone and the actual fracture site are digitally assembled to form a composite 3D model.

Fig 3

Processing steps during generation of a composite 3D bone model. (a) The approximate location of the fracture is determined on the radiogram. (b) A generic intact bone model is cut accordingly. (c) A 3D model of the fracture site obtained from the 3D fluoroscopy scan and the collar of intact bone shaft isolated on each side of the fracture (images magnified). (d) The collars from the fracture scan are matched to the generic bone model by rigid registration. The composite 3D model is created by independently applying the determined transformation matrices to both ends of the generic bone model (b).

Fig 4

Results of matching the fracture site into different intact generic bone models. Each generic bone has its characteristic morphology. The matching algorithm is robust enough to account for variable surface mesh resolution, inhomogeneous shaft diameters and differing cross-sections.

Fig 5

Comparison of genuine and virtual radiograms. (a) Orthogonal radiograms of the fractured bone. (b) Virtual radiogram synthesized from a 3D model of a full-length CT scan of the fractured bone. (c) Virtual radiogram of the composite model composed of the fracture scan merged (arrows) into an intact generic femur model. The composite model shows high congruence with the reference images in a and b.

Fig 6

Seven superposed composite models. In this example, the manually measured position of the fracture within the shaft was intentionally varied in 5-mm steps to ±15 mm. The algorithm is robust enough to largely correct such an imprecision. This is demonstrated in the only minor deviations between the individually colored composite models.