Manual and Computer-Assisted Pedicle Screw Placement Plans: A Quantitative Comparison

Bone density optimized pedicle screw instrumentation improves screw pull-out force in lumbar vertebrae

Pedicle screw instrumentation is performed in the surgical treatment of a wide variety of spinal pathologies. A common postoperative complication associated with this procedure is screw loosening. It has been shown that patient-specific screw fixation can be automated to match standard clinical practice and that failure can be estimated preoperatively using computed tomography images. Hence, we set out to optimize three-dimensional preoperative planning to achieve more mechanically robust screw purchase allowing deviation from intuitive, standard screw parameters. Toward this purpose, we employed a genetic algorithm optimization to find optimal screw sizes and trajectories by maximizing the CT derived bone mechanical properties. The method was tested on cadaveric lumbar vertebrae (L1 to L5) of four human spines (2 female/2 male; age range 60-78 years). The main boundary conditions were the predefined, level-dependent areas of possible screw entry points, as well as the automatically located pedicle structures. Finite element analysis was used to compare the genetic algorithm output to standard clinical planning of screw positioning in terms of the simulated pull-out strength. The genetic algorithm optimization successfully found screw sizes and trajectories that maximize the sum of the Young's modulus within the screw's volume for all 40 pedicle screws included in this study. Overall, there was a 26% increase in simulated pull-out strength for optimized compared to traditional screw trajectories and sizes. Our results indicate that optimizing pedicle screw instrumentation in lumbar vertebrae based on bone quality measures improves screw purchase as compared to traditional instrumentation.

Joint Vertebrae Identification and Localization in Spinal CT Images by Combining Short- and Long-Range Contextual Information

Automatic vertebrae identification and localization from arbitrary computed tomography (CT) images is challenging. Vertebrae usually share similar morphological appearance. Because of pathology and the arbitrary field-of-view of CT scans, one can hardly rely on the existence of some anchor vertebrae or parametric methods to model the appearance and shape. To solve the problem, we argue that: 1) one should make use of the short-range contextual information, such as the presence of some nearby organs (if any), to roughly estimate the target vertebrae; and 2) due to the unique anatomic structure of the spine column, vertebrae have fixed sequential order, which provides the important long-range contextual information to further calibrate the results. We propose a robust and efficient vertebrae identification and localization system that can inherently learn to incorporate both the short- and long-range contextual information in a supervised manner. To this end, we develop a multi-task 3-D fully convolutional neural network to effectively extract the short-range contextual information around the target vertebrae. For the long-range contextual information, we propose a multi-task bidirectional recurrent neural network to encode the spatial and contextual information among the vertebrae of the visible spine column. We demonstrate the effectiveness of the proposed approach on a challenging data set, and the experimental results show that our approach outperforms the state-of-the-art methods by a significant margin.