Computer Science Tools for Manual Editing of Computed Tomographic Images: Impact on the Quality of 3D Printed Models

Abstract

Background: Three-dimensional printing (3DP) technologies are a trendsetting topic, also in the field of surgery. Preoperative planning for maxillofacial and neurological surgery, for instance, increasingly involves skull models obtained by 3DP. However, the cranial replicas currently used in this context have been shown to not reproduce the exact anatomy of the individual patient undergoing surgery. Objective: The present study aimed at investigating the extent to which manual editing, using current computer software tools, can improve skull models derived from medical images. Methods: Skull computed tomography (CT) was obtained on three cadavers and sent to three institutions that provide preoperative 3DP services. Each institute independently performed 3D reconstructions, including routine manual editing, and subsequently produced the replicas. The models were then qualitatively compared with the respective original skull. For quantitative comparison surface scans of particular regions of interest were made and the deviations assessed using 3-matic software (Materialise, Leuven, Belgium). Results: Routine manual editing of CT images resulted in replicas that were clear improvements over automatically generated reconstructions. This was particularly the case for teeth artefacts and thin-walled entities (e.g. paranasal sinuses). Conversely, however, many anatomical structures remained incorrectly rendered (e.g. orbitae, pterygoid processes, and sella turcica). Extraosseous calcifications had regularly not been removed. After extensive manual editing, however, replicas were able to provide largely submillimeter accuracy (mean deviation 0.2496 mm; standard deviation ±0.2276 mm). Conclusions: This study confirms that manual editing with current computer science tools does improve the quality of CT-based 3D printed skull models. But, it also demonstrates that a number of structures remain largely incorrectly rendered when edited in the presently used surgical framework. Conversely, it shows that highly accurate replicas are feasible, provided that extensive manual editing is performed.