Surgical navigation and 3D printing in hemipelvic osteotomy

Abstract

Purpose: Surgery of pelvic tumours is a complex procedure due to resection of large masses, pelvic geometry and presence of critical anatomical structures [1]. Patient-specific surgical guides are designed from preoperative computed tomography (CT) images, 3D printed and placed on particular pelvic areas to define cutting planes [1]. However, each guide does not cover the whole cutting plane so no depth information beyond that tool is available. Surgical navigation overcomes this limitation and also provides intraoperative orientation by using patient’s preoperative images. A reference frame (RF) with optical passive markers placed on the iliac crest has been previously reported to account for pelvic movements during surgical navigation with an optical tracking system (OTS) [2]. Several anatomical landmarks were used to register real and virtual worlds, but optimal fiducials may not be visible in the initial steps of hemipelvic osteotomy or are difficult to capture accurately inside surgical cavity. A patient-specific RF placed on a particular area of iliac crest could avoid registration. We have tested this approach in two clinical cases with no complications during surgery related to frame attachment. However, target registration error (TRE) increases in those points far from the three principal axes defined by the optical markers of RF [3]. We propose to combine two tools designed specifically for each patient and 3D printed in hospital to reduce TRE in both the iliac and pubic cutting planes: a RF and a pubis-specific tool (PST) that includes registration landmarks. TRE was evaluated using pelvic phantoms obtained from those previous clinical cases, comparing values when using the RF only with those when adding PST. Methods: The first step of our approach is surgical planning (segmentation of pelvis and tumor, and definition of cutting planes [ilium and pubis]) using GNU open-source software Horos on preoperative CT and positron emission tomography (PET) images. After that, patient-specific RF and PST are modelled with freely available software MeshMixer by extruding characteristic surfaces obtained from segmentation of iliac crest and superior pubic ramus respectively. RF includes three screws to attach optical markers. PST has three conical holes (Ø4 x 3 mm depth) on its surface to be used as landmarks (registration step). These tools are 3D printed with a desktop fused deposition modelling device (Witbox-2, BQ) and polylactic acid. This thermoplastic was chosen due to its extrudability, nontoxic properties and possibility of being sterilized by ethylene oxide. During surgery, RF is attached to the iliac crest with a Kirschner wire and two metallic screws, and PST is placed on the superior pubic ramus, both tools at same positions as in the modelling step. Then, position of those conical holes in the real world is obtained with the tip of a tracked electrosurgical scalpel (Fig. 1). Navigation is performed with an OTS (OptiTrack, NaturalPoint) connected to open-source navigation software: Plus Toolkit, 3D Slicer platform and SlicerIGT extension. These fiducials and other three landmarks located on RF (ideal positions, without scalpel) are used in the registration step. With this virtual-to-real world transformation applied, surgeons can navigate, delimiting both cutting planes with the tracked scalpel. The evaluation of this approach was done with data from two clinical cases: pelvic angiosarcoma and epidermoid carcinoma of anal canal. 3D models of pelvises and tumours were 3D printed after segmenting CT and PET images (Fig. 1). Several conical holes (13, Ø4 x 3 mm depth) delimiting both cutting planes were included for the validation step. For each case, a CT image of the whole setting (pelvis, tumour, RF and PST) was acquired to obtain its 3D model since pelvises were 3D printed in several pieces. For each pelvis, TRE was estimated twice by using the validation holes in order to assess error improvement: firstly, after attaching RF and, secondly, after also attaching PST and carrying out registration. This process was repeated three times, removing both patient’s specific tools each time. Results: Table 1 shows TRE in both scenarios without and with PST. Theoretic values were also calculated by applying Fitzpatrick et al’s expected TRE [3] and assuming a fiducial localization error (FLE) of 1.6 mm (error of scalpel tip error measured at the centre of camera field of view, Fig. 1). Theoretic and experimental TREs decrease when adding PST. In most cases, experimental TREs are larger than theoretic values that do not account for anisotropic FLE (shown in OTS), misplacement of RF and PST and slight flexibility of scalpel tip. Conclusion: This study merges navigation, patient’s specific tools and 3D printing for hemipelvic osteotomy. The suggested approach removes the step of anatomical landmark selection, facilitating clinical workflow during surgery. The use of a pubis-specific tool to adjust virtual-to-real world registration improves TRE in both cutting planes.