The information of brain disease from neuroimaging techniques was traditionally displayed on films or screens, facilitating the diagnosis. However, this would be changed by the emerging technique called three dimensional (3D) printing. This technique deposit materials, such as plastic or metal, layer by layer to construct solid 3D models. It makes the disease information from a patient available “by touch”. Surgical treatment of complex cerebrovascular lesions, such as arteriovenous malformations (AVMs) and aneurysm, require careful planning before surgery. The 3D-printed models have the potential to help surgeons to rehearse the cases beforehand and reduce operative risk. Time-of-flight (TOF) MR angiography is a widely used sequence of evaluating cerebral arterial diseases. TOF MRA could provide high spatial resolution and good blood/tissue contrast without the need for the injection of contrast agent on modern clinical 3T scanner. Thus, image data from TOF MRA is a good source for constructing 3D printable cerebrovascular models. In some case, the surrounding skull anatomy is also needed to determine the operative approach. These information can be acquired from a routine CT scan. In this work, we describe a semi-automatic workflow that combine TOF-MRA data and CT data to construct a printable 3D model for cerebrovascular surgical planning.

A 57 years old male patient suspected to aneurysm was involved in this work. MRA data was acquired on a clinical 3T scanner with an 8-channel phase array head coil using 3D TOF sequence (TR=20ms, TE=3.59, Flip angle=18) with 240 x 240 matrix over a field of view of 240 x 240 mm and 80 axial slices of 1mm thickness and -20% slice distance. CT scan was performed on 128 detectors CT scanner (80kV, 358-410mA, 1mm slice thickness). High resolution T1 weighted anatomical images were also acquired for registration purpose using MP-RAGE sequence (TR=8.5ms, TE=3.5ms, TI=400ms, Flip angle=12, 1mm³ isotropic voxel size).

Both MRA and CT data were co-registered to T1w image to unify the resolution and imaging space using affine registration tool of Advanced Normalization Tools (ANTs)¹. The signal inhomogeneity of MRA data was corrected using N4BiasFieldCorrection tool from ANTs package before registration². The segmentation of vessel and skull were performed using 3D level set algorithm implemented in ITK-SNAP software³. Mathematic morphological processing was then used on the raw segmentation to fill the cavity inside the skull and remove unconnected parts (Figure 1). The segmented binary images of vessel and skull were converted to mesh data and export to STL format.

Model for 3D printing need to be watertight and manifold, which means the model cannot have holes on its surface and every triangle edge is shared by two and only two triangles. However, current model may have separated vertices and the topology of edges is mussy. This would cause unpredictable error during printing. To fix this problem, we use the ZRemesher function from ZBrush software (Pixologic inc.) to generate high quality retopology as well as remove unnecessary polygons to meet the requirement (Figure 2). Now the model is ready to be printed.

In this work, we proposed a semi-automatic workflow of constructing printable vessels and skull models of an individual patient based on MRA and CT images. The full construction can be done within one hour with minimal manual intervention. The only two steps need manual operation is to place initial surfaces at main branch of vessels for level set evolution and to draw guide lines for mesh refinement. The constructed models were consistent with the findings of MRA in our case, a 4mm aneurysm at M1 section of right middle cerebral artery (Figure 3). Surgeons can physically hold the 3D models, view them from different angles, practice the operation with real instruments and get tactile feedback, especially for deep vessels that are very tricky to operate on. The 3D printed models may be even more meaningful for surgical planning of interventional therapy or medical education.

In conclusion, our proposed workflow may improve the feasibility of using 3D printing technique in the treatment of intracranial artery diseases. Further investigations are warranted to optimize this technique and translate it into research and clinical practice.