Enhanced resolution and speckle‐free three‐dimensional printing of macular optical coherence tomography angiography

Abstract

Volume rendering of optical coherence tomography angiography (OCTA) is a rapidly evolving imaging tool, which has been shown to preserve the three-dimensional (3D) architecture of various retinal diseases including diabetic retinopathy, retinal vein occlusion and macular telangiectasia type 2 (Spaide 2015). This form of imaging avoids flattening of subvolumes of tissue as is done in en face imaging and does not require the use of segmentation, which often is incorrect in retinal disease. Volume rendering can illustrate the close relationship between the flow signal and structural optical coherence tomography (OCT) information from which it is derived. By extending the OCT volume rendering technology into stereolithography, the whole 3D experience can be made physically tangible by producing life-like models as well as surgical templates and implants on a larger scale for an additional conceptualization and tactile feedback. The introduction of stereolithography in medicine has already been demonstrated to be a useful tool for surgical planning in congenital heart surgery, reconstructive surgery and in simulation training for aneurysmal surgery. Such 3D models might also be important in ophthalmology for surgical training or planning of microsurgical procedures and for teaching purposes for students and patients (Choi et al. 2016). In this report, we demonstrate the first stereolithographic retinal vessel models based on standard OCTA. Two methods were developed to obtain a printable 3D model (Fig. 1). Nine repeats of tracked OCTA measurements (3 mm × 3 mm scan area, 245 × 245 pixel), were performed on one healthy right macula of a 35-year-old emmetropic female with Zeiss Cirrus HD-OCT Model 5000 with angioplex (Review software 9.0.0.281; Carl Zeiss Meditec, Jena, Germany). The nine OCTA volumes were aligned and averaged into one final “enhanced resolution” OCTA volume (“erOCTA”), and a print model was saved in obj format. Another method for the 3D printing used single volume rendering and was tested first and freed from speckle noise using a recently developed 3D speckle denoiser. The 3D data stack was imported in open source imagej (v1.467; ref – Rasband, W.S., imagej, US National Institutes of Health, Bethesda, MD, USA, https://imagej.nih.gov/ij/, 1997–2016) and after thresholding, exported in obj format. This 3D mesh was modified with a view to seal vessel gaps and remove digital artefacts with the digital sculpting tool zBrush 4R7 (Pixologic inc., Los Angeles, CA, USA). A printable prototype was transferred to the 3D printing service company i.materialise (Materialise HQ, Leuven, Belgium) and printed in transparent resin constructed from a hardened liquid. The material is strong, hard, stiff and water-resistant and is suited for models that require a smooth, good-quality surface with a transparent look. Design specifications for 3D printing included minimum wall thickness of 1 mm, minimum details of 0.5 mm, accuracy ±0.2% and a size of 200 × 200 × 15 mm, although 3D prints 300 × 300 × 28 mm have been made (Fig. 2). This corresponds to a magnification of 66.7–100 times. Finally, to accentuate the details of the retinal vessels, one stereolithographic replica was charged by conduction by submerging it in a silver bath and copper bath for 48 hr and subsequently in a gold bath, 24 karat. The 3D print depicts the typical arrangement of the superficial vascular complex vessels that lie in a linear pattern along the inner retinal surface with vertical branches into deeper retinal vascular networks. Video S1 demonstrates a 3D print of a normal 3DOCTA, especially the foveolar avascular zone (FAZ). These vessels do not necessarily look round but are thickened in the Z-axis because of decorrelation tails. The 3D rendering showed partially irregular thickening of the vessels. In addition, multiple, small, wart-like protrusions were found on the vessel surface, which appeared to be localized above and perpendicular to the vessel direction and in direction of the laser beam, respectively. These presumed signal artefact protrusions show themselves most impressively and most frequently on the large superficial vessels, and diminish in the direction to the FAZ, but are also present in smaller numbers and on the innermost FAZ vessels. They change in size and shape over time on repeat imaging. Most were found in contiguous with the vessel, others appear marginally separated from the vessel itself. The protrusions show a relatively uniform distance from one another, but can aggregate sometimes. In addition, small defects in visualizing flow were present as little dents (Fig. 2D). Both new 3D OCTA volume rendering findings are depicted as signal protrusions (“OCTA warts”) and signal constrictions (“OCTA dents”), respectively, in video S2It is possible that these localized changes on the vessels are a result of algorithms that generate the image and do not reflect true anatomical structure. The 3D model clearly demonstrated that the FAZ is outlined by only few vessels of changing depth within the retina, which is not detectable in 2D en face OCT. This fact may be of particular interest, as it is worth discussing whether it may be accurate to apply in a conventional 2D OCTA a planar 2D measurement of the OCTA surface, which in itself represents but a 3D structure. In the posterior en face view, efferent vessels united in about fourteen collecting vessel networks that were detected in the 3D as has been described by others (Bonnin et al. 2015). We introduce 3D-printed OCTA as a novel technique for visualization of retinal vessels and the connection between the superficial and deeper vascular networks. Acquisition of size-calibrated 3D images of the retinal vascular network enables the characterization of the spatial relationship within the retina with preservation of valuable volumetric relationships, thus enabling a longitudinal assessment of the retinal vasculature that might have been missed from the cross-sectional 2D data set. This feature may help improve diagnostic assessment of retinal vascular diseases and may serve as a platform for therapeutic interventional strategies and drug development. Moreover, 3D imaging has been shown to be important in molecular imaging techniques in human subjects. This new field of medicine offers the possibility of visualizing in vivo molecular processes after the administration of extrinsic contrast agents. Several approaches are currently being explored so to enable OCT-based molecular imaging in vivo and multimodal imaging with incorporation of high-resolution 3D imaging as a promising strategy potentially capable of probing both morphological and biochemical pathological changes in the retina and its vasculature (Ramos de Carvalho et al. 2014). Another potential application of this technology includes aiding conceptualization of retinal disorders and for educational purposes. We used such 3D printouts at the University of Basel and Moorfields Eye Hospital, London, for teaching of retinal anatomy to students and even blind patients who were able to explore and understand the retinal architecture using tactile stimulation. Two inherent difficulties of our study include some data loss during data processing with speckle-denoising and thresholding and neglect of smaller vessels loops due to lack of attachment to a larger vessel. Furthermore, OCTA data are prone to artefacts that were automatically removed by the manufacturers' software (Spaide et al. 2015). Covering the model with a thin gold layer reduced the size of the intervessel spaces, but increased the robustness and is aesthetically more pleasing. Modern 3D printer is capable of printing complex structures, but the relative high printing costs may represent a limit and be outside of the budget of most labs. Printing costs were for transparent resin red with model size of 129.29 × 71.92 × 200 mm, mammoth resin natural white with model size of 129.29 × 71.92 × 200 mm, transparent resin red with model size of 200 × 200.07 × 14.83 mm and transparent resin red with model size of 300 × 300.42 × 28.42 mm, 72, 207, 240 and 666 Euro, respectively. It is likely the cost of 3D printing will decrease substantially as the technology becomes more widely available across other industries. This study reports the first 3D printing of OCTA data sets. It provides a novel structural representation of the OCTA data and gives insight into the 3D architecture of the retinal vessels and the interactions of the different vascular networks in the retinal microcirculation. Furthermore, 3D printouts are a useful visual and tactile tool, which has the potential to aid in patient and surgeon education. One foreseeable direction of this technology may be to move from printing with resins to the reconstruction of the retinal microcirculation with viable tissue so to visualize how retinal vessels are perfused and interact. In addition, this technology may be adapted to other imaging techniques with the aim to provide spatiotemporal characterization of 2D cross-sectional imaging of molecular and biochemical processes. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.