ETD

• 3D reconstruction
• Bioprinting
• FLight
• Osteochondral defects

Lesions of osteochondral tissues of diarthrodial joints, such as the knee, pose a significant clinical challenge due to the limited regeneration capabilities of articular cartilage, often causing pain and limiting joint functionality, as well as increasing the risk of post-traumatic osteoarthritis. Current cell-free and cell-based surgical approaches implemented to treat osteochondral lesions are often expensive, typically require multiple interventions and are limited in availability, leading to arthroplasty for restoration of joint functionality and pain relief. Within this context, in-situ bioprinting methodologies arise as a promising alternative to traditional approaches, enabling controlled deposition and arrangement of biomaterials and cells directly inside the lesion site during the surgical operation. Particularly, solutions combining multiple bioprinting techniques, such as extrusion-based bioprinting (EBB) and novel light-based approaches, such as Filamented-Light (FLight) bioprinting, are especially relevant, as they allow to control both biomaterial depositions within the defect site, as well as its macro and micro-architecture through pattern-based photo-crosslinking, enabling the fabrication of constructs mimicking the natural structure of the osteochondral tissue and promoting tissue growth and maturation.
This work presents the development and preliminary validation of ex-vivo and in-silico models for an in-situ bioprinting approach combining EBB with FLight, for one-stage regeneration of osteochondral defects. The proposed models aim to: i) generate 3D digital models of osteochondral lesions to estimate the biomaterial volume to be extruded for complete filling of the defect, using arthroscopic videos of the lesion as an input, and ii) simulate the FLight bioprinting process within the extruded material, allowing for quantitative prediction of the final architecture of bioprinted constructs.
To generate digital models of osteochondral defects, a 3D reconstruction pipeline was developed based on state-of-the-art techniques combining Structure from Motion (SfM), Multi-view Stereo (MVS) and Poisson Surface Reconstruction algorithm. Despite their effectiveness in many computer vision tasks, such methods find limited application in arthroscopy, as the low texture of articular cartilage and the limited contrast inside the knee joint often lead to instabilities and failures of the reconstruction process. In order to improve the applicability of these approaches to the arthroscopic setting, a pipeline was developed and optimized using 3D printed simulated osteochondral defects as reference objects. After tuning the pipeline parameters to ensure robust and repeatable reconstructions, validation experiments were performed on real osteochondral defects induced in ex-vivo bovine knee joints, in which the reconstructed sizes of osteochondral lesions were compared to their true dimension.
In order simulate the FLight bioprinting process within materials extruded inside osteochondral lesions, a computational model, implemented in Matlab® was developed.
The model takes parameters regarding the light source, the photo-crosslinkable resin and the printing pattern as inputs, and provides the 3D distributions of the refractive index and deposited light-dose within the material at the end of the printing process as an output. Analysis of local maxima and minima of the refractive index distribution allows for quantification of the micro-architecture of the printed constructs, while volumetric segmentation of the deposited light dose within the material enables porosity assessments. The Matlab implementation of the FLight model was developed starting from the open-source optical propagation solver BPM-Matlab, building a coupled framework that combines light propagation physics with models describing the temporal evolution of the optical properties of the biomaterial undergoing cross-linking.
To validate the FLight model, gelatin methacryloyl (GelMA) was chosen as reference material due to its known photo-crosslinking ability, using lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP) as a photo-initiatior. Model validation experiments were performed on different GelMA compositions with varying material and photo-initiator concentrations (GelMA 5% with 0.05% LAP, GelMA 5% with 0.10% LAP and GelMA 10% with 0.10% LAP, concentrations were expressed as percentage of weight over volume). Photo-crosslinking properties of the GelMA resins were assessed through photo-rheological tests at 37°C and 25°C, and through dose tests using a FLight bioprinting device. Optical properties of the materials were evaluated through refractometer measurements. Validation of the FLight model was performed by quantitatively comparing the micro-architecture of printed constructs obtained experimentally and through simulations under the same printing settings, for each material composition. Automated micro-architecture quantification was carried out using a Matlab script taking microscopy images for experimental data, and the output of the FLight model for simulated constructs, as inputs.
Finally, a proof of concept combining the two models was proposed: a real osteochondral defect induced in a bovine knee joint was reconstructed using the established pipeline, and the measured dimensions of the digital model were then used to estimate the volume of resin needed to fill the lesion. After depositing the material inside the defect, FLight printing was performed in-situ, and the cross-linked resin was subsequently removed for micro-structure characterization. Lastly, the micro-architecture of the printed construct was compared to the one obtained through the FLight computational model, under the same printing conditions, at different depths.
The proposed models showed promising results, as the reconstruction pipeline provided diameter estimations of the defects close to the real values (Root Mean Squared Error of 0.48 mm with defect sizes of 7-8 mm), while showing a consistent underestimation of defect depth (approximately by a factor of 2), which can be compensated through the introduction of an appropriate scaling factor. The FLight computational model was able to produce outputs matching the real architecture of the printed constructs for each tested material (Root Mean Squared Error of mean micro-filament size of 0.34 µm, with distribution mean values ranging from 4.9 to 6.0 µm).
Future developments will focus on the fine-tuning of the reconstruction pipeline for accurate defect depth estimation (e.g., projection of structured light pattern in the osteochondral defects during arthroscopic acquisitions), and optimization of the FLight model for improved assessment of size distribution of micro-channels in printed constructs (e.g., introduction of additional material parameters, such as volumetric shrinkage coefficients). Further work will involve the validation of the FLight model on new materials, such as hyaluronic acid methacrylate (HAMA) and collagen-based photo-resins.