Simulation / Modeling / Design

New AI-Powered 3D Printing Can Help Surgeons Rehearse Procedures

Researchers at Washington State University (WSU) unveiled a new AI-guided 3D printing technique that can help physicians print intricate replicas of human organs. Surgeons can then use these organ models to practice before performing the actual surgery, which gives doctors more tools to improve surgical results. 

The AI algorithm was trained on images and key attributes of human kidneys and prostates, including characteristics like weight, size, porosity, and vascular architecture. The algorithm works with 3D printers in a process of improvement. It helps find the best settings for three important parts of 3D printing: how accurate the model is, how light it is, and how long it takes to print. 

One of the co-authors of the study, Kaiyan Qiu, an assistant professor of mechanical and materials engineering at WSU, said that AI optimizations can materially shorten the time it takes to create viable 3D models. The algorithm adjusts  key 3D printing variables, including a printer’s nozzle size and travel speed, the pressure printing materials are dispensed at, and the height of each printed layer.. It then guides the printer in creating an appropriate model for a specific use-case.

“For pre-surgical organ models, we know surgeons will need high fidelity models that can be printed out quickly and with low labor intensity,” Prof. Qiu said. “We imagine a scenario where a surgeon receives an MRI and CT scan [of a patient] in the morning. She has two hours to prepare everything for surgery. The AI can optimize the parameters, and print out a model organ in half-an-hour, and the surgeon can then spend the remaining time practicing [on the organ replica].” 

Qiu and his co-author on the paper, WSU computer science professor, Jana Doppa, used a multi-objective Bayesian Optimization (BO) approach using BoTorch to improve the efficiency and precision of the 3D printing process. The BO algorithm uses a probabilistic surrogate model to approximate the relationship between printing parameters and the quality of the printed organ models. This process captures uncertainties in the printing process, allowing for more robust optimization.

Flow-chart schematic of multi-objectiveBO assisted 3D-printing of presurgical organs models with three input parameters in tangent with four output parameters. The cycle starts with generating input values based on the current dataset of inputs and corresponding outputs through BO, which are used to produce printing pathways for direct-ink-writing (DIW). After the model is 3D-printed via DIW, image processing is applied to the model to reconstruct a mesh object. The mesh object is then adjusted for comparisons with the ideal model for measurements regarding positive and negative geometrical precisions. The time of model printing and porosity measurements are also calculated. Once all the output measurements are completed, their individual values are re-entered into the BO algorithm to yield new input parameters.
Figure 1. The methodology for machine learning assisted 3D-printing is a four-step recursive process.

The researchers used NVIDIA A40 GPUs to train their AI model, and NVIDIA NGP Instant NeRF to reconstruct a mesh object of the 3D printed model.

The AI process the researchers have pioneered is also broadly generalizable. Besides printing model organs, the algorithm can guide printers to create prototypes of implantable medical devices, like pacemakers or stents. The technology can also be used to make models of airplane and robot parts, batteries, or even shoes customized for you. 

To learn more about this research, you can read this article, or this research paper.

Featured image credit: Washington State University

Discuss (0)

Tags