A generative AI framework for disease-specific lung microtissue bioengineering

Abstract

Generative Lung Architecture Modeling (GLAM) is an integrated bioengineering framework that couples high-resolution three-dimensional tissue imaging with generative artificial intelligence to de novo design and 3D-bioprint anatomically detailed lung microtissue models. Native extracellular 3D matrix architectures of pulmonary parenchyma were extracted from healthy, fibrotic, and emphysematous in vivo mouse disease models and processed through a computational pipeline containing pre-trained image segmentation and 3D mesh generation. The resulting datasets were used to train a U-Net generative diffusion model with attention layers capable of synthesizing healthy and diseased lung tissue architectures. Microtissue cubes of about 200 - 300 {micro}m edge length of native and synthetic datasets were fabricated through high-resolution two-photon stereolithography with gelatin-methacryloyl biomaterial ink and successfully seeded with cells, demonstrating biological compatibility. In closing the loop between biological imaging, generative modeling, and high-resolution biofabrication, this integrated framework establishes generative AI as a functional design layer for tissue engineering. The resulting lung microtissues retained architectural features of the native and original tissues, making them an application-ready platform for customizable and scalable fabrication of biological tissue surrogates for preclinical modeling, drug testing, and precision regenerative bioengineering.