Search papers, labs, and topics across Lattice.
The paper introduces a geometry-driven data synthesis pipeline to generate realistic 3D myotube microscopy volumes for training instance segmentation models. This pipeline models myotube morphology using polynomial centerlines, varying radii, branching, and ellipsoidal end caps, and incorporates realistic noise and domain adaptation via CycleGANs. A 3D U-Net trained solely on this synthetic data achieves a mean IPQ of 0.22 on real data, surpassing existing zero-shot segmentation models.
High-quality 3D biomedical instance segmentation is now possible in annotation-scarce domains, thanks to a biophysics-driven data synthesis pipeline that bridges the gap between synthetic and real microscopy data.
Myotubes are multinucleated muscle fibers serving as key model systems for studying muscle physiology, disease mechanisms, and drug responses. Mechanistic studies and drug screening thereby rely on quantitative morphological readouts such as diameter, length, and branching degree, which in turn require precise three-dimensional instance segmentation. Yet established pretrained biomedical segmentation models fail to generalize to this domain due to the absence of large annotated myotube datasets. We introduce a geometry-driven synthesis pipeline that models individual myotubes via polynomial centerlines, locally varying radii, branching structures, and ellipsoidal end caps derived from real microscopy observations. Synthetic volumes are rendered with realistic noise, optical artifacts, and CycleGAN-based Domain Adaptation (DA). A compact 3D U-Net with self-supervised encoder pretraining, trained exclusively on synthetic data, achieves a mean IPQ of 0.22 on real data, significantly outperforming three established zero-shot segmentation models, demonstrating that biophysics-driven synthesis enables effective instance segmentation in annotation-scarce biomedical domains.