Domain-Adaptive Self-Supervised Pretraining for Bacterial Flagellar Motor Detection in Cryo-Electron Tomography

To automate bacterial flagellar motor detection in cryo-electron tomography despite low signal-to-noise ratios, limited annotations, and complex structures.

Template-matching pipelines are unscalable. YOLO models face a domain gap from ImageNet pretraining. This project closes the domain gap for cryo-ET.

Quantifying bacterial flagellar motor copy number and spatial distribution, motor counting, spatial distribution mapping, and subtomogram averaging seeding.

—Automated detection of bacterial flagellar motors in cryo-electron tomography (cryo-ET) is challenged by extremely low signal-to-noise ratios, limited annotations, and complex macromolecular structures. We propose DA-SSL, a domain- adaptive self-supervised learning framework that pretrains a YOLOv 8 s backbone on unlabeled cryo-ET data to learn domain- specific noise characteristics and structural priors. Pretraining is guided by a masked reconstruction objective with a structural consistency regularizer, followed by systematic weight transfer andsupervisedfine-tuningwithaninitialbackbone-freezestrategy. The proposed approach achieves m AP@0.5 = 84.47% and precision = 85.05%, outperforming supervised baselines while maintaining real-time inference capability. Ablation studies show that both pretraining and freezing contribute non-redundant performance gains, and evaluation on unseen data confirms improved generalization.