Scott T. Retterer

Current segmentation models are capable of routine identification of biological features in noisy cryogenic electron microscopy (cryoEM) images. However, there are still challenges with complete segmentation of high boundary, thin objects such as bacterial cell envelopes and flagella. Moreover, ultralow-dose cryoEM images pose as an additional challenge to boundary distinctions between the object and background. Here, we present TileBac, a benchmark dataset of ultralow-dose montage tiles of Pantoea sp. YR343 to segment bacterial inner and outer membranes for evaluation of model effectiveness. We show that foundation models outperform convolutional neural networks at continuous bacterial cell envelope segmentation despite having lower performance metrics. We release the TileBac benchmark dataset on Hugging Face for further insights into model architecture development.

Deep learning neural networks provide a powerful approach for segmenting low-contrast cryogenic electron microscopy (cryoEM) images. However, model performance can vary significantly across imaging conditions and may hinder downstream quantitative analyses. Here, we present a structured evaluation workflow to systematically screen segmentation models based on performance, inference speed, robustness across imaging conditions, and reliability of downstream quantitative measurements. Using the Bacterial Cell Envelope Thickness Tool (BCET) as a test case, we evaluate multiple architectures (YOLOv11, YOLO26, U-Net, Detectron2, and SAM3) under low-dose and ultralow-dose cryoEM conditions. While several models achieve high metrics, model choice strongly influences downstream measurements of envelope thickness. Models optimized for high F1-scores may produce unreliable segmentation masks from object crowding, interpolation artifacts or imaging conditions. Our results reveal distinct trade-offs between performance, speed, and robustness amongst models. YOLOv11 provides the highest fidelity membrane segmentation for quantitative measurements and the Meta-based model SAM3 offers improved robustness under ultralow-dose conditions with competitive inference performance. This work provides practical guidance for model selection in cryoEM workflows, emphasizing that optimal choice depends on experimental priorities and downstream analysis requirements rather than metrics alone. These findings are broadly relevant to cryoEM workflows as AI-based analysis expands beyond the biological sciences. Graphical

Scientific user facilities routinely generate large-scale microscopy datasets across diverse instruments and vendors, differing substantially in file formats, dimensionality, and resolution. Beyond these inconsistencies, datasets are frequently fragmented living across isolated instruments and constrained by security policies and uneven metadata practices. Consequently, tracking, standardizing, processing, and visualizing these datasets in a manner compatible with modern machine learning and autonomous experimentation workflows remains a major challenge. While existing initiatives address data archiving, standardization, or analysis individually, few provide integrated solutions that bridge instrument-level acquisition and scalable ML workflows within heterogeneous, security-constrained user facilities. Here, we establish a deployable, facility-scale infrastructure that bridges instrument-level data generation with cloud-based ML analytics while remaining compliant with institutional network constraints. Our framework integrates on-premises cloud computing, the in-house Pycroscopy ecosystem, and an open-source metadata management platform to transform heterogeneous microscopy datasets into standardized, ML-ready representations. We demonstrate this approach across distinct microscopy modalities through end-to-end workflows encompassing metadata capture, format harmonization, automated database ingestion, segmentation-based ML inference, and interactive visualization. By structurally separating acquisition from cloud-based analysis services, the framework enables scalable model deployment and iterative refinement without direct connectivity to instrument computers. Together, this work provides a reproducible blueprint for facility-scale data and AI infrastructure, enabling ML-ready analytics, metadata traceability, and future autonomous experimentation workflows in microscopy-driven research

Atomic force microscopy (AFM) is a widely used tool for nanoscale characterization across materials science, energy research, and biology. However, its adoption in high-throughput materials discovery and statistically driven studies remains limited by a strong dependence on expert operator input and by the scarcity of annotated experimental AFM datasets needed to enable data-driven automation. Here, we introduce SimuScan, a synthetic-data–driven framework that enables reliable AFM feature identification, segmentation, and targeted imaging without requiring large manually labeled experimental datasets. SimuScan generates tunable, high-fidelity synthetic AFM images of defined morphologies while incorporating realistic experimental artifacts, including tip–sample convolution, noise, flattening distortions, and surface debris. These datasets are shown to support scalable, label-free training of modern deep learning models for AFM analysis. When integrated into data-driven AFM workflows, SimuScan-trained models can locate and analyze nanoscale structures across large datasets and guide targeted follow-up imaging. We validate this approach on nanostructured surfaces, DNA assemblies, and bacterial cells, demonstrating robust generalization across diverse sample types with minimal operator intervention. More broadly, this work establishes a general strategy for generating explicitly conditioned, task-relevant synthetic data to improve the reliability of downstream models in autonomous microscopy.

Bacteria adapt by undergoing dynamic structural changes in response to environmental cues, which are often indicative of deeper phenotypic shifts in physiology and behavior. Understanding these changes across length scales is crucial for elucidating bacterial lifecycles, informing antifouling surface design, enhancing pathogen detection, and advancing renewable energy applications. Cryogenic electron microscopy (cryoEM) enables high-resolution imaging of bacterial structures in hydrated, biologically relevant conditions. However, current quantitative analysis strategies that extract structural information from bacterial samples remain labor-intensive. This work presents an AI-driven segmentation workflow tailored to low-dose cryoEM datasets to rapidly analyze bacterial ultrastructural features from Pantoea sp. YR343, a Gram-negative bacterium isolated from the rhizosphere of Populus deltoides that forms robust biofilms along plant roots. YOLOv11 image segmentation quantifies inner and outer membrane thickness and flagella length, enabling automated analysis of bacterial ultrastructure. The workflow reliably distinguishes membranes from carbon edges of the TEM grid and contaminant crystalline ice while matching manual measurements and increasing throughput. Flagella detection routines additionally quantify nearest-neighbor proximity between bacterial envelopes and flagella. A field-of-view module further detects bacteria at low magnification for rapid screening. Together, these tools provide a scalable and automated framework for high-throughput, quantitative analysis of bacterial ultrastructure in cryoEM data.

The mechanisms by which two sister chromosomes separate and partition into daughter cells in bacteria remain poorly understood. A recent theoretical model has proposed that out-of-equilibrium central dogma reactions involving mRNA and ribosomes play a significant role in this process. Here, we test this idea in the Escherichia coli model system using high-throughput fluorescence microscopy in microfluidic devices. We compare our experimental observations with predictions from a reaction–diffusion model that includes central dogma-related reactions and excluded volume interactions between ribosomal subunits, polysomes, and chromosomal DNA. Our results show that the nonequilibrium reactions of ribosomes cause them to aggregate at the midcell, and this process facilitates the separation of the two daughter chromosomes. However, the observed effects are weaker in live cells than our one-dimensional reaction–diffusion model predicts. Rather than relying solely on active mRNA–ribosome dynamics, our data suggest that the closing division septum via steric interactions and potentially entropic forces between two DNA strands coupled to cell elongation act as additional mechanisms to ensure faithful partitioning of the nucleoids to two daughter cells.

Atomic force microscopy (AFM) underpins nanoscale research across materials, energy, and biology. Yet its utility is restricted by low imaging throughput and the requirement for a highly skilled operator. Progress toward autonomous operation is hindered by the scarcity of annotated datasets and the poor generalizability of existing models. To overcome this, we introduce a synthetic data generator tailored for AFM that produces tunable, high-fidelity images of arbitrary morphologies while embedding realistic artifacts such as tip convolution, noise, flattening errors, and debris. SimuScan datasets enable scalable, label-free training of deep learning models including YOLOv8, U-Net, and Mask R-CNN. Integrated with autonomous AFM control, the trained vision system can not only locate, segment, and analyze nanoscale structures across millimeter-scale areas in real time, but also recursively act on them by acquiring new images or adapting scan parameters, achieving robust performance without human labeling nor experimental intervention. We validate these autonomous closed-loop capabilities on nanostructures, DNA, and bacterial cells, demonstrating reliable high-resolution imaging with minimal operator input. By coupling synthetic data generation with deep learning and adaptive AFM control, this platform accelerates imaging workflows, broadens accessibility, and creates opportunities for high-throughput applications in nanomaterials discovery, biomedical diagnostics, and wafer-scale inspection.

Combinatorial magnetron sputtering and electrical characterization were used to systematically study the impact of compositional changes in the resistive switching of transition metal oxides, specifically the Zr_xTa_1−xO_y system. Current-voltage behavior across a range of temperatures provided insights into the mechanisms that contribute to differences in the electrical conductivity of the pristine Ta₂O₅ and ZrO₂, and mixed Zr_xTa_1−xO_y devices. The underlying conductive mechanism was found to be a mixture of charge trapping and ionic motion, where charge trapping/emission dictated the short-term cycling behavior while ion motion contributed to changes in the conduction with increased cycling number. ToF-SIMS was used to identify the origin of the “wake-up” behavior of the devices, revealing an ionic motion contribution. This understanding of how cation concentration affects conduction in mixed valence systems helps provide a foundation for a new approach toward manipulating resistive switching in these active layer materials.