The platform

The platform we are developing is an AI-driven, fully automated system for high-throughput 5D fluorescence imaging and data analysis. It is designed to overcome the major limitations of current microscopy approaches:

• Low throughput (typically one cell at a time).
• Heavy user dependence (manual training and oversight required).
• Insufficient data volume for advanced ML/AI applications.

By integrating cutting-edge optics, robotics, and artificial intelligence, the platform will autonomously acquire, process, and analyse very large datasets of living cells (>10,000 cells/condition) in 5D (x, y, z, time, and multiple spectral channels) with high spatiotemporal resolution

Core components

1. Optical Imaging System

• Microscope design: A custom single-objective light-sheet system, based on an improved oblique plane microscopy (OPM) concept.
• Capabilities:
  • Simultaneous multichannel volumetric imaging of large fields of view (300 × 300 × 20 μm³).
  • Spatial resolution of ~300 nm (xy) and ~800 nm (z), with acquisition speeds of 100–500 ms per cell volume.
  • Parallel acquisition of ≥60 cells with ≤30 s time resolution.
• Features under development:
  • OP-SLIM (Oblique Plane Scattered Light-Sheet Imaging): A novel approach to generate high-contrast volumetric reference images in parallel with fluorescence imaging, useful for ML-based label-free inference.
  • Structured Illumination Microscopy (SIM): Integrated with OPM to achieve ~150 nm lateral resolution while maintaining throughput.

2. Automation and Robotics

• Automated sample handling:
  • Robotic arm with gripper for moving multiwell plates (96-well or higher).
  • Barcode readers for sample tracking.
  • Automated pipetting and drug application systems.
  • Temperature-controlled microplate storage (“hotels”).
• Live-cell environment:
  • Automated stage-top incubator with CO₂ and temperature control.
  • Silicone-oil dispenser for long-term, stable imaging.
• Continuous operation: 24/7 autonomous imaging, enabling collection of 4,000–10,000 cells per day.

3. Software Framework

• FPGA and Python-based control for all hardware components.
• User interface:
  • Graphical User Interface (GUI) for experiment setup and data visualization.
  • Python API for extensibility and integration of custom hardware/software.
• AI integration:
  • Automated cell search and selection based on user-defined criteria.
  • Adaptive imaging parameter optimisation (laser power, exposure, sampling strategy).
  • Natural language interface (in development) allowing users to interact with the system using written instructions.

Data Output and Integration

• 5D datasets: Large-scale single-cell volumes in multiple fluorescent channels, aligned with reference label-free imaging.
• Metadata: Complete acquisition parameters, environmental conditions, and experimental annotations.
• Analysis-ready: Data streamed directly into ML pipelines for segmentation, generative modelling, and population-level statistical analysis.
• Throughput: Orders of magnitude higher than current high-resolution microscopy approaches, providing the scale required for AI/ML-driven biological discovery.

Applications

The platform is designed to be modular and generalisable, making it suitable for a wide range of biological applications. Our platform will provide a powerful tool for any research area requiring large-scale, high-resolution live-cell imaging, including cell signalling, developmental biology, and systems pharmacology.