Design

Problem

Research laboratories are constantly developing new microscopes that are are complex systems that rely on the tight integration between many components. Instead of focusing on the microscopy, these teams spend substantial time and effort in software development to control their new microscopes. Such software is limited to the specific microscope setup and components and becomes unshareable, difficult to maintain, or incur ongoing licensing costs. A huge amount of time, effort, and resources is wasted as scientists in different labs, and even in the same lab, repeatably implement solutions to the same problems.

Project aims

  • Community targeted. We want to use this ourselves but we could write something simpler if it was only for use by ourselves. We want other people building or using microscopes to be able to use it and contribute back. This means we should consider use cases other than our own on design choices.
  • Free (libre) software. Even though we consider other use cases when designing it we will still be resources limited on what we can implement. The users should be able to help themselves and contribute the fixes back to everyone else.
  • Ease of use. Microscopes are built by physicists and biologists. They are the main target for Microscope, not software engineers. They should be able to use and install it without much trouble.

Choices

Why not using µManager

µManager is an existing open source software for the control of microscopes. It is written in Java and based on top of ImageJ, an open source image analysis program. However, µManager design has issues when controlling more complex microscopes where devices are spread over multiple computers, with multiple cameras, and devices are synchronised with TTL signals. In addition, while Java makes it easier to access ImageJ it makes more difficult to use the whole of numpy and scipy.

Python

Python has multiple features:

  1. it is widely used in the scientific community. This increases the odds that users of Microscope will be capable to participate in its development.
  2. unlike other widespread languages in the scientific community, it is a general purpose programming language and not mainly for numerical or symbolic computation.
  3. while Python is not firstly a language for numerical computations, numpy and scipy are the basis for this. Most algorithms for image analysis are available in Python.

Use cases

Microscope GUI

This provides the device and experiment interface that a GUI microscope interface would need to be viable.

Devices on local machine

The simplest typical microscope. All devices are controlled from a single computer, the same computer where the user is. The synchronisation between devices is all done in software.

Devices over multiple computers in local network

The user is in one computer but the devices are actually connected to multiple other computers. This may be for performance but also because different microscope devices may require different incompatible OS.

Synchronisation of devices during an experiment will likely be performed by a separate device with high time precision based on a table of events.

Controlling an independent device

There is no actual a microscope, only experiment with a single device. For example, just testing of a camera or deformable mirror. This device may be in a local or a remote machine.

Programmed image acquisition based on image analysis

Automated image analysis can make decisions during image acquisition. For example, scanning slides for specific features; tracking of moving particles; and automatically changing imaging parameters over time.

Multiple microscopes controlled by one central server

If image acquisition is automated, a single system can automatically control multiple microscopes without user interaction.

Realities

We have to deal with the reality which is less perfect and puts limits on our implementation and design of Microscope.

State machine

Devices do not report back their state which prevents from modelling Microscope as a state machine. For example, users will change objectives or move the stage.

History

Development of Python Microscope started at Oxford Micron Bioimaging Unit to provide remote control of microscope devices independent of hardware specifics. Locally, development was guided to support development of a new version of cockpit, a graphical user interface for the control of microscopes.