Intelligent Imaging

ImageJ drives primary scan (multiple fields of view tiling the sample), image analysis and secondary scan (detected targets) on a Leica SP5 confocal microscope or a widefield microscope driven by Micro-Manager.

Motivation and Benefits

In High-Throughput Microscopy experiment, automation makes the life of the user much easier, the microscope acquires in a controlled manner over many fields of view and the user does not need to stay at the microscope. However, depending on the type of experiment, not all cells (or specimen in general) are actually relevant to the experiment. Take a classical example, if you study mitosis but cannot synchronize your cells, how many cells will be in mitosis when you fix your sample ? And among these, how many will be in metaphase ? Issues raise if you need to perform high resolution microscopy of cells in metaphase ! A regular screening microscope will just record every position at high resolution, you will generate 100 GB of data, and in the best case only 1GB of this data will be relevant to you, while you will trash 99GB to the bin, together with 99% of the hours booked at the microscope to generate junk data. In microscopy facilities, time at the instrument is shared among users and inefficient experiments affect everyone.

To tackle these issues, we developed an application which takes control over the microscope to teach it which specimen is relevant, and which is to be left in peace. Written in the ImageJ environment (in java and in macro language), the application communicates through the CAM interface (Leica, available add-on module for confocal and widefield microscopes), and performs image processing on a primary 'Low-resolution' scan to determine where the relevant specimen are. After positive hits detection, ImageJ commands the confocal microscope to go back to these positions for a secondary 'High Resolution' scan. Only relevant images are acquired at high resolution, hence shortening both the global duration of the experiment and the amount of data stored. The protocol also allows for more reproducible acquisition parameters.

Hence, from the side of the user, samples are not screened manually through the ocular for the right target, no analysis needs to be performed on huge datasets to select and keep the relevant data, and data transfer is faster. The length and cost of the experiments are significantly reduced for the same outcome, and finally the use of the microscope is efficiently optimized.

Example Applications and Technical Description

Documentation

See AutoScanJ repository.