Nowadays, deconvolution in cell and tissue imaging has matured into a standard restoration technique that is accessible to large fraction of the microscopy community thanks to steadily improving algorithms. Still, deconvolution is often the rate-limiting step in the analysis of the acquired data, even at today's computer performance. Here, we present the Huygens Remote Manager, an open-source, efficient, multi-user web-based interface for parallel batch deconvolutions.

Processes at the subcellular level can today be followed in vivo, dynamically, and imaged in two or three dimensions. The choice of the microscopy technique to use for the application of interest, however, depends on a variety of parameters, like the resolution limit, the sensitivity of detection, the speed of acquisition, the viability and thickness of the specimen, and even the price of the microscope. As a consequence, often no single microscope system or technique will be best, and compromises will have to be made, often resulting in a less-than-optimal image resolution and/or signal to noise ratio.
Deconvolution and image restoration in general [1, 2, 3] are computer processing techniques designed to overcome the limitations of the different microscope modalities. They make use of models of the optics of the microscope (the 3D point-spread function), the noise, and to some extent also of possible aberrations.
The Huygens Deconvolution Software by Scientific Volume Imaging (http://www.svi.nl) is the most advanced deconvolution package available today (fig. 1). It can process images from all current optical microscope types and can run on a large number of processors in parallel. As one of its products, SVI offers Huygens Scripting, an engine that exposes all processing functionality of the Huygens Compute Engine (the processing kernel at the core of the Huygens software) for the automated restoration of datasets through scripts written in the Tcl scripting language. This allows the processing of datasets in batch mode locally or remotely on a dedicated server. However, it requires the users to program their own scripts, which is both time-consuming and error-prone.

Furthermore, this approach cannot efficiently scale to larger user bases, since it lacks a mechanism to organize the jobs into a queue and forces users to work with a first come first served policy.
To address the limitations of Huygens Scripting for a multi-user environment, we wrote the Huygens Remote Manager (HRM), a web-based deconvolution platform primarily targeted to microscopy and imaging facilities in academic and industrial environments that must handle large user bases.

HRM consists of two main components, both written in PHP: a web-based interface to the Huygens Compute Engine and a queue manager. Briefly, the web interface allows users to create deconvolution jobs from remote through any reasonably recent web browser. The thus created jobs are dispatched by the queue manager to any of the dedicated servers running the Huygens software.
In more detail, HRM requires a web server with a database backend (to store deconvolution parameters, job descriptions and, optionally, user accounts), a file server (to temporarily store input and restored datasets), and one or more processing servers running the Huygens software. The setup is highly configurable, since (i) servers and queue manager can either be all hosted by the same machine (fig. 2 a) or be distributed onto two, three or more computers (fig. 2 b); (ii) a large number of databases are supported through the ADODB database abstraction library for PHP (http://adodb.sourceforge.net); (iii) parts of HRM are optional, as is the case for the embedded user management system, that can be turned off if user administration is required, for example, at facility level.
Figure 3 shows the HRM welcome screen. Here, existing users can login and new users can register. After login, the user can create and launch deconvolution jobs. All parameters and settings are stored in the database (fig. 4), therefore creating a personalized working environment for each user. This way, repeating jobs can be created with just a few clicks.
A job definition consists of three parts (fig. 4): (i) a set of optical parameters that allow the Huygens software to calculate a model of the 3D PSF for deconvolution or, alternatively, a pointer to a file containing a measured PSF; (ii) a set of task settings, that define the parameters for restoration; (iii) the input files list to be processed and the file format of the results. The created job is stored into the database, from which the queue manager will pick it and dispatch it to the first free processor on the dedicated server(s). The status of the queue can be visualized at any time through HRM. At the end of the job, HRM will inform the user via a summary e-mail that the restored datasets are ready on the file server to be retrieved.