Experiments In Extracting Digital Slides From Video
While digital pathology systems are becoming more and more affordable and thus diffused, for occasional, non diagnostic acquisitions (e.g., teaching) it might not be economically viable to buy a scanner. In addition to that, it is relatively common that videos from microscope are published for teaching purposes in particular in fields related to medical imaging . Recently, a research group published a series of papers related to digital slide acquisition from non-automated microscopes 2 , 3. The developed technique allows to reconstruct a digital slide starting from a video flow acquired during slide examination. The proposed method seems interesting, and stimulated us in investigating further alternatives. In fact, the proposed process very closely resembles the creation of panoramas from multiple pictures through stitching, or the simulation of medium format photog- raphy through the Brenizer method  .
Aim of the present work is to explore techniques for extracting digital slides from microscope video record- ings, possibly using already available and open source software.
We searched for panorama stitching software and found a number of packages, including a widely used open source software (Hugin)  . We concentrated on it and on a commercial product (AutoPano). We designed a two steps experiment:
1) For a first approach, we decided to use a screencast of a digital slide. This to reduce the possible sources of error, and also to have a gold standard to compare with.
2) At a second step, we recorded video from a microscope (Olympus Provis AX70) using a commercial camera (Sony Nex-6) on an easily replicable hardware setup (fig.1).
We devised a sequence of steps to obtain digital slides as if they were panoramas recorded in video. This included pre-processing of the video source, and selection of parameters of the software for best reconstruction
The designed workflow include:
- record a video from the microscope while browsing a case, taking care in covering the whole sample in a continuous way.
- Transform the video in a sequence of images to feed the stitching software. This can be done with free software like ffmpeg and derivatives. It is not needed to have every frame of the video: the number is related to the speed of examination. In our experiment, 2-4 frames per second have been sufficient.
- With Hugin (or AutoPano), open all the images and set them as if they were captured with a teleobjective with very long focal length (e.g., 1000mm). This to minimize abherrations in reconstruction.
Figure 1 shows a digital slide obtained with Hugin from digital microscope screencast; figure 2 shows a digital slide obtained from a real microscope. The former is much better, and this indicates a possible way for occasional scanning.
Open issues include:
- the method should be validated by comparing with a gold standard, like a digital slide of the sample;
- if the focus position changes during the video, focusing should occur and further processing could be needed for applying also focus stacking to the operations.
- While looking at a slide at higher magnifications, it is difficult to ensure to visit the whole sample during visualization. From this point of view, the realtime implementation presented in  allows for continu- ously checking what it has been acquired.
- This work has been partially founded by the EU FP7 program within the Marie Curie Project “AIDPATH”, grant number 61247
 Gherardi A, Bevilacqua A, (2013), Real-time whole slide mosaicing for non-automated microscopes in histopathology analysis, J Pathol Inform, 1-6, 4
 Gherardi A, Bevilacqua A, (2014), Manual stage acquisition and interac- tive display of digital slides in histopathology, IEEE J Biomed Health Inform, 1413-1422, 18
 Naryškin R, (2012), Advanced Photography Technique: Brenizer Method Panorama, Photography Life, https://photographylife.com/advanced-pho- tography-techniques-brenizer-method-panorama, 2016-01-30
 VV AA, (2015), Hugin, http://hugin.sourceforge.net, 2016-01-30
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
Authors who publish with this journal agree to the following terms:
1. Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
2. Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
3. Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).
4. In case of virtual slide publication the authors agree to copy the article in a structural modified version to the journal's VS archive.