Elsevier

IRBM

Volume 38, Issue 6, November 2017, Pages 325-338
IRBM

Original Article
High Resolution Image Registration for Micro-Colonies Monitoring on Petri Dishes

https://doi.org/10.1016/j.irbm.2017.10.004Get rights and content

Highlights

  • Real-time monitoring of micro-colonies in Petri dishes by high resolution imaging.

  • Image registration algorithm validated at 95.8% on 1154 Petri dishes samples.

  • Detection of acquisition defects and of agar breaks.

  • The device is currently being commercialized.

Abstract

Background To reach faster diagnosis of sanitary control and to reduce human manipulations, rapid microbiology develops methods and systems allowing to detect these contaminations early. The company Advencis conceived an innovative device able to follow in real-time the contaminations during the incubation process.

Methods The aim is to keep the traditional method (Petri dishes with lid), and this, without any modification of the sample. This method is based on the monitoring of the Petri dishes surface by image processing while incubating the samples. The iterative registration method developed here is based on small objects lying on the growing surface.

Results A large number of image sequences of various samples were tested. The results show that this method is quite efficient to follow and detect the variations on the sample.

Conclusion Some specific cases may be improved in a future algorithm version.

Introduction

Rapid microbiology methods (RMMs) can provide results in hours rather than days for classical techniques. For industrial quality control RMMs give opportunity to improve product process. For clinical diagnosis, rapidity in getting results often allows for defining the most appropriate treatment earlier, and thus to save lives. The Advencis company, a bioMérieux subsidiary, designed an innovative system for real-time detection and monitoring of contaminations during the incubation process [1]. The aim of the method is to rely the compendial method and its standard consumables (Petri dishes), while keeping identical both workflow and sample (no manipulation, closed lid, no staining step, etc.). This method is based on high resolution imaging of the Petri dishes surface. A scanning technique is used to acquire surface image of the Petri dishes. This technique induces lower-than-1 mm shifts between two consecutive images. To follow the growth of the microorganisms, a precise image registration method is needed. This article presents the registration method developed specifically for this application and evaluated on a significant number of real data.

Firstly, rapid microbiology and its interest are presented. Then, different instruments used in rapid microbiology are introduced and their pros and cons are analyzed. Subsequently evaluating the need for a new system, the Advencis instrument is described. The image registration need is described by introducing the mechanical offset related to the system and the offset of biological origin. Several existing registration methods are presented before developing our own solution. A validity control of this registration method is presented, along with the obtained results, which are finally discussed.

Biology consists in studying bacteria, yeast and fungi. Those are microorganisms not visible by naked eye, with a size of 0.5 to 10 micrometers. Some of these germs are pathogenic for humans and animals. It is therefore important to be able to detect contamination in food, medicines and others. The detection of bacterial micro-colonies is done in fields ranging from sanitary control in food industries to patients infection diagnosis in clinical field to sterility control of medicines and vaccines. Traditionally, the major part of these tests is done into Petri dishes. The samples under interest are disposed on a nutritional medium where the eventually present living cells can multiply to form micro-colonies with a diameter of 20 to 500 micrometers (μm), then visible colonies when larger. These biological samples are incubated with a constant temperature during 4 to 16 days [2]. Once the germs have grown, microbiologists study the dishes by naked eye or by using a magnifying glass. The Petri dishes can also be analyzed by colony counters, especially when an important concentration of germs prevents from counting them precisely, or if too many samples need to be counted [3], [4], [5], [6]. They usually detect colonies from a size of 500 μm and above, but the Scan® 1200 from Interscience claims a minimal detection size of 50 μm. However, there is a lack of counting precision at these sizes (high number of false positives). To detect germs precociously, with an interesting cost per test, rapid microbiology devices are developed.

The samples are usually analyzed in solid or liquid media. The contamination detection time can be significantly reduced to a few hours for molecular methods like PCR (Polymerase Chain Reaction), which directly targets microorganisms DNA, compared to a few days for culture methods on solid media (agar media). However, all kind of samples cannot be reached with this method. Furthermore, in liquid media, cells are spread around, and they are usually detected at a concentration of 105–106 cells per milliliter (mL), whereas the solid state tests allow for segregating individual cells, when forming micro-colonies, at a concentration of 102–103 cells per mL. Different actors using solid states tests can be found on the rapid microbiology market (Cf. Fig. 1). Three systems using solid media closest to the Advencis system are compared by their observing frequency. Others solid media systems developed by bioMérieux, the ChemScan using cytometry [7] and BacTAlert [8] using colorimetry, are much more different from the systems Quantum, Milliflex® and GrowthDirect, so are not considered in this work.

The Quantum device, from early 2010s, allows for the analysis of a single Milliflex® dish, a specific dish format. It uses a viability staining that is metabolized by the cells to obtain a fluorescent signal [9]. This instrument, developed by Merck-Millipore, has a detection time around 21 to 48 hours. When the sample is placed into the device, the microorganisms are revealed by fluorescence and a picture of the sample can be taken by an embedded camera. However, the staining process is coercing. This system is approximately two times faster than the traditional method by naked eye.

The Milliflex® Rapid device, the forerunner of the Milliflex® Quantum, was also developed by Merck-Millipore in 2008. It uses the natural bioluminescence process, measured by ATP-metry [10]. A specific protein, the adenosine triphosphate (ATP), is sprayed on the filterable membrane. It is a label of cell viability. Milliflex® Rapid allows for the analysis of a single Milliflex® sample at a time. The contamination detection time is around four times faster than the traditional method.

The Rapid Microbio System is one of the first rapid microbiology devices, invented in 2004. It is not yet widely spread, because of its high cost. It combines auto-fluorescence and real time image processing [11]. The samples are automatically placed under the optical sensor. Auto-fluorescence images are regularly taken. The growing of luminous objects is looked for to validate a contamination. The dishes have a specific format. The detection time is here two times faster than the traditional method.

Working in solid state on traditional Petri dishes allows for analyzing samples that cannot be transformed into liquids (food, tissues), and moreover allows for visualizing the samples and the micro-colonies of bacteria. Staining (color, fluorescence) makes the detection faster but is potentially destructive, complicates the preparation workflow, creates the risk of a cross-contamination and increases the cost per test. A method avoiding additional handlings of the samples may thus be useful. Using Petri dishes allows for maintaining the same manipulation workflow and for visually checking the sample at the end of the incubation. This involves a method that allows rapid detection of sample contamination and naked-eye observation at the end of incubation. The proposed solution may use the growing of these micro-colonies to eventually define a contamination of the sample. A first detection of these elements at a size between 40 and 150 μm leads to a confirmation by naked eye at a size between 150 and 300 μm few images later, considering the cells replication rate. For the moment no industrial device respects these specifications.

Section snippets

Description of the Advencis device

The Advencis company developed a rapid microbiology system (Cf. Fig. 2a) for analyzing Petri dishes. Those are disposed on a tray, placed on a drawer. The samples are incubated inside a temperature-controlled incubation chamber. An embedded image acquisition system provides a real time monitoring of the development of the microorganisms. The Advencis Bio-System Software (Cf. Fig. 2b) allows for choosing the test parameters (temperature, number and sample size, etc.), launching the images

Results

The performances of the GIR method are measured on the different control levels available, summarized below (Cf. Table 4).

The results given in the tables of this section take into account the last control level available, some sets having additional steps. The intermediary results for previous control levels are detailed only if relevant for the considered set.

Discussion

According to our results, the GIR algorithm works on the wide majority of the tested samples. The developed registration method is able to detect two combined (⊕) displacements into the sample's images (for instance: mechanical ⊕ (dish + lid)/tray; mechanical ⊕ lid/dish; mechanical ⊕ agar/dish). In the case of a shock on the instrument, three displacements on the sample are typical (mechanical ⊕ dish/tray ⊕ lid/dish; mechanical ⊕ dish/tray ⊕ agar/dish; mechanical ⊕ lid/tray ⊕ agar/dish). This

Conclusion

Good registration leads to good superimposition of the objects present on the growing surface along the image sequence. Thus, the following steps of pairing and tracking of all the objects detected on the growing surface, including micro-colonies of living organisms, can be done with confidence. This allows for identifying easily growing objects as living objects, and therefore alerting the user of a contamination of the sample by microorganisms, as expected. The Petri dishes monitoring is

Ethical and disclosure statement

Not concerned: All data are coded anonymously, by date, no animal testing.

Contributors, authorship and role of the funding source

Empty CellStudy conceptionData acquisitionData analysisArticle draftingArticle intellectual contentFinal approvalSubmission authorization
Marlena Betzner (Advencis)YESYES (part)YESYESYESYES/
Joseph Pierquin (Advencis)YESYES (part)//YESYES/
Sophie Kohler (UHA)///YESYESYES/
Alain Dieterlen (UHA)///YESYESYES/
bioMérieux/////YESYES

Declaration of interest

  • 1.

    Marlena Betzner: Advencis employee (industrial thesis).

  • 2.

    Joseph Pierquin: Advencis employee (manager).

  • 3.

    Sophie Kohler: Conflicts of interest: none.

  • 4.

    Alain Dieterlen: Conflict of interest: none.

Acknowledgements

This work was supported by the Advencis company and the ANRT (Association Nationale pour la Recherche et la Technologie, Grant number 2012/1471).

References (21)

  • D.P. Mukherjee et al.

    Bacterial colony counting using distance transform

    Int J Bio-Med Comput

    (1995)
  • J. Marotz et al.

    Effective object recognition for automated counting of colonies in Petri dishes

    Comput Methods Programs Biomed

    (2001)
  • B. Zitová et al.

    Image registration methods: a survey

    Image Vis Comput

    (2003)
  • Council of Europe

    European pharmacopoeia

    (2013)
  • J. Dahle et al.

    Automated counting of Mammalian cell colonies by means of a flat bed scanner and image processing

    Cytometry Part A

    (2004)
  • S.D. Brugger et al.

    Automated counting of bacterial colony forming units on agar plates

    PLoS ONE

    (2012)
  • A. Onadipe et al.

    A method for the rapid detection of microbial contaminants in animal cell culture processes

    PDA J Pharm Sci Technol

    (2001)
  • T.C. Thorpe et al.

    BacT/Alert: an automated colorimetric microbial detection system

    J Clin Microbiol

    (1990)
  • A. Baumstummler et al.

    Detection of microbial contaminants in mammalian cell cultures using a new fluorescence-based staining method

    Lett Appl Microbiol

    (2010)
There are more references available in the full text version of this article.

Cited by (0)

View full text