The study of protozoa population in wastewater treatment plants by image analysis

Abstract

Protozoa are important microorganisms for the ecosystem balance in wastewater treatment plants. A procedure for their semi-automated identification and counting based on image analysis is proposed. The main difficulty is segmentation of the protozoa as most of them are in contact with the sludge. The protozoa are characterized by the size of their silhouette (area and length) and three shape factors (elongation, circularity and eccentricity). They are identified after projecting the resulting 5D space into a 3D space of principal components. The rate of automated identification is actually higher than 50% for some of the species commonly found in activated sludge.

Protozoa; wastewater treatment; image analysis


THE STUDY OF PROTOZOA POPULATION IN WASTEWATER TREATMENT PLANTS BY IMAGE ANALYSIS

M.da Motta1, M.N.Pons1* * To whom correspondence should be addressed , H.Vivier1, A.L.Amaral2, E.C.Ferreira2, N.Roche 1,3 and M. Mota2

1 Laboratoire des Sciences du Génie Chimique, CNRS-ENSIC-INPL,

1, rue Grandville, BP 451, F-54001 Nancy Cedex, France, Phone: +33 3 83 17 52 77,

Fax: +33 3 83 17 53 26, E-Mail: Marie-Noelle.Pons@ensic.inpl-nancy.fr

2 Centro de Engenharia Biológica - IBQF, Universidade do Minho, P-4710 Braga, Portugal

3 IUT de Marseille, Dept. GCGP, UPRES 884, Université de Marseille, Travers C. Susinni,

BP 157, F-13388 Marseille Cedex, France

(Received: October 5, 2000 ; Accepted: January 15, 2001)

Abstract - Protozoa are important microorganisms for the ecosystem balance in wastewater treatment plants. A procedure for their semi-automated identification and counting based on image analysis is proposed. The main difficulty is segmentation of the protozoa as most of them are in contact with the sludge. The protozoa are characterized by the size of their silhouette (area and length) and three shape factors (elongation, circularity and eccentricity). They are identified after projecting the resulting 5D space into a 3D space of principal components. The rate of automated identification is actually higher than 50% for some of the species commonly found in activated sludge.

Keywords: Protozoa, wastewater treatment, image analysis.

INTRODUCTION

The efficiency of wastewater treatment by activated sludge is related to the bacterial population but also to the protozoa (Nicolau et al., 1997). Different species can be found and have been listed by various authors: Curds and Cockburn (1970a), Martin-Cereceda et al. (1996), Richard (1991), Sasahara and Ogawa (1983), etc. Under normal conditions their concentration is larger than 106 protozoa/L. 107 protozoa/L corresponds to a very good pollution abatement. On the contrary, concentrations lower than 105 protozoa /L are indicative of the low efficiency of the plant (Drakides, 1978). In terms of biomass, protozoa represent between 0.17 and 0.44% of the sludge during the colonization phase but can represent up to 9% at steady state (Madoni, 1994a). Curds and Cockburn (1970b) established relationships between the abundance of some species and the sludge loading and associated the species with the quality of the effluent according to the biological oxygen demand (BOD). Table 1 summarizes the predominant groups of protozoa as a function of organic loading. These protozoa have an important role in maintaining a good balance in the biological ecosystem: they eliminate the excess bacteria and stimulate their growth and they promote flocculation (Gerardi et al., 1995). By consuming the free bacteria, they help to decrease the turbidity of the effluent as well as its BOD and its suspended matter content (Curds et al., 1968).

Most of the protozoa found in the sludge are ciliated and they can be classified in four main groups: free-swimming, crawling, attached and carnivorous. Table 2 shows that the predominance of one group or the other can be an indicator of the efficiency of a wastewater treatment plant using activated sludge. Several authors have applied statistical methods to express the relationships between the protozoa and the operational conditions of the plants. Martin-Cereceda et al. (1996) used a partial correlation analysis to examine the protozoa of ten wastewater treatment plants in Madrid (Spain) and established relationships between the protozoa and plant efficiency (effluent quality and settleability). Using principal component analysis (PCA) with Varimax rotation, Genoveva et al. (1991) expressed 73% of the process variability in terms of six principal components: the first of these components explains 25% of the variability and takes into account the ciliates.

The protozoa identification and counting needed for the previously mentioned studies was done manually; this is a very tedious task for an expert. Amaral et al. (1999) developed a procedure for the semi-automated recognition of protozoa by image analysis. The image analysis section, called ProtoRec V0, is embedded into a VisilogÔ 5.1 environment (Noésis, Les Ulis, France). The results (size and shape descriptors) are later analyzed by a multivariate method (PCA) for the identification of the protozoa from a database. This procedure was validated using samples regularly taken in a full-scale municipal wastewater treatment plant over a summer period of two months (June and July 1998). However, since that date, other species have been noticed in the samples and the amount of filamentous bacteria has increased drastically, which causes problems in image treatment. Here a new version is developed to cover the filamentous bacteria and to increase the size of the database.

MATERIALS AND METHODS

Sampling and Image Grabbing

Sludge samples are regularly taken in the wastewater treatment plant of Nancy-Maxéville (350 000 PE). The delay between sampling and image grabbing is about 30 min. The image grabbing system is based on an optical microscope (Leitz Dialux 20) and a monochrome camera (Hitachi CCTV) connected to a PC via a Matrox Meteor board. A mixed liquor drop is deposited on a glass slide and carefully covered with a slip to avoid any mechanical stress on the microorganisms. For most images a 400x magnification (normal illumination) is used, except in the case of sets of protozoa (Opercularia for instance) or large rotifers, where a 250x magnification is needed. For each sample 50 images of live protozoa are grabbed by a systematic examination of the slide.

Image Treatment

The procedure is called ProtoRec V1 and it is implemented in VisilogÔ5.1: its aim is the calculation of size and shape parameters describing the silhouette of the protozoa. The gray-level image is pre-treated to enhance the contours of the protozoa and is segmented. This is a key step as many protozoa are in contact with the flocs and validation by the operator is requested at some points of the procedure. The main steps are presented in Figure 1.

Measurements

The protozoa are characterized by their size (projected surface, A, and length, L, given by the maximal Feret diameter, Fmax) and shape descriptors (elongation, FS, circularity, C, and eccentricity, E, calculated from the second-order moments (M2x, M2y and M2xy)):

where P is the perimeter of the silhouette

The presence of a flagellum or a stalk is helpful in the identification step, but it is not always possible to obtain complete protozoa (with flagella or stalk).

Figure 2 gives the percentage of each species present in the database. From the total population of protozoa a training set has been defined, with protozoa identified by an expert (Jahn et al., 1979; Madoni, 1994b). A principal component analysis (PCA) (XlstatÔ, T. Fahmy, Paris, France) is run on the training data se, which contains several individuals of 14 protozoa species, to take into account the variability within each species (Einax et al., 1997).

RESULTS

Table 3 give the eigenvalues obtained from the correlation matrix. The first two principal components, f1 and f2, explain 79% of the variability in the training data set. With three components, f1, f2 and f3, 95% of the variability can be explained. No further improvement is obtained by addition of another component.

The correlation circle (Figure 3) summarizes the relationships between the variables. They are relatively well distributed, indicating that these descriptors can really help to discriminate between the species. As seen in Table 4, L, E and C have a strong effect on f1, A on f2 and FS on f3.

Equations 4 to 6 give the relationships between the coordinates in the principal component space (Coij) for each protozoa species i along axis j.

where mi is the mean value taken for parameter i for the whole set of protozoa and si, the corresponding standard deviation.

In Figure 4 the average position of each species has been plotted in the 3D space of the principal components. It can be seen that V. microstoma without stalk, Aspidisca and Colpidium are very close one to another. V. microstoma can be isolated when its stalk is considered. The same improvement can be obtained for V. convalaria and Opercularia; the stalk makes identification easier.

The location of each species and the standard deviation due to the variability within each species are given in Table 5. Flagella and stalks increase the standard deviations as they can have various positions, but they nevertheless improve identification as the average positions differ considerably, depending on whether or not the stalk is considered. The recognition rate doubles when the stalk can be taken into account. Peranema exhibits very large standard deviations along the three axes due to its small size, its flagellum and its mobility.

Figure 5 gives the percentage of each protozoa present, imaged during one week and identified by the operator. Some species have not been included in the database yet and about 22% of the protozoa could not be clearly identified. The semi-automated recognition is applied only to the protozoa previously identified by the expert. The protozoa coordinates in the PCA space are computed using equations 4 to 6: the distance of each protozoa to the characteristic position of each species, as given in Table 5, is calculated. The protozoa is assigned to the species for which the distance is minimal. The results obtained by the automated classification are compared with those found by the operator. Figure 6 gives the rate of successful recognition for the species included in the database.

The rate is larger than 50% for Zoothamnium, Microstoma and Convallaria, which are relatively abundant in the population, as well as for Trachelophyllum and Tetrahymena. Some species are particularly difficult to recognize: Peranema, Chilodonella and Aspidisca (Figure 7a and b). Peranema and Chilodonella are new species that have recently been introduced into the database and the limited number of individuals could be a reason for the bad rate of recognition. Aspidisca is a small protozoa which is often found over the sludge flocs (Figure 7c and d).

CONCLUSIONS

Protozoa are known to be an important indicator of the efficiency of wastewater treatment plant. However, their manual identification and counting is a tedious task. A procedure was developed to perform these tasks semi-automatically. Segmentation of the protozoa from the sludge flocs is a key step in the image treatment, which cannot be fully automated at this point. Identification is based on size and shape descriptors of the protozoa silhouette. A database of several individuals belonging to 14 protozoa species was built. A multivariate analysis of the descriptors is used for the identification of the protozoa.

Although the procedure needs improvements, the initial results are promising. Further work is currently being conducted to improve the method of segmentation of the images and identification by introducing new shape descriptors to characterize the silhouette of the protozoa. In parallel the database is being gradually enlarged by the addition of new protozoa and introduction of metazoa such as nematodes.

ACKNOWLEDGEMENTS

The authors are thankful to the National Council of Scientific and Technological Development of Brazil (CNPq), the Embassy of France in Portugal and ICCTI.

  • Amaral A.L., Baptiste C., Pons M.N., Nicolau A., Lima N., Ferreira E.C., Mota M. and Vivier H., Semi-Automated Recognition of Protozoa by Image Analysis, Biotechnology Techniques, 13, 111-118 (1999).
  • Coster M. and Chermant J.L., Précis dAnalyse dImages, 2nd edition, CNRS Editions, Paris (1989).
  • Curds C.R., Cockburn A. and Vandike J.M., An Experimental Study of the Role of the Ciliated Protozoa in the Activated-Sludge Process, Wat. Pollut. Control, 67 , 312-329 (1968).
  • Curds C.R. and Cockburn A., Protozoa in Biological Sewage-Treatment Process I. A Survey of the Protozoan Fauna of British Percolating Filters and Activated-Sludge Plants, Water Research, 4, 225-236 (1970a).
  • Curds C.R. and Cockburn A., Protozoa in Biological Sewage-Treatment Process II. Protozoa as Indicator in the Activated-Sludge Process, Water Research, 4, 237-249 (1970b).
  • Drakides C., LObservation Microscopique des Boues Activées Appliquée ŕ la Surveillance des Installations dEpurations : Technique détude et Interprétation, TSM-LEAU, 2, 85-98 (1978).
  • Einax J.W., Zwanziger H.W. and Geiss S., Chemometrics in Environmental Analysis, VCH, Weinheim (1997).
  • Genoveva E., Tellez C. and Bautista L.M., Dynamics of Ciliated Protozoa Communities in Activated-Sludge Process, Wat. Res., 25, No. 8, 967-972 (1991).
  • Gerardi M.H. et al, Wastewater Biology: The Microlife, A Special Publication, Water Environment Federation, 2nd edition, Alexandria, Virginia (1995).
  • Jahn T.L., Boove E.C. and Jahn F.F., How to Know the Protozoa, The Pictured Key Nature Series, Wm. C. Brown Company Publishers, 2nd edition, Dubuque, Iowa (1979).
  • Madoni P., Estimates of Ciliated Protozoa Biomass in Activated Sludge and Biofilm, Bioresource Technology, 48 , 245-249 (1994a).
  • Madoni P., La Microfauna nellanalisi di Qualita Biologica dei Fanghi Attivi. Manuale di Applicazione, Azienda Gas Acqua Consorziale di Reggio Emilia, Universtitŕ degli Studi di Parma (1994b).
  • Martin-Cereceda M., Serrano S. and Guinea A., A Comparative Study of Ciliated Protozoa Communities in Activated-Sludge Plants, FEMS Microb. Ecol., 21, 267-276 (1996).
  • Nicolau A., Lima N., Mota N. and Madoni P., Os Protozoários como Indicadores da Qualidade das Lamas Activadas, Boletim de Biotecnologia, 56, 14-19 (1997).
  • Richard M., Activated Sludge Microbiology, The Water Pollution Control Federation, 2nd edition, Virginia (1991).
  • Russ J.C., The Image Processing Handbook, 2nd edition, CRC Press, Boca Raton (1991).
  • Sasahara T. and Ogawa T., Treatment of Brewery Effluent. Part VIII : Protozoa and Metazoa found in the Activated Sludge Process for Brewery Effluent, Monatsschrift für Brauwissenschaft, 11, 443-448 (1983).

  • *
    To whom correspondence should be addressed

Publication Dates

  • Publication in this collection
    25 May 2001
  • Date of issue
    Mar 2001

History

  • Accepted
    15 Jan 2001
  • Received
    05 Oct 2000
Brazilian Society of Chemical Engineering Rua Líbero Badaró, 152 , 11. and., 01008-903 São Paulo SP Brazil, Tel.: +55 11 3107-8747, Fax.: +55 11 3104-4649, Fax: +55 11 3104-4649 - São Paulo - SP - Brazil
E-mail: rgiudici@usp.br