Review
Application of image analysis techniques in activated sludge wastewater treatment processes
Ewa Liwarska-Bizukojc
Department of Environmental Engineering, Technical University of Lodz, Al. Politechniki 6, 90-924, Lodz, Poland (Fax: +48-42-6313517; E-mail: ewaliwar@p.lodz.pl)
Received 20 May 2005; Revisions requested 5 June 2005; Revisions received 19 July 2005; Accepted 19 July 2005
Key words: activated sludge, flocs morphology, image analysis, wastewater treatment
Abstract
Image analytical techniques have been extensively developed to evaluate complex microbial aggregates such as sludge flocs and biofilms. This review covers the latest contributions concerning the application of image analysis to the activated sludge systems with respect to the most frequently used morphological parameters and relations between them and traditional wastewater treatment parameters. Recent developments have indicated that image analysis can be successfully used for the quantification of flocs and filamentous bacteria in the operating wastewater treatment plants, which enables prediction of bulking events and pinpoint flocs formation.
Introduction
The characteristics of activated sludge flocs are important not only for researchers but also in the everyday operation of many wastewater treatment plants (Li & Ganczarczyk 1990, Grijs- peerdt & Verstraete 1997, Contreras et al. 2004).
There are two major reasons behind this. First of all, the activated sludge process is one of the most often used methods for microbiological degradation of the contaminants present in wastewater. Secondly, microbial aggregates, de- noted as flocs, are undoubtedly the major com- ponent of this wastewater purification system.
They determine the quality of effluent and the overall efficiency of wastewater treatment plants.
On the one hand, the activated sludge flocs influence the efficiency of wastewater treatment processes due to the impact on substrate transfer, sludge recirculation and separation processes. On the other hand, the internal structure and mor- phology of flocs depend on many factors, such as substrate composition, operational conditions
and the type of the aerated tank. These correla- tions are schematically shown in Figure 1.
The investigations of activated sludge flocs usually concern the following issues: (1) mor- phology, i.e. the size and shape of flocs; (2) com- position of flocs, i.e. exploration of their internal structure, for example distribution of microbes;
(3) the identification of microbial species; (4) spa- tial arrangement of microorganisms.
Recently, a great variety of analytical and microscopic techniques have been developed to analyse activated sludge flocs and biofilms. In this study, several latest contributions concerning the morphology of activated sludge flocs are dis- cussed. A closer look is taken at the definition and description of the most common morpholog- ical parameters used to quantify the flocs and filamentous bacteria.
Image analysis of flocs
Image analysis has been widely used to quantita- tively describe many different biological
Biotechnology Letters (2005) 27: 1427–1433 Ó Springer 2005
DOI 10.1007/s10529-005-1303-2
processes with both suspended and immobilised cultures (Pons & Vivier 1999). Initially, image analysis was applied to characterise the morphol- ogy of filamentous species such as fungi and filamentous bacteria. The application of auto- mated image analysis procedures has subse- quently been extended to mixed cultures of aerobic and anaerobic sludges (Li & Ganc- zarczyk 1990, Grijspeerdt & Verstraete 1997, Al- ves et al. 2000, da Motta et al. 2001a,b).
Li & Ganczarczyk (1990) analysed stained microtome sections of flocs to determine several morphological parameters with the help of image analysis. Some authors have used the image anal- ysis of activated sludge, at high magnifications, to detect filamentous bacteria in the sludge (Watanabe et al. 1990). Later, low magnification microscopy (50 or 100) of unstained or fixed slides combined with image analysis became more common to quantify the size and shape of activated sludge flocs (Grijspeerdt & Verstraete 1997, Alves et al. 2000, da Motta et al. 2001a, b).
In this case, sample preparation is a simple and non-laborious task. Additionally, the application of automated procedures makes the measurement
more objective and reproducible, especially in comparison to the traditional microscopic obser- vations. The automated image analysis proce- dures aim at quantification of the size and shape of activated sludge flocs, however, they do not allow for a detailed identification of the bacterial species and a visualisation of filamentous bacte- ria inside the flocs.
There are several examples of commercial im- age analysis software packets which are usually offered by the companies that deliver micro- scopes and imaging systems. Additionally, image analysis programmes are elaborated by groups of scientists and available as a public domain. Here, an example is ImageJ 1.33 (http://rsb.info.- nih.gov/ij/) elaborated by the Research Services Branch of National Institutes of Health (USA) and DAIME (Digital Image Analysis in Micro- bial Ecology), created in the Department of Microbial Ecology of Vienna Ecology Centre (University of Vienna). DAIME is a novel com- puter programme that integrates digital image analysis and 3-D visualisation functions. It can analyse the digital images from epifluorescence microscopes and confocal image stacks (CLSM)
Fig. 1.Correlations between operational conditions, flocs and effluent quality in the activated sludge wastewater treatment system.
(Daims et al. 2005). It enables exploration of the spatial arrangement of microorganisms, for example in biofilms and flocs (www.microbial- ecology.net/daime/).
An image analysis procedure can be divided into four steps: (1) sample and slide preparation, (2) imaging and grabbing, (3) image processing, (4) image analysis, i.e. measurement of morpho- logical parameters (Russ 1990, Pons & Vivier 1999). Within the first stage of image analysis procedure a suitable slide, either stained or un- stained, vital or fixed, should be prepared. In the next step, an image is obtained with the help of an optical, fluorescence or confocal laser scan- ning microscope (Lopez et al. 2005). Later, the images are taken by means of CCD cameras and saved on magnetic or optical data carriers with the use of a relevant software. Image processing is a set of operations which are performed to transform an image in order to enable the mea- surement of the observed objects. Image process- ing also improves the quality of an image by reducing noises, enhancing objects and detecting their edges. Basic tools of the image processing are point operations and filtration in the fre- quency and space domain by means of linear and non-linear filters. The processed image is then subjected to segmentation and as a result a bin- ary image is obtained. Finally, the size of objects and other morphological parameters are mea- sured. Figure 2 shows the flowsheet of processing and analysis of a single activated sludge floc.
In order to perform the image analysis proce- dure, a sufficient number of images should be taken. Grijspeerdt & Verstraete (1997) found that at least 150 objects, which correspond to about 10 images, should be analysed in order to obtain the statistically relevant results. da Motta et al. (2001) considered, however, that the num- ber of examined images should not exceed 70 grabbed images because this number is adequate to provide stable results. Liwarska-Bizukojc &
Bizukojc (2005) confirmed that the analysis of 40 images was sufficient to obtain statistically rele- vant results.
Generally, the image analysis is used for quantification of the morphological parameters of sludge flocs. However, the image analysis proce- dures combined with molecular biology methods such as fluorescent in situ hybridization (FISH) or denaturating gradient gel electrophoresis
(DGGE) allow us to identify the key microorgan- isms as well as to quantify the number of cells.
Image analysis, in combination with FISH and environmental 16S rRNA libraries, have contrib- uted to the discovery and understanding of the processes and microorganisms pertinent to bio- logical wastewater treatment. It applies, for example, to the anaerobic oxidation of ammo- nium, polyphosphate accumulating organisms (PAO) responsible for biological phosphorus re- moval or various nitrification organisms such as Nitrospira sp. (Wilderer et al. 2002). An online database for rRNA-targeted oligonucleotide probes elaborated by Loy et al. (2003) is accessi- ble in the Internet (www.microbial-ecology.net/
probeBase). Further information and an overview of novel molecular biology methods currently available to investigate microbial aggregates is provided by Wilderer et al. (2002).
Structure and composition of flocs
Sludge flocs consist mainly of organic matter, which makes about 70% of their dry weight (Hartmann 1992). From the biological point of view, activated sludge suspension is a complex ecosystem, which consists mainly of bacteria and protozoa (da Motta et al. 2001b). Generally, the four main constituents can be discriminated in flocs: microorganisms (viable and dead cells), extracellular polymeric substances (mainly carbo- hydrates and proteins), water and inorganic par- ticles (sand). The structure of a typical activated sludge floc is shown in Figure 3.
Li & Ganczarczyk (1990) investigated in detail the composition and internal structure of activated sludge flocs using microtome section- ing, staining and image analysis procedures.
Microorganisms, water and extracellular poly- meric substances (EPS) were irregularly dispersed within the floc although the cross-sectional mor- phology of the flocs appeared similar. Therefore, flocs could be characterised by the fractal con- cept within a certain size limit. It was also con- firmed that the large amount of extracellular polymeric substances were present within flocs.
Extracellular polymeric substances acted on sub- strates and products transferred to and from the microbial cells in the flocs. Substances to be transferred have to overcome not only the 1429
diffusional resistance of water but also of the EPS, which surround most of the microbial cells.
The EPS matrix is very heterogeneous and con- sists of a variety of polymeric compounds as car- bohydrates, proteins, lipids and nucleic acids.
However, the dominant components are carbo- hydrates and proteins. The presence and compo- sition of EPS deliver data on activated sludge age and organic loading rates. The total amount of EPS decreases at higher organic loading rates.
A large amount of EPS is excreted at low growth stages of microorganisms due to cell autolysis.
Also the composition of EPS, especially the pro- tein to carbohydrates ratio depends on sludge retention time and organic loading. Carbohy- drates are synthesised extracellularly for specific functions, while proteins usually come from the
excretion of intracellular polymers or cell lysis (Lee et al. 2003).
The internal floc structure is also explored with the help of the most modern microscopic tech- niques. Lopez et al. (2005) evaluated different microscopic techniques from epifluorescence microscopy to confocal laser scanning microscopy (CLSM) and two photon excitation laser scanning microscopy (TPE-LSM) combined with fluores- cent stains to investigate complex microbial aggregates such as activated sludge flocs (Lopez et al. 2005). The applicability limits of these microscopic techniques were estimated by analy- sing activated sludge samples taken from three different sources after staining with a fluorescent viability indicator. The selection of an appropri- ate microscopic technique depended on the type
Fig. 2. Image analysis flowsheet for activated sludge flocs.
1430
of microbial aggregates analysed. Generally, epi- fluorescence and CLSM proved to be sufficient to analyse the aggregates. However, for flocs of high cell density, only TPE-LSM images revealed the internal structure of flocs.
Morphological parameters of flocs
The morphological parameters of activated sludge flocs, which are obtained on the basis of the automated analysis of microscopic images, can be divided into two groups. The first group covers parameters representing the size of flocs.
These are mean projected area, diameter, perime- ter and equivalent circle diameter. Mean pro- jected area (A) is the basic image analysis parameter and is found easily by pixel count and its multiplication by scaling factor. The other size parameters are the derivatives of mean projected area. For example, equivalent circle diameter (De), which was used by Grijspeerdt & Verstraete (1997) and da Motta et al. (2001a, b), is defined as:
De¼ 2 ffiffiffiffiffiffiffiffiffiffiffi Area p r
Taking the size of flocs into account, the three groups of flocs can be distinguished: small flocs (diameter below 100 lm); mean flocs (diameter
between 100 and 500 lm) and large flocs (diame- ter above 500 lm) (Eikelboom & van Buijsen 1992). Additionally, another commonly used parameter, which describes the size and amount (concentration) of flocs, is the field area (FA).
This is the ratio of the area occupied by the flocs detected in an image to the total image area (Russ 1990).
The second group of morphological parame- ters describe the shape of flocs, mainly with re- spect to their circularity and regularity. Here, the most often used parameters are roundness (RD) or circularity index (Cx). They both indicate to what an extent the measured floc is similar to the true circle. Roundness varies from 0 to 1 and a circle has a RD of one. The circularity index is also equal to one when the object is a circle, however it increases when the object becomes less circular. This is caused by different expres- sions of these shape factors. Roundness is calcu- lated from the projected area:
RD¼ 4 Area p Length
Circularity index (Cx) is the ratio of the observed perimeter (P) to the perimeter of a circle of the same area as the measured one:
Cx ¼ Perimeter 2 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
p Area p
A next important and popular shape factor is the fractal dimension, which is a measure of the irregularity of the perimeter. The fractal dimen- sion is usually determined with the ‘‘mosaic amalgamation’’ algorithm described by Russ (1990). da Motta et al. (2001b) reported that the fractal dimension decreased before bulking had started.
Image analysis plays an important role in the quantifications of filamentous bacteria. Generally, the amount of filamentous bacteria is estimated by manual counting under a microscope, which is laborious, time-consuming and can be subjective.
da Motta et al. (2001b) developed a general image analysis procedure to evaluate the amount of fila- mentous bacteria from the optical microscopy images. A key parameter expressing the number of filamentous bacteria within this procedure was total filament length per image (Lf). Additionally, a number of filaments per image and surface ratio of filaments per flocs were measured. The total
Fig. 3.The composition of a typical activated sludge floc, im- age taken at magnification of 1000 with phase contrast.
1431
filament length (Lf) is the sum of filament length of all the individuals present in the image (da Motta et al. 2001a). da Motta et al. (2001a, b) re- ported that an increase of sludge volume index was accompanied by the increase of filament length per image in pilot-scale experiments as well as in a large municipal wastewater treatment plant. The obtained results indicated that image analysis could be applied to detect bulking events in the wastewater treatment plant.
Quantification of filamentous bacteria
The problems of separating solids from the trea- ted effluent still occur in many wastewater treat- ment plants (Contreras et al. 2004). The majority of settling failures can be attributed to the bul- king of sludge caused by the excessive growth of filamentous microorganisms. Misbalance between the filamentous bacteria and flocs-forming bacte- ria induces bulking and foaming problems and, as a result, bad quality of the treated effluent.
Therefore, a systematic quantification of filamen- tous bacteria with the use of automated image analysis procedures can be a useful tool for the classification of bulking events.
Contreras et al. (2004) used image analysis to quantify the fractions of filamentous microorgan- isms (FM) and non-filamentous microorganisms (NFM) in the activated sludge flocs. In order to classify the particles as filamentous or not fila- mentous two shape parameters were used: round- ness (RD) and reduced gyration radius (Rg).
Roundness is equal to 1 for a circle and decreases if the object becomes more elongated.
The second parameter was reduced gyration ra- dius (Rg). Rg is equal to 0.707 for a circle and the more elongated the object is, the higher it becomes. Contreras et al. (2004) gave a detailed definition of Rg, which is as follows.
Rg¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Mx2þ My2
p ffiffiffiffiffiffiffiffiffi pN=p
where: N is the object area (pixels); Mx2and My2
are the central second moments with respect to x-axis and y-axis, respectively.
The calibration of image analysis procedures was based on pure cultures of Sphaerotilus natans (FM) and strain E932 (NFB). The distributions
of Ro and Rg corresponded very closely both to NFB and FM. For example, area-based histo- grams showed that for NFB roundness and re- duced gyration radius reached maximum at Ro=1 and Rg=0.75, respectively. In the case of FM, the position of the maximum values in the distribution corresponded to Ro=0.16 and Rg= 1.5. It means that both proposed morphological parameters can be applied to quantify filamen- tous microorganisms (Contreras et al. 2004).
Moreover, it occurred that at low dilution rates FM dominated, whereas at high dilution rates a rapid growth of NFB was observed. It should be added that the experiments were conducted in a continuous reactor, where both microorganisms competed for a single substrate.
Relation between image analysis and traditional wastewater treatment parameters
The simplest and most often used methods for the biomass determination in the process of activated sludge wastewater treatment are volatile sus- pended solids (VSS) or chemical oxygen demand (COD). In spite of their simplicity, these methods do not distinguish between living cells, dead bio- mass and nonviable organic particles. According to the literature data, viable biomass in the typical activated sludge makes only 5 to 20% of VSS on average (Jørgensen et al. 1992, Liwarska-Bizukojc
& Ledakowicz 2003). Moreover, these simplest methods do not allow for the estimation of the quantity of flocs formed or filamentous bacteria present in the system. Therefore, new analytical techniques, including image analysis, are still being developed for the determination of biomass physiological state or morphology.
Literature data indicate that some image analy- sis parameters can also be good biomass indica- tors. Moreover, a relation between image analysis parameters and standard parameters of wastewa- ter treatment were established. Grijspeerdt & Ver- straete (1997) found a linear correlation between the field area and activated sludge concentration.
This linear correlation was not valid when the activated sludge was too concentrated, above 4 g l)1. Liwarska-Bizukojc & Bizukojc (2005) con- firmed this linear relationship between total sus- pended solids and the field area. Additionally, the linear relation between the mean projected area of
flocs and total suspended solids in lab-scale batch experiments was established (Liwarska-Bizukojc &
Bizukojc, 2005). Activated sludge concentration within the batch experiments was below 2 g l)1.
The settleability of sludge depends greatly on the formation of compact flocs. Ideally, sludge flocs should be firm and spherical to achieve the best settling properties. Therefore, finding a correlation between the morphological parame- ters of flocs and settleability indexes would help significantly in the everyday operation of a wastewater treatment plant. Grijspeerdt &
Verstraete (1997) indicated that it was possible to join the morphology of sludge flocs with set- tling properties. However, da Motta et al.
(2001a) reported that a global and direct rela- tion between sludge volume index (SVI) and morphological parameters such as average floc diameter and roundness within full-scale acti- vated sludge monitoring could not be found.
Nevertheless, an increase of SVI was accompa- nied by the increase of filament length per image, especially in the pilot-scale experiments (da Motta et al. 2001a, b).
Summary
Microscopic techniques ranging from optical microscopy to confocal laser scanning micros- copy can help to evaluate the morphology, internal structure, identification and spatial arrangement of microorganisms in microbial aggregates such as sludge flocs and biofilms.
Low magnification light microscopy for vital un- stained or fixed slides is sufficient to quantify the size and shape of activated sludge flocs. The same image analysis procedures could be applied to analyse both municipal and industrial acti- vated sludge. The measurement of morphologi- cal parameters of flocs and filamentous bacteria supply valuable data for everyday operation of wastewater treatment plants. These data help to detect the bulking events and pinpoint flocs for- mation. Simultaneously, they enable estimation of biomass concentration. Therefore, image analysis could be incorporated into control sys- tems for monitoring of a wastewater treatment plant.
References
Alves M, Cavaleiro AJ, Ferreira EC, Amaral L, Mota M, da Motta M, Vivier H, Pons MN (2000) Characterisation by image analysis an anaerobic sludge under shock conditions.
Water Sci. Technol. 14: 207–214.
Contreras EM, Gianuzzi L, Zaritzky NE (2004) Use of image analysis in the study of competition between filamentous and non-filamentous bacteria. Water Res. 38: 2621–2630.
Daims H, Lucker S, Wagner M (2005) Daime, a novel image analysis program for microbial ecology and biofilm research.
Environ. Microbial. in press.
da Motta N, Pons MN, Roche N, Vivier H (2001a) Charac- terisation of activated sludge by automated image analysis.
Biochem. Eng. J. 9: 165–173.
da Motta M, Pons MN, Roche N (2001b) Automated moni- toring of activated sludge in a pilot plant using image analysis. Water Sci. Technol. 43/7: 91–96.
Eikelboom DH, van Buijsen HJJ (1992) Handbuch fu¨r die mikrobiologische Schlammuntersuchung, 3rd edn.Munich: F Hirthammer Verlag GmbH.
Grijspeerdt K, Verstraete W (1997) Image analysis to estimate the settleability and concentration of activated sludge. Water Res. 31: 1126–34.
Hartmann L (1992) Biologische Abwasserreinigung, 3 Berlin, Heidelberg: Springer Verlag.
Jørgensen PE, Eriksen T, Jensen BK (1992) Estimation of viable biomass in wastewater and activated sludge by determination of ATP, Oxygen Utilization Rate and FDA hydrolysis. Water Res. 26: 1495–1501.
Lee W, Kang S, Shin H (2003) Sludge characteristic and their contribution to microfiltration in submerged membrane bioreactors. J. Membr. Sci. 216: 217–227.
Li DH, Ganczarczyk JJ (1990) Structure of activated sludge flocs. Biotechnol. Bioeng. 35: 57–65.
Liwarska-Bizukojc E, Bizukojc M (2005) Digital image analysis to estimate the influence of sodium dodecyl sulphate on activated sludge flocs. Proc. Biochem. 40: 2067–2072.
Liwarska-Bizukojc E, Ledakowicz S (2003) Estimation of viable biomass in aerobic biodegradation processes of organic fraction of municipal solid waste (MSW). J. Biotechnol. 101:
165–172.
Lopez C, Pons MN, Morgenroth E (2005) Evaluation of microscopic techniques (epifluorescence microscopy, CLSM, TPE-LSM) as a basis for the quantitative image analysis of activated sludge. Water Res. 39: 456–468.
Loy A, Horn M, Wagner M (2003) ProbeBase-an online resource for rRNA-targeted oligonucletide probes. Nuleic Acids Res. 31: 514–516.
Pons MN, Vivier H (1999) Biomass quantification by image analysis. Adv. Biochem. Eng./Biotechnol. 66: 133–184.
Russ JC (1990) Computer Assisted Microscopy: The Measure- ment and Analysis of Images, New York: Plenum Press.
Watanabe S, Baba K, Matuzaki H, Hiraoka M, Tsumura K (1990) Operation condition diagnosis system for activated sludge processes based on microorganism image informa- tion. In: Briggs R, ed. Advances in Water Pollution and Control, London: Pergamon Press, pp. 683–690.
Wilderer PA, Bungartz HJ, Lemmer H, Wagner M, Keller J, Wuertz S (2002) Modern scientific methods and their potential in wastewater science and technology. Water Res.
36: 370–393.
1433
