4.3.1 Clean and Biofilm image analysis
4.3.1.2 Bulk bio-physical data
The process outlined in steps 6 and 7 of Section 4.2.7 was used to visualize differences between the Clean and Bio scans. Figure 4.19a-d illustrates the results for the middle slice of the ROI of each experiment as an example of the column data.
Figure 4.19. Differentiating image between Clean and Bio scans for (a) BioLightLong (b) BioDarkLong (c) BioLightShort (d) BioDarkShort in which blue and green indicates differences between the before and after scans and white and black indicate areas where the solid and liquid fraction were consistent between the before and after. A red circle indicates an area that does not match up between scans, a purple circle indicates an area where the two scans align well and an orange circle indicates an area where the gravel can be seen as shifting between the two scans
Two main observations are apparent from Figure 4.19: Firstly, in all of the experiments individual grains of gravel seemed to have shifted between the Clean and Bio scans. This is evident when the shape of the grain remains, but one side of the grain shows a green area and the other an equivalent blue area (Fig 4.20a). The size of these coloured regions need not be exactly identical as grain movements are in 3D space and may not occur planar to the image slice.
Analysis indicates that for the majority of movements the displacement is
a b
c d
Chapter 4 MRI of biofouling and metal transport in a gravel filter ___________________________________________________________________________________
122
equivalent to 1 pixel (0.3mm). Whilst this is equivalent to the estimated error possible from manual misalignment of the column in the MRI bore (Section 4.4.5), it is clear that this effect is only present for some, not all, of the grains and varies within slices from 1-5 pixels in displacement length; thus, image acquisition cannot be responsible for these changes. Instead, such movement are likely a result of the unstable grains moving locally; this may be due to:
instability under pore-pressure; vibrations during manual removal of the column from the MRI between scans; or, disturbance due to pressure exerted when connecting the flow system to the column. This problem appears most pronounced on the BioDarkLong (Figure 4.19b) experiment and may explain the unexpected porosity data of Table 4.2 (discussed above).
Figure 4.20. (a) Example of local movement of grain (b) distinct area of green without blue compensation
Secondly, it is clear from Figure 4.19 that partial or entire pore filling (green regions) is prevalent in the wall regions of the chambers. Fig 4.20b shows that this is not compensated locally by pore emptying regions and thus is a process distinct from the aforementioned process of grain movement. This is believed to represent biofilm growth areas responsible for pore clogging. Visual examination of MRI slices (e.g. Figure 4.19) shows that local pore filling is most apparent in the upper, right and lower wall regions of the BioLightLong experiment (Figure 4.19a); here 4 pores appear to have entirely filled (100%) over the 6 month growth period. To a lesser extent the BioLightShort (Figure 4.19c) indicates similar locations of growth (all be it to a lesser extent in that only 1 pore is entirely filled; i.e. partial-filling of pores is more common). This data supports the notion of biological growth filling these pores, as phototrophs have preference for lit conditions near the chamber surface. Interestingly, the Dark experiments also indicate regions of biological pore filling in near-wall regions;
as growth is not ubiquitous around the circumference of the chamber detailed review of Figure 4.19 appears to show that growth in dark conditions is greatest
___________________________________________________________________________________
123
in connected, larger pore areas near the wall. Thus, greater porosity in the near- wall region of the column also appears to promote greater fluid exchange/flow and greater probability of biofilm settlement.
In light of the findings above, analysis was undertaken to attempt to quantify the relative importance of the two processes of grain shifts and biological growth. A basic mathematical approach was utilised in that: (i) the coloured (blue or green) area was determined for each of the 140 slices; (ii) its percentage relative to the total area of the slice was then calculated; (iii) bulk average statistics were then calculated for the volume space. Results are provided in Table 4.3. Since much of the green and blue areas are thought to be local movement of grains between the Clean and Bio scans, subtracting the blue area from the green area may be indicative of biofilm growth in the column.
Table 4.3. Percentage area of pixel analysis illustrating differences between the Clean and Bio scans
From Table 4.3 BioLightLong appears to be the only column with more green than blue; that is, there is an increase in solid region over six months of biological growth when analysed over the bulk volume of the chamber. This is surprising, as earlier data from Figure 4.19 undisputedly highlights regions of biofilm growth in all experiments such that porosity losses would be expected in all data sets of Table 4.3. Thus, three issues are raised: (i) are we correct in assuming that biofilm images as a solid? (ii) have we fully considered issues of gravel movement? (iii) are bulk measurements appropriate to quantifying the
Reported as % Pixels Throughout 140 Slice ROI
Solid/Gravel (Black)
Water/Pore Space (White)
Liquid in Clean Solid in Bio (Green)
Solid in Clean Liquid in Bio (Blue)
% Pore Blockage due to Biofilm
BioLightLong Clean 66 34
BioLightLong Bio 67 33
BioLightLong Added 56 23 11 10 1.6
BioDarkLong Clean 68 32
BioDarkLong Bio 67 33
BioDarkLong Added 55 21 12 12 -1.2
BioLightShort Clean 67 33
BioLightShort Bio 66 34
BioLightShort Added 57 24 8.9 10 -3.7
BioDarkShort Clean 67 33
BioDarkShort Bio 67 33
BioDarkShort Added 58 23 9.4 10 -0.57
Chapter 4 MRI of biofouling and metal transport in a gravel filter ___________________________________________________________________________________
124
impact of local biophysical changes in gravel columns? The first two questions are discussed below, whilst the final question requires separate analysis provided in Section 4.3.1.3.
To address the first of these questions, a look at the original high resolution images is needed to demonstrate a correlation with, and reasoning for use of binary analysis as well as correlation with visual evidence of biofilm growth in the BioLightLong experiment. Since image analysis has been done using binary images thus far and has resulted in what appears to be biofilm growth imaging as a solid rather than a liquid, Figures 4.21a-c show an example (BioLightLong) of the original high resolution Clean (a) and Bio (b) scan images followed by the subtraction of Clean from Bio (c) in which similar trends near the top and right of the column can be visualized. Using the original high resolution scans illustrates that the same differences (little change on the left side of the column, major variations at the top of the column, and what appears to be small
‘fuzzy’ changes on the right of the column) are seen between the Clean and Bio scans without being segmented into binary. What is also apparent is the intensity inhomogeneity differences seen between the Clean and Bio Scans. This inherent side effect of the image acquisition (Vovk et al. 2007) is noticeable where higher degrees of green versus blue appear in the Clean scan over the Bio scan. For this reason, histogram analysis and thresholding of the multitude of colours in the original high resolution scans would be difficult and incomparable and thus, why binary image analysis was chosen as a more appropriate method for comparison and determination of biofilm growth.
___________________________________________________________________________________
125
Figure 4.21. Original high resolution images of slice 76 of BioLightLong (a) Clean (b) Bio (c) Bio subtracted from Clean
Since BioLightLong was the only column that demonstrated a decrease in porosity and increase in green regions, the following points summarizes the observations and highlights further verification of what is believed to be imaged biofilm growth in BioLightLong:
1. It is speculated that biofilm growth would occur near the edge of the column and near the inlet. Growth around the edge would likely occur as this is the closest area of the column to the light source and growth would occur near the inlet as this is where the first entry of the inoculated pond water occurs, thus enhancing colonization in these areas of the column.
2. The areas where the blue and green variations occur in BioLightLong correlate to where the phototrophic biofilm is seen growing in the column. Figure 4.24 shows photos of the biofilm growth of the column on the right side (a) and the left side (b).
Chapter 4 MRI of biofouling and metal transport in a gravel filter ___________________________________________________________________________________
126
Figure 4.22. (a) Photo of the right side of BioLightLong column after 6 month growth period (b) Photo of the left side of BioLightLong column after 6 months growth period.
3. Most of the green variations are seen in the BioLightLong experiment, which is the only experiment where phototrophic biofilm was visually seen to grow as a thick slimy coating. While biofilm may have grown in the other three columns, it was not as visually prominent as BioLightLong.
Thus, for the highlighted reasons above, it does appear that biofilm growth is imaging and thresholding as a solid.
To address the second of the questions casting uncertainty over Table 4.3, the full extent of gravel movement in the columns does appear to be significant. The negative values generated in determining differences between the Clean and Bio scans of BioDarkLong, BioLightShort and BioDarkShort indicate an increase in pore space, even though areas of biofilm growth are clearly seen (in BioDarkLong in particular) and it is not possible for the solid gravel fraction to ‘disappear’.
Thus, the pore space gain is thought to be due to each of the columns gravel settling under gravity to different degrees, as well as porewater pressures experienced during recirculation of pond water and vibrations caused by the MRI process. A change in the original packing arrangement due to the freedom of movement of grains then has the possibility to move and image in different 3D slices, calculating as a ‘loss’ of solid gravel. Therefore, the actual settlement is conceived to be higher than the percentage reported in Table 4.3, in which case percentages of calculated biofilm growth are thought to be an underestimate of actual biofilm growth.
Top of column
Right side of column
Left side of column Most
biofilm growth
Least biofilm growth
Light source:
2nd half of growth period Light source:
1st half of growth period
Top of column
Inlet Inlet
___________________________________________________________________________________
127