Clin Oral Invest (2015) 19:429–436 DOI 10.1007/s00784-014-1266-6
ORIGINAL ARTICLE
Particle size distributions determined by optical scanning and by sieving in the assessment of masticatory performance of complete denture wearers Lydia Eberhard & Sophie Schneider & Constantin Eiffler & Stefanie Kappel & Nikolaos Nikitas Giannakopoulos
Received: 16 October 2013 / Accepted: 19 May 2014 / Published online: 1 August 2014 # Springer-Verlag Berlin Heidelberg 2014
Abstract Objectives Standard procedure for the measurement of masticatory performance is the fractionated sieving of fragmented test food, which is substantially time consuming. The aim of this study was to introduce a less laborious, comparable, and valid technique based on scanning. Methods Fifty-six Optocal chewing samples were minced by wearers of complete dentures with 15 and 40 chewing strokes and analyzed by both a sieving and a scanning method. The sieving procedure was carried out using ten sieves (5.6, 4.0, 2.8, 2.0, 1.4, 1.0, 0.71, 0.5, 0.355, and 0.25 mm) and measuring the weight of the specific fractions. Scanning was performed with a flatbed scanner (Epson Expression1600Pro, Seiko Epson Corporation, Japan, 1,200 dpi). Scanned images underwent image analysis (ImageJ 1.42q, NIH, USA), which yielded descriptive parameters for each particle. Out of the 2D image, a volume was estimated for each particle. In order to receive a discrete particle size distribution, area–volumeconversion factors were determined. The cumulated weights yielded by either method were curve fitted with the Rosin– Rammler distribution (MATLAB, The MathWorks, Inc., Natick, USA) to determine the median particle size X50. Results The Rosin–Rammler distributions for sieving and scanning resembled each other and showed an excellent correlation in 15/40 chewing strokes (r=0.995/r=0.971, P62; ≤647
0.5 0.71 1.0 1.4 2.0 2.8 4.0 5.6
>647; ≤1,694 >1,694; ≤3,251 >3,251; ≤7,366 >7,366; ≤13,196 >13,196; ≤23,652 >23,652; ≤42,975 >42,975; ≤86,582 >86,582
432
a) Sieve aperture (edge length) (mm)
Number of particles
0.25 0.355 0.5 0.71 1.0 1.4 2.0 2.8 4.0 5.6
121 204 145 145 107 43 59 53 105 4
Sum
986
Pixel
Table 2 Description of the “counted sample,” which consisted of primarily sieved particles. Four to 204 particles were taken from the overweight sieve fractions and used as a reference
Clin Oral Invest (2015) 19:429–436
Sieve Number (decreasing aperture), Imaging parameters „Minor“ (le) and „MinFeret“ (right)
differentiates well between sieves 1 to 4, but hardly among sieves 5 to 9. The second discriminating function makes only a small additional contribution to discrimination. It is, furthermore, unable to distinguish between sieves 2 and 9, 3 and 6 as well as 4 and 5. As a result, many particles were misclassified from the fifth or seventh into the sixth, or from the eighth into the ninth sieve. Deficiencies in discrimination, especially with large particles, substantially compromised the particle size distribution. Statistical approach b showed that the variables “area,” “minor,” and “minFeret” had the best discriminatory power (Fig. 3). Threshold values were determined by taking the 5 % percentile of the area values for the particles from the upper sieve and the respective 95 % percentile for particles from the lower sieve and calculating the mean of both, because there was usually an overlap (Fig. 3).
Pixel
b)
Sieve Number (decreasing aperture), Imaging parameter „Area“
Fig. 3 Box plots of three scanning properties: a minor axis of the bestfitting ellipse and minimum Feret’s diameter; b area for the categories sieve 1 to sieve 9 (aperture 5.6 to 0.355 mm, respectively). Pixel = unit of the original scanned image
In a second step, the third dimension of each particle was estimated and corrected with the factors listed in Table 3.
Table 3 Correction factors for the calculation of weights of scanned particles
Fig. 2 Superimposed values of the first two discriminating functions for the sieve fractions from 5.6 to 0.355 mm sieve aperture. The range of the functions is arbitrary
Sieve aperture (edge length) (mm)
Factor
0.25
10.1632
0.355 0.5 0.71 1.0 1.4 2.0 2.8 4.0 5.6
0.8380 0.6496 0.7109 0.6812 0.8036 0.7542 0.7327 1.4592 1.8025
Clin Oral Invest (2015) 19:429–436
Statistical analysis The particle size distributions found by the two different methods were compared by Pearson correlation. As the correlation coefficient merely describes the degree of interconnection of values but does not necessitate their being identical, a Bland–Altman plot was created [17]. All procedures were conducted with SPSS 18.0 (SPSS Inc., Chicago).
Results Particle size distribution The Rosin–Rammler distributions obtained from scanning and sieving resembled each other (Fig. 4a) in both experiments with 15 or 40 chewing strokes, respectively. There was reasonable agreement between the weights of sieve fractions for sieve apertures from 4.0 to 1.0 mm. The difference was greater for smaller and larger particles. The variation in the 5.6-mm sieve fraction can be attributed to two groups: one shows a small positive difference, indicating slight overestimation of the scanned weight; a second one shows a greater negative difference, indicating marked underestimation of the scanned weight. This reflects a threshold effect: if present, particles on the 5.6-mm sieve have a substantial mass of approximately 0.2 g; often, however, no particles were retained by that sieve.
433
The values of X50 for 40 chewing strokes varied between 2.39 and 5.43 for sieving and between 3.04 and 5.28 for scanning. Thus, scanning overestimated X50 values on average by 2.4 % (Table 4). The agreement of the weight distributions derived from scanning and sieving was excellent for both 15 chewing strokes (0.995, N=22, P
ORIGINAL ARTICLE
Particle size distributions determined by optical scanning and by sieving in the assessment of masticatory performance of complete denture wearers Lydia Eberhard & Sophie Schneider & Constantin Eiffler & Stefanie Kappel & Nikolaos Nikitas Giannakopoulos
Received: 16 October 2013 / Accepted: 19 May 2014 / Published online: 1 August 2014 # Springer-Verlag Berlin Heidelberg 2014
Abstract Objectives Standard procedure for the measurement of masticatory performance is the fractionated sieving of fragmented test food, which is substantially time consuming. The aim of this study was to introduce a less laborious, comparable, and valid technique based on scanning. Methods Fifty-six Optocal chewing samples were minced by wearers of complete dentures with 15 and 40 chewing strokes and analyzed by both a sieving and a scanning method. The sieving procedure was carried out using ten sieves (5.6, 4.0, 2.8, 2.0, 1.4, 1.0, 0.71, 0.5, 0.355, and 0.25 mm) and measuring the weight of the specific fractions. Scanning was performed with a flatbed scanner (Epson Expression1600Pro, Seiko Epson Corporation, Japan, 1,200 dpi). Scanned images underwent image analysis (ImageJ 1.42q, NIH, USA), which yielded descriptive parameters for each particle. Out of the 2D image, a volume was estimated for each particle. In order to receive a discrete particle size distribution, area–volumeconversion factors were determined. The cumulated weights yielded by either method were curve fitted with the Rosin– Rammler distribution (MATLAB, The MathWorks, Inc., Natick, USA) to determine the median particle size X50. Results The Rosin–Rammler distributions for sieving and scanning resembled each other and showed an excellent correlation in 15/40 chewing strokes (r=0.995/r=0.971, P62; ≤647
0.5 0.71 1.0 1.4 2.0 2.8 4.0 5.6
>647; ≤1,694 >1,694; ≤3,251 >3,251; ≤7,366 >7,366; ≤13,196 >13,196; ≤23,652 >23,652; ≤42,975 >42,975; ≤86,582 >86,582
432
a) Sieve aperture (edge length) (mm)
Number of particles
0.25 0.355 0.5 0.71 1.0 1.4 2.0 2.8 4.0 5.6
121 204 145 145 107 43 59 53 105 4
Sum
986
Pixel
Table 2 Description of the “counted sample,” which consisted of primarily sieved particles. Four to 204 particles were taken from the overweight sieve fractions and used as a reference
Clin Oral Invest (2015) 19:429–436
Sieve Number (decreasing aperture), Imaging parameters „Minor“ (le) and „MinFeret“ (right)
differentiates well between sieves 1 to 4, but hardly among sieves 5 to 9. The second discriminating function makes only a small additional contribution to discrimination. It is, furthermore, unable to distinguish between sieves 2 and 9, 3 and 6 as well as 4 and 5. As a result, many particles were misclassified from the fifth or seventh into the sixth, or from the eighth into the ninth sieve. Deficiencies in discrimination, especially with large particles, substantially compromised the particle size distribution. Statistical approach b showed that the variables “area,” “minor,” and “minFeret” had the best discriminatory power (Fig. 3). Threshold values were determined by taking the 5 % percentile of the area values for the particles from the upper sieve and the respective 95 % percentile for particles from the lower sieve and calculating the mean of both, because there was usually an overlap (Fig. 3).
Pixel
b)
Sieve Number (decreasing aperture), Imaging parameter „Area“
Fig. 3 Box plots of three scanning properties: a minor axis of the bestfitting ellipse and minimum Feret’s diameter; b area for the categories sieve 1 to sieve 9 (aperture 5.6 to 0.355 mm, respectively). Pixel = unit of the original scanned image
In a second step, the third dimension of each particle was estimated and corrected with the factors listed in Table 3.
Table 3 Correction factors for the calculation of weights of scanned particles
Fig. 2 Superimposed values of the first two discriminating functions for the sieve fractions from 5.6 to 0.355 mm sieve aperture. The range of the functions is arbitrary
Sieve aperture (edge length) (mm)
Factor
0.25
10.1632
0.355 0.5 0.71 1.0 1.4 2.0 2.8 4.0 5.6
0.8380 0.6496 0.7109 0.6812 0.8036 0.7542 0.7327 1.4592 1.8025
Clin Oral Invest (2015) 19:429–436
Statistical analysis The particle size distributions found by the two different methods were compared by Pearson correlation. As the correlation coefficient merely describes the degree of interconnection of values but does not necessitate their being identical, a Bland–Altman plot was created [17]. All procedures were conducted with SPSS 18.0 (SPSS Inc., Chicago).
Results Particle size distribution The Rosin–Rammler distributions obtained from scanning and sieving resembled each other (Fig. 4a) in both experiments with 15 or 40 chewing strokes, respectively. There was reasonable agreement between the weights of sieve fractions for sieve apertures from 4.0 to 1.0 mm. The difference was greater for smaller and larger particles. The variation in the 5.6-mm sieve fraction can be attributed to two groups: one shows a small positive difference, indicating slight overestimation of the scanned weight; a second one shows a greater negative difference, indicating marked underestimation of the scanned weight. This reflects a threshold effect: if present, particles on the 5.6-mm sieve have a substantial mass of approximately 0.2 g; often, however, no particles were retained by that sieve.
433
The values of X50 for 40 chewing strokes varied between 2.39 and 5.43 for sieving and between 3.04 and 5.28 for scanning. Thus, scanning overestimated X50 values on average by 2.4 % (Table 4). The agreement of the weight distributions derived from scanning and sieving was excellent for both 15 chewing strokes (0.995, N=22, P
