This research work was supported by the Hanns-Seidl-Stiftung, Munich, Germany (PhD grant to Andrea Czermak) Acknowledgement Conclusion Age at death evaluation by tooth cementum annulation (TCA) – a software for an automated incremental line counting Andrea Czermak 1* , Adrian Czermak 2 , Hartmut Ernst 2 , Gisela Grupe 1 1 LMU Biozentrum, Ludwig Maximilians University Munich, 82152 Planegg-Martinsried, Germany ; 2 Faculty of Computer Sciences, University of Applied Science Rosenheim, Germany Figure 12. Example for an Auto-TCA analysis of incremental lines from individuals of different age groups (determined according to morphological criteria). The software was tested on samples of individuals with unknown age-at-death from historical excavation. The incremental line numbers obtained from one thin- section typically displays a Gaussian distribution. Numbers obtained from different sites (ROIs) of one thin section largely coincide. In comparison, younger individuals tend to show a more narrow Gaussian distribution indicating a more definite result (A). With advanced age the distribution curve tend to be more flattened and the resulting numbers are accordingly more scattered (B). In individuals of advanced age the numbers differ between thin sections from near the tooth crown and thin-sections from the middle of the root (C). With increasing age this difference seems to become more prominent (D), demonstrating that incremental lines are not generated consistently along the tooth root. Thus, the thin-section position can substantially affect the outcome results and has to be taken into consideration (Czermak 2006, 2011). A valid age at death estimation is required in historical and forensic anthropology. Tooth cementum annulation (TCA) is a method for age at death estimation of adult individuals. The method uses light microscopic images acquired from tooth root cross sections. The age is then estimated by counting the number of visible tooth cement incremental lines and adding the result to the assumed age of tooth eruption. Manual line counting, however, is time consuming, potentially subjective and the number of individual counts is insufficient for statistical evaluations. Here a custom-made AutoTCA software is presented that allows automated evaluation of TCA images using Fourier analysis and algorithms for image analysis and pattern. It involves “line-by-line” scanning and the counting of gray scale peaks within a selected region of interest (ROI). Each scanning process of a particular ROI yields up to 400 counts, thus minimizing the potential error induced by manual line counting. This simple and time saving program can substitute manual counting and provides consistent and reproducible and user independent, unbiased results. In either case, however, reliability of the results depends largely on the state of preservation of the analyzed material, the preparation, the choice of the thin section and image quality. These factors have to be standardized to get consistent and reproducible results. Figure 4. Sample preparation. Tooth sample embedded in an epoxide resin bloc (Biodur E12) and fixed in a microtome saw (Leica 1600). The tooth crown of the sample is cut off. Principle of tooth cementum annulation (TCA) References Figure 10. Power spectrum of the rotated ROI. (A) Vertically orientated incremental lines are grouped near the central horizontal line of the power spectrum. In contrast, non-linear structures or linear structures orientated in different angles to the incremental lines in the original image (e.g., saw blade marks) are represented by pixels that are more distant from the horizontal line. To eliminate interfering artifacts a point symmetric angular filter mask was applied (B; !=20°) to result in a masked power spectrum of the rotated ROI (C). Figure 8. Scatterplot after 11x11 Gauss low- pass filtering and DFFT. The ROI is segmented into equally sized rectangles (128 pixels x length of ROI). Each segment is rotated to a horizontal position, using the outer dark line as a reference. Gauss low-pass filter and discrete fast Fourier transformation (DFFT) is applied. The resulting scatter plot indicates a preferred direction (A). The best rotation angle is then extracted by fitting a straight line (B). Single counts Applied data 1 tooth ! 8000 each tooth mean value of all images ! 5 images ! 1600 each image mean value of all ROIs ! 4 ROI each image ! 400 each ROI mean value of the mode of the single counts of a ROI ! 400 Single-counts each ROI Mode of each ROI Figure 5. Bright-field image of a thin-section. Cementum-layer with incremental lines (right side), dentine-layer (left ) (40x objective). Image should be recorded with the incremental lines oriented vertically. Raw image (tif, bmp) is required for further processing. Sample preparation and microscopy Figure 3. Physical sectioning for bright-field imaging. Section are cut at a 90°-angle to the root orientation and show concentric ring structures of alternating bright and dark lines (medium box). Thin-section along the axis of the tooth root are not vertical oriented to the growing line of the cementum and they show less well-defined rings which are shifted and not completely overlapping (lower box) (Maat et al. 2006). Sections in the upper and medium part of the root show the most distinct lines (modified after Maat et al. 2006). Figure 1. Tooth scheme (longitudianal cut). The root surface is surrounded by the tooth cementum which is added in layers on the bone- side of the tooth, comparable with tree-rings. The cementum is interwoven by collagen fibers (“sharpey fibers”) fixing the tooth to the alveolar bone (modified after Schroeder 2001). Figure 2. Correlation of bands, fiber orientation and season. Dark and bright lines, visible in a bright- field microscope presumably correlate with variable orientation and different mineralization of the collagen fibers (assumed orientation of the sharpey fibers in the course of one year). Bright bands appear to develop in winter, dark lines in the summer season (Liebermann 1993, 1994; Stutz 2002). A changeover of the bands happens in March/April and September/October (Wedel 2007). (Figure: Czermak) Figure 9. ROI after rotation correction. Incremental lines are now accurately orientated in vertical direction. The lines are often disrupted by optical artifacts (Fig. 6), dirt, partial decomposition, or by linear kerf marks caused by the saw blade (arrows). These kerf marks are similar to incremental lines and may affect the counting result. Program workflow Quantitative evaluation using Auto-TCA software Figure 7. Region of interest (ROI). User defined polygon is set around the region to be evaluated. The boundary should follow the bright “eruption line” on one side and the dark border to the embedding resin on the other side. Figure 6. Diffraction artifacts. Artifact lines (arrows) are often visible on the interface of preparation and embedding material (magnified from image taken with 40x objective). These may be mistaken for incremental lines, in particular when using lower magnifying objectives (Czermak 2006). (1) Time saving compared to manual counting. (2) The algorithm is more sensitive to intensity differences than the human eye. (3) High number of counts (>1000 per sample) minimizes statistical errors. (4) User-independent, consistent and reproducible results. Abstract Aims of this project (1) Substitute manual line counting (2) Optimize sample preparation and imaging (3) Quantitative evaluation using Auto-TCA software Digital image processing Detection of the ROI Vertical assembly of the “real” incremental lines B A A B C Figure 11. Image processed example ROI and line scan analysis. After Fourier back- transformation of the masked spectrum all interfering structures are eliminated from the image (A). The applied algorithm then generates a line- by-line scan of pixel-by-pixel gray scale values of the entire processes ROI to determine local maxima and minima. Local maxima correspond to the number of incremental lines in this row. These are detected and counted by a “peak-finder” algorithm. An example of one line scan is shown in B. Elimination of interfering structures Vertical orientation of the “real” incremental lines after image processing Table 1. Typical numbers obtained from quantitative data evaluation with Auto-TCA. Each quantification comprises 300-500 single counts (depending on the ROI size). The Auto-TCA software lists the peak numbers of each individual line within the ROI (sorted by the amount of counted peak numbers). The “most often counted number”, the modal value, determined for each ROI is than considered as the most probable value and used for further evaluation. grayscale level width of ROI Evaluation of counting results [1] Czermak, A.; Czermak, A.M.; Ernst, H.; Grupe G. (2006): A New Method for the Automated Age-at-Death Evaluation by Tooth-Cementum Annulation (TCA). Anthopologischer Anzeiger 64 (1), 25-40. [2] Czermak A. (2011): Social Stratification in the Early Middle Ages - Evidence by Demography, Physical Stress and Nutrition. (Soziale Stratifizierung im frühen Mittelalter – Aussage und Nachweismöglichkeiten anhand von biologischen Indikatoren). Dissertation, München [6] Schroeder, H. E. (2001). Orale Strukturbiologie. Stuttgart, Thieme. [7] Stutz, A. J. (2002). "Polarizing Microscopy Identification of Chemical Diagenesis in Archaeological Cementum." Journal of Archaeolgical Science 29: 1327-1347. [8] Wedel, V. L. (2007). "Determination of Season at Death Using Dental Cementum Increment Analysis." Journal of Forensic Science 52(6): 1334-1337. [3] Liebermann, D. E. (1993): Life History Variables Preserved in Dental Cementum Microstructure. Science 261: 1162-1164. [4] Liebermann, D. E. (1994): The Biological Basis for Seasonal Increments in Dental Cementum and their Application to Archaeological Research. Journal of Archaeological Science 21: 525-539. [5] Maat, G. J. R.; Gerretsen, R. R. R., et al. (2006). "Improving the visibility of tooth cementum annulations by adjustment of the cutting angle of microscopic sections." Forensic Science International 159(S): S95-S99. Frühadult 0 5 10 15 20 25 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 Incremental lines Häufigkeit [%] Schnitt3/Zählung1 Schnitt3/Zählung2 Schnitt3/Zählung3 Schnitt4/Zählung1 Schnitt4/Zählung2 Schnitt4/Zählung3 Schnitt5/Zählung1 Schnitt5/Zählung2 Schnitt5/Zählung3 Mitteladult 0 5 10 15 20 25 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 Incremental lines Häufigkeit [%] Schnitt3/Zählung1 Schnitt3/Zählung2 Schnitt3/Zählung3 Schnitt5/Zählung1 Schnitt5/Zählung2 Schnitt5/Zählung3 Schnitt6/Zählung1 Schnitt6/Zählung2 Schnitt6/Zählung3 Spätadult 0 5 10 15 20 25 5 10 15 20 25 30 35 40 45 50 Incremental lines Häufigkeit [%] Schnitt3/Zählung1 Schnitt3/Zählung2 Schnitt3/Zählung3 Schnitt4/Zählung1 Schnitt4/Zählung2 Schnitt4/Zählung3 Schnitt7/Zählung1 Schnitt7/Zählung2 Schnitt7/Zählung3 Schnitt8/Zählung1 Schnitt8/Zählung2 Schnitt8/Zählung3 Mittelmatur 0 5 10 15 20 25 15 20 25 30 35 40 45 50 55 60 Incremental lines Häufigkeit [%] Schnitt2/Zählung1 Schnitt2/Zählung2 Schnitt2/Zählung3 Schnitt3/Zählung1 Schnitt3/Zählung2 Schnitt3/Zählung3 Schnitt8/Zählung1 Schnitt8/Zählung2 Schnitt8/Zählung3 Occurrence Occurrence Occurrence Occurrence slice3/count1 slice4/count1 slice5/count1 slice3/count2 slice4/count2 slice5/count2 slice3/count3 slice4/count3 slice5/count3 slice3/count1 slice5/count1 slice6/count1 slice3/count2 slice5/count2 slice6/count2 slice3/count3 slice5/count3 slice6/count3 slice3/count1 slice4/count1 slice7/count1 slice8/count1 slice3/count2 slice4/count2 slice7/count2 slice8/count2 slice3/count3 slice4/count3 slice7/count3 slice8/count3 slice2/count1 slice3/count1 slice8/count1 slice2/count2 slice3/count2 slice8/count2 slice2/count3 slice3/count3 slice8/count3 early adult (aged 20-24) middle adult (aged 25-31) late adult (aged 32-38) middle mature (aged 46-52) TCA: aged 27 TCA: aged 41 TCA (1): aged 37 TCA (2): aged 40 TCA (1): aged 45 TCA (2): aged 52 czermak_andrea@web.de *Correspondence to: Auto TCA advantages: (1) Good quality microscope image is required (i.e., largest magnification that still allows the cementum layer to fit into the field of view). (2) The software detects bright and dark lines. Each banded structure can be counted (also applicable to annual rings of trees or animal prints). (3) The Auto-TCA software makes the quantitative analysis more accurate and reliable (but not the TCA method itself). Notes: A B C D