An automated method to detect and quantify fungiform papillae in the human tongue: Validation and relationship to phenotypical differences in taste perception

Abstract

Determination of the number of fungiform papillae (FP) on the human tongue is an important measure that has frequently been associated with individual differences in oral perception, including taste sensitivity. At present, there is no standardised method consistently used to identify the number of FP, and primarily scientists manually count papillae over a small region(s) of the anterior tip of a stained tongue. In this study, a rapid automated method was developed to quantify the number of FP across the anterior 2cm of an unstained tongue from high resolution digital images. In 60 participants, the automated method was validated against traditional manual counting, and then used to assess the relationship between the number of FP and taste phenotype (both 6-n-propylthiouracil (PROP) and Thermal Taster Status). FP count on the anterior 2cm of the tongue was found to correlate significantly with PROP taster status. PROP supertasters (PSTs) had a significantly higher FP count compared with PROP non-tasters (PNTs). Conversely, the common approach used to determine the number of FP in a small 6mm diameter circle on the anterior tongue tip, did not show a significant correlation irrespective of whether it was determined via automated or manual counting. The regional distribution of FP was assessed across PROP taster status groups. PSTs had a significantly higher FP count within the first centimetre of the anterior tongue compared with the PNT and PROP medium-tasters (PMT), with no significant difference in the second centimetre. No significant relationship was found with Thermal Taster Status and FP count, or interaction with PROP taster status groups, supporting previous evidence suggesting these phenomena are independent. The automated method is a valuable tool, enabling reliable quantification of FP over the anterior 2cm surface of the tongue, and overcomes subjective discrepancies in manual counting.

Keywords: Automated counting; Colour segmentation; Fungiform papillae; PROP; Thermal taster status; Tongue images.

Fig. 1

a) Standardised positioning of the tongue for image capturing, with ruler used to scale images, b) An example of manual selection of the three regions of interest (ROIs) - ‘fungiform papillae’, ‘filiform papillae’, and ‘tongue base’, c) Algorithm segmentation of the tongue into ‘fungiform papillae’, ‘filiform papillae’, and ‘tongue base’- based on colour learning.

Fig. 2

a) A significant correlation between manual fungiform papillae count within a 6 mm circle for the dyed tongue versus undyed tongue, r = 0.87, p < 0.001. Significant correlations between manual and automated counts for b) the 6 mm circle (r = 0.86, p < 0.001) and c) the whole anterior 2 cm region of the tongue (r = 0.72, p < 0.001). d) Correlation coefficients (r) between the automated and manual count of papillae in each grid across the anterior 2 cm of the tongue. Grids are numbered from 1 to 4 in Band 1, and from 5 to 8 in Band 2 for the left (L) and right (R) side of the tongue.

Fig. 3

Distribution of fungiform papillae computed using the automated method in a) Band 1 and Band 2, and b) each grid. Error bars indicate standard error (SE).

Fig. 4

a) Bar chart of the manual and automated fungiform papillae count in anterior 2 cm of the tongue plot for each PROP group. The fungiform papillae count was significantly higher in PSTs compared to PNTs for the manual (p = 0.006) and automated (p = 0.019) count, as assessed by post-hoc Tukey's analysis. b) The correlation of bitterness rating of PROP with the automated fungiform papillae count on the anterior 2 cm of the tongue. c) Automated fungiform papillae count on the tongue in Band 1 and Band 2 across PROP groups. d) Automated fungiform papillae count on the tongue in each grid across PROP groups. Error bars indicate standard error (SE) and * indicate significance difference.