Age estimation based on pulp/tooth volume ratio using CBCT: a study on mandibular canines

Abstract

Background

This study aimed to estimate chronological age by evaluating the pulp/tooth volume ratio (PV/TV) in mandibular canine teeth using cone-beam computed tomography (CBCT) images and to assess its correlation with age.

Methods

A total of 240 CBCT scans of individuals aged 17–72 years, with a balanced sex distribution, were analyzed retrospectively. Pulp and tooth volumes were measured using 3D Slicer, and the PV/TV ratio was calculated. Correlations with chronological age were assessed using Pearson’s correlation and linear regression. Model performance was evaluated via mean absolute error (MAE) and root mean square error (RMSE).

Results

The PV/TV ratio showed a strong negative correlation with age (r = − 0.710, p < 0.001). Regression analysis yielded R² values of 0.549 for males, 0.455 for females, and 0.504 for the total sample. The model achieved a prediction error of ± 3.69 years (60% CI) and ± 7.24 years (95% CI), with MAE and RMSE values of 8.41 and 11.19 years, respectively.

Conclusion

The pulp/tooth volume ratio measured from CBCT images can serve as a moderately reliable predictor for dental age estimation. The regression model developed using PV/TV provides an acceptable error margin, particularly within forensic and anthropological contexts.

Introduction

Age determination plays a crucial role in forensic dentistry and finds applications in genetic research, criminal investigations, legal proceedings, and anthropological and archaeological studies [1–3]. Teeth, being the hardest structures in the human body, are capable of withstanding degradation for thousands of years and are less affected by physical and environmental conditions compared to other bodily tissues. This exceptional durability makes teeth highly valuable for age estimation purposes [4–6]. Dental age estimation involves comprehensive evaluation of developmental, morphological, and biochemical changes in teeth. These evaluations not only enable precise age estimation but also offer insights into an individual’s chronological development and physiological changes over time. The resilience of dental tissues to environmental and physical influences makes them ideal for scientific analyses, such as assessing growth patterns, wear, and pathological changes that contribute to understanding aging at both individual and population levels [7–9]. The scope of dental age estimation, which includes both morphological and radiographic assessments, extends beyond forensic dentistry. It also plays a fundamental role in anthropological research by providing insights into growth, health, and development in past populations [10].

However, biochemical techniques for age estimation are limited by their complexity, invasiveness, and unsuitability for use in living individuals. Consequently, radiographic methods are predominantly preferred. Over time, a variety of dental age estimation techniques have been developed, encompassing morphological, histological, and radiographic approaches. Among these, the method developed by Demirjian is particularly notable for its systematic evaluation of tooth development stages through panoramic and periapical radiographs [5, 11]. This method is based on the principle that pulp volume decreases due to secondary dentin formation, a process that begins after tooth eruption and continues throughout life.

Although many studies have utilized panoramic and periapical radiographs for age estimation, two-dimensional imaging often fails to provide the volumetric detail necessary for accurate assessment of dental structures. This limitation highlights the importance of advanced three-dimensional imaging technologies [7, 12].

CBCT significantly improves diagnostic precision by minimizing anatomical superimposition and providing high-resolution images in coronal, sagittal, and axial planes [13–15]. It offers superior sensitivity in capturing the three-dimensional architecture of dental tissues, enabling more accurate age estimations than traditional two-dimensional radiographs. CBCT allows for comprehensive volumetric analysis of both tooth and pulp structures [5, 16, 17].

Measurements of tooth and pulp volumes from CBCT images are typically performed using specialized software such as 3D Slicer. This open-source platform provides a robust set of tools for segmentation, registration, and quantitative analysis [18, 19]. These capabilities facilitate precise volumetric calculations and detailed morphological evaluations, thereby enhancing the reliability and accuracy of age estimation procedures. Recent studies also highlight the growing importance of automated or semi-automated 3D volumetric tools in forensic age estimation, contributing to reproducibility and cross-population application. By allowing meticulous processing of CBCT data, 3D Slicer has become an indispensable tool in forensic and anthropological research [18].

This study aims to evaluate whether the PV/TV ratio, assessed in mandibular canines—a single-rooted and anatomically stable tooth type [20]—using a standardized CBCT protocol, can serve as a robust and moderately accurate indicator of chronological age for forensic and anthropological purposes.

Material and method

Human ethics and consent to participate

Ethical approval for this study was granted by the Ethics Committee of the Necmettin Erbakan University, Faculty of Dentistry (decision dated 29/07/2021 and numbered 2021/08–77). Informed consent was obtained from all participants included in the study. Between 2019 and 2022, CBCT and digital panoramic images from 976 patients who visited the Necmettin Erbakan University, Faculty of Dentistry for various purposes were retrospectively reviewed. Funding: This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. Clinical trial number: not applicable.

Power analysis

A priori power analysis was performed using G*Power 3.1.9.7 software to determine the minimum required sample size for the study. The statistical test selected was linear multiple regression (fixed model, R² deviation from zero). The parameters were set as follows: medium effect size (f² = 0.15), significance level (α) of 0.05, test power (1 − β) of 0.95, and one predictor variable (pulp/tooth volume ratio). Based on these parameters, the analysis revealed that a minimum of 89 participants would be required to detect a statistically significant relationship. The present study included a total of 240 participants, thus exceeding the required sample size and ensuring sufficient statistical power.

From the initial pool, diagnostically adequate CBCT images of 240 patients aged between 17 and 72 years were selected from the archives of the Department of Oral and Maxillofacial Radiology.

All images were obtained using a 3D Accuitomo 170 (J Morita MFG Corp., Kyoto, Japan) unit with parameters 90 kVp, 5 mA and 15–18 s exposure time according to the manufacturer’s recommended protocol. A total of 240 individuals were retrospectively screened and included in the study. No participants were excluded after applying the eligibility criteria. All selected cases had diagnostically sufficient CBCT images suitable for pulp and tooth volume measurements.

The chronological age of the individuals was calculated by subtracting the date of birth from the date the radiograph was taken, and their gender information was recorded. The inclusion criteria excluded teeth with dental caries, periapical pathologies, severe attrition, abrasion, erosion, or developmental anomalies. Additionally, mandibular canine teeth were excluded if they exhibited restorations, fractures, previous endodontic treatment, malocclusion, rotation, or incomplete root development, or if veneer crowns were present. Teeth affected by occlusal trauma or with a history of craniofacial surgery or trauma were also excluded. A thorough literature review indicates that the choice of right or left canine has no significant effect on age estimation outcomes [21, 22]. Therefore, the canine with the best image clarity was selected for each patient in both digital panoramic and CBCT images, regardless of their side.

Segmentation was performed using 3D Slicer software (version 5.6.2; www.slicer.org), specifically the Segment Editor module. Initially, coronal, axial, and sagittal planes were determined using the orientation windows in the CBCT dataset. The region of interest (ROI) was isolated by applying the “volume rendering” function along with the “Display ROI” and “Crop” features to focus exclusively on the target mandibular canine. Segmentation proceeded semi-automatically using a combination of the Threshold and Grow from Seeds effects. Threshold values were determined based on gray-level intensity differences between pulp tissue and surrounding dentin. A Hue-Saturation threshold range of 370–445 was also specified to enhance visualization. In complex regions such as the apical and cervical thirds of the root, where automatic segmentation was less accurate, manual refinements were performed to ensure accurate margin determination. An example of the orientation adjustment and segmentation steps is shown in Fig. 1. All segmentation steps were completed by a single trained observer to ensure methodological consistency and reproducibility.

Fig. 1.

3D Slicer software program, the working screen shows the pulp volume in coronal, axial, sagittal view, respectively

Thus, the volumes of both the tooth and the pulp in a total of 240 mandibular canine teeth were automatically calculated in cubic millimeters (mm³) using software. Panoramic reconstruction images were obtained from the CBCT scans of 240 patients.

Intra-observer reliability was evaluated by re-examining 20 randomly selected images at two-week intervals.

Statistical analysis

Statistical analysis was performed using the Statistical Package for Social Sciences (SPSS) software (version 21.0, SPSS Inc., Chicago, USA) to evaluate the relationship between age and change in pulp volume.

A regression model was developed using the volume ratio variable (PV/TV), which demonstrated the strongest statistical correlation with age. To evaluate the performance of this regression model, the mean absolute error (MAE) and root mean square error (RMSE) were calculated.

The differences between the actual ages and the ages predicted by the regression formulas across gender and age groups were analyzed using Student’s t-test.

Normality of the pulp and tooth volume distributions was assessed using Q-Q plots (Fig. 2), which demonstrated acceptable conformity with a normal distribution.

Fig. 2.

Q-Q plots assessing the normality of (A) pulp volume (PV) and (B) tooth volume (TV) distributions across the total sample (n = 240). Both plots show that the data points closely follow the reference line, indicating that the distributions of PV and TV approximate normality and satisfy the assumptions required for parametric statistical analyses

The reliability of intra-observer measurements for linear assessments from volumetric measurements from CBCT images was evaluated using the paired-samples t-test. Intra-observer reliability was quantified using Cronbach’s alpha coefficient.

Results

A total of 240 patients aged between 17 and 72 years were initially examined in this study. The final sample consisted of 240 individuals, including 120 females (50%; age range: 18–70 years, mean age: 43.66 ± 15.01 years) and 120 males (50%; age range: 17–72 years, mean age: 43.40 ± 15.01 years). The participants were categorized into three age groups: 17–34 years, 35–54 years, and 55 years and older (Table 1).

Table 1.

Age-based distribution

Gender	17–34	35–54	55 and older	Total
Female	39	47	34	120
Male	40	44	36	120
Total	79	91	70	240

CBCT measurements

Intra-observer reliability was assessed through the re-evaluation of 20 randomly selected patients. The Cronbach’s alpha values for pulp volume and tooth volume were 0.993 and 0.998, respectively, indicating excellent measurement reliability.

The correlation between the PV/TV ratio and age was calculated using Pearson’s correlation coefficient. A statistically significant negative correlation was observed (r = − 0.710, p < 0.001).

The correlation between age and PV/TV ratio is further illustrated in Fig. 3, showing a linear trendline and distribution of data points.

Fig. 3.

Scatter diagram demonstrating the correlation between PV/TV ratio and age among 240 participants

Given this strong correlation, a regression analysis was performed to model the linear relationship. The resulting regression equations are presented in Table 2. When stratified by gender, the male group (n = 120) demonstrated r = − 0.741 and R² = 0.549, while the female group (n = 120) showed r = − 0.675 and R² = 0.455. These results suggest that the model explains a substantial portion of age variation in both sexes, with slightly greater predictive accuracy in males.

Table 2.

Predicted values ​​of age determination obtained with the regression formula created using the PV/TV variable

Group	Model	Regression coefficient	Standard error	Significance		Regression equation		Significance
				t	p			F	p
Total	Constant	83.114	2.666	31.180	0.000*	Age=83.114−(217.015× PV/TV)	0.504	241.976	0.000*
	PV/TV	-217.015	13.951	-15.556	0.000*
Female	Constant	81.755	4.027	20.300	0.000*	Age= 81.755−(208.548× PV/TV)	0.455	98.494	0.000*
	PV/TV	-208.548	21.014	-9.924	0.000*
Male	Constant	84.235	3.567	23.615	0.000*	Age=84.235−(224.293× PV/TV)	0.549	143.471	0.000*
	PV/TV	-224.293	18.726	-11.978	0.000*

The age estimation range using the regression formula based on the PV/TV ratio was ± 3.69 years at the 60% confidence interval and ± 7.24 years at the 95% confidence interval.

The 60% and 95% confidence intervals (± 3.69 and ± 7.24 years, respectively) were calculated based on the standard deviation of the residuals from the regression model, assuming a normal distribution of prediction errors.

The predictive accuracy of the regression model was further evaluated using the mean absolute error (MAE) and root mean square error (RMSE). For the main dataset (n = 240), the model yielded an MAE of 8.41 years and an RMSE of 11.19 years. When applied to an independent test set of 40 individuals (20 females and 20 males), the MAE and RMSE remained consistent at 8.41 and 11.19 years, respectively, indicating that the model generalizes well to unseen data.

To explore whether predictive accuracy varied across different age ranges, the sample was stratified into three age bands: 17–34 years, 35–54 years, and ≥ 55 years. The mean absolute error (MAE) values for these groups were 7.89, 8.12, and 9.23 years, respectively. These findings suggest slightly lower prediction errors in younger individuals and a modest increase in older age groups, indicating a relatively consistent but mildly age-dependent model performance.

To further evaluate the regression model, an independent test dataset comprising 40 CBCT scans (20 females, 20 males; age range: 17–71 years) was obtained from the same institutional archive but was not included in the initial 240-participant training dataset. The age and sex distribution of the test set was proportionally matched to that of the training set to ensure comparability. Age estimations were performed on this test group using the regression equations derived from the training data to assess the model’s generalizability. These estimated values were then compared with the actual chronological ages, and the corresponding results are presented in Table 3.

Table 3.

Average age values ​​by gender in the regression model created with PV/TV

Age (mean ± standard deviation)
Sex	Chronological age	Estimated age	t-test	p
Female	369.24	37.893.97	-0.021	0.981*
Male	39.3314.81	41.245.66	-0.765	0.426*
Total	37.5412.08	39.954.42	-0.438	0.659*

* No difference between estimated age and chronological age (p > 0,05)

Discussion

The present findings confirm a moderate inverse association between PV/TV ratio and chronological age (R² 0.504 for the total sample), consistent with previous studies linking secondary dentin deposition to pulp volume reduction over time [5, 23]. Regression analysis indicated that the model explains a considerable portion of age variation, particularly in males, suggesting the method’s practical relevance in forensic applications.

The use of three-dimensional volumetric analysis offers several advantages over two-dimensional methods by providing more accurate and reproducible assessments of internal tooth structures. The application of semi-automatic segmentation using 3D Slicer enabled efficient and standardized measurements of pulp and tooth volumes. This approach is in line with recent methodological advances in dental age estimation [24, 25]. Additionally, prior research has shown that voxel size plays a critical role in segmentation precision and model performance; finer voxel resolutions allow for more detailed pulp/tooth boundary delineation, thereby improving correlation strength with chronological age [5].

Compared to Kvaal’s method [26], which relies on radiographic length and width measurements from panoramic or periapical images, the volumetric approach adopted in this study yielded greater predictive accuracy. Previous studies have reported R² values ranging from 0.12 to 0.35 when using Kvaal’s method [26, 27], whereas our CBCT-based regression model achieved an R² of 0.504 in the total sample. This suggests that CBCT-derived volume measurements may be more sensitive to age-related morphological changes within the pulp cavity.

The prediction errors observed — ±3.69 years at the 60% confidence interval, and ± 7.24 years at the 95% confidence interval— support the model’s utility in age estimation tasks where moderate precision is sufficient. The correlation between age and PV/TV ratio is further illustrated in Fig. 3, showing a linear trendline and distribution of data points. Although the margin of error is greater than that of high-resolution histological methods, it remains within the acceptable range for both forensic and clinical applications [28]. In forensic contexts, an error margin of approximately ± 7–11 years, as found in the present study, is generally deemed acceptable, especially when age estimation is used as a complementary tool. While this level of accuracy may be insufficient for definitive legal determinations near critical age thresholds (e.g., 18 years), it can still offer meaningful guidance when interpreted alongside other indicators such as clinical, skeletal, or documentary evidence. For jurisdictions requiring greater precision, integrating multiple estimation methods may enhance the reliability of age assessments.

A notable finding was the slightly better performance of the model in male participants (R² = 0.549) compared to females (R² = 0.455). While this may suggest the influence of biological factors such as dentin deposition rates or sample heterogeneity, there is currently insufficient direct evidence to confirm a hormonal basis. Therefore, this interpretation remains speculative and should be investigated further in future studies. Studies in forensic odontology have indicated that the process of secondary dentin deposition—a key factor in dental age estimation—varies between sexes [29, 30].

In addition to correlation coefficients, prediction error metrics were used to evaluate the model’s precision. The mean absolute error (MAE) and root mean square error (RMSE) for the main dataset were calculated as 8.41 and 11.19 years, respectively. These values reflect a moderate degree of prediction error and are consistent with previously reported CBCT-based models [16, 31–33]. Notably, when the model was applied to an independent test dataset of 40 individuals, the MAE and RMSE values remained consistent (8.41 and 11.19 years), indicating the model’s stability and generalizability across different samples.

For instance, Shruthi et al. [29] compared various dental age estimation methods in a South Indian population and reported that differences between male and female subjects could be explained by variations in the rate and pattern of secondary dentin deposition. This finding supports our observation that female participants demonstrated a slightly higher coefficient of determination in regression models, possibly due to lower variability in biological aging markers compared to males.

Moreover, Bajpai and Rahman [30] demonstrated that age estimation methods based on secondary dentin deposition, root translucency, and cementum apposition exhibited distinct error margins that may be attributed to biological discrepancies between sexes. Their findings suggest that hormonal influences—potentially affecting the rate of dentin deposition—can lead to gender-specific differences in chronological age estimation using dental structures. Hormonal effects, as highlighted in animal models, are further supported by the work of Smid et al. [34], who showed that growth hormone status influences dentin thickness and morphology. Although conducted in a murine model, their study provides mechanistic insight into hormonal regulation that may contribute to the observed gender-specific differences in dental aging. Collectively, these studies support the interpretation that the slightly higher predictive accuracy observed in female participants in our regression models may reflect underlying biological and hormonal influences on dentin deposition, as well as sample heterogeneity. These findings emphasize the importance of incorporating gender-specific variables or adjustments in forensic age estimation protocols to enhance accuracy and reliability.

Salemi et al. [31] utilized CBCT-based measurements of canine teeth to estimate age and reported statistically significant correlations between dental morphological variables and chronological age. Notably, they achieved a correlation coefficient of 0.88 in their regression model, suggesting that CBCT-derived measurements—whether area- or volume-based—can yield strong predictive performance when appropriate anatomical landmarks and standardized protocols are applied. Although their study focused on area ratios in maxillary canines, the significant negative correlation reported is consistent with our findings, reinforcing the notion that regressive changes in pulp volume are reliable indicators of chronological age. In addition, Abdinian et al. [35] investigated pulp-tooth volume ratios in anterior teeth using CBCT and proposed regression equations for age estimation, supporting the validity of volumetric parameters for forensic applications, similar to the regression models developed in the present study (R² ≈ 0.504).

Furthermore, studies involving different geographic populations, such as the work by Alqarni et al. [32], provide additional support. Their study using pulp–tooth area ratios demonstrated a significant inverse relationship with age, affirming the robustness of CBCT-derived parameters across various demographic groups. Fayek et al. [33] demonstrated that volumetric analysis of the pulp chamber-to-crown volume ratio on CBCT images yielded regression equations with high coefficients of determination (up to R² = 0.812), suggesting that the precision of age estimation can be improved through optimized segmentation and measurement protocols.

To increase interpretability and facilitate comparison across studies, key methodological parameters and performance indicators from recent CBCT and Kvaal-based age estimation studies [5, 26, 33, 36, 37] are outlined in Table 4.

Table 4.

Summary of recent studies on dental age Estimation using pulp/tooth ratios

Author (Year)	Method	Sample Size	Tooth Type	Imaging Modality	Key Metric (R² / MAE)
Kvaal et al. (1995)	Length Ratios	100	Various	2D Radiographs	R² = 0.30–0.50
Adisen et al. (2020)	Pulp/Tooth Volume	200	Maxillary canines	CBCT	R² = 0.52; MAE ≈ 8.2 y
Molina et al. (2021)	Pulp/Tooth Volume	313	Max/Mand anterior teeth	CBCT	(Max insisors) R² = 0.366
de Souza et al. (2023)	Kvaal linear regresyon	554	Maxillary incisos/canines	OPG	R² = 0.335; MAE ≈ 5.3–5.7 y
Fayek et al. (2024)	Pulp/Tooth Volume	1000	All types teeth	CBCT	R² ≤ 0.812
Current Study (2025)	PV/TV Ratio	240 (+ 40 test)	Mandibular canines	CBCT	R² = 0.504; MAE = 8.41 y

Yousefi et al. [38] presented systematic evidence in their meta-analysis that three-dimensional imaging yields more reproducible and accurate correlations between pulp volume and chronological age, underscoring the advantage of CBCT-based assessments.

While Rudolphi et al. [1] applied modified Gustafson criteria on panoramic radiographs and reported lower predictive accuracy (R² = 0.22–0.35), the current volumetric method demonstrated stronger correlations, suggesting that the pulp/tooth volume ratio provides a more precise metric for age estimation.

In contrast to Kotze et al. [8], who utilized two-dimensional pulp/tooth area ratios based on periapical and stereomicroscopic images of maxillary canines, our fully volumetric CBCT approach offers improved accuracy by eliminating issues related to projection geometry and enamel-associated measurement variability. Although Kotze et al. [8] eported a relatively high R² of 0.76 when combining stereomicroscopy and digital radiographs, their method produced a mean absolute error (MAE) of 6.97 years—comparable to the error margin observed in our CBCT-based model. Moreover, their study was conducted on cadaveric specimens, which limits the generalizability of their findings to in vivo settings. By contrast, the present study used radiographic data from living individuals, enhancing the clinical relevance and forensic applicability of the PV/TV approach.

Additionally, the gender-based variability observed in our regression models—with R² values of 0.455 in females and 0.549 in males—was also reported by Rathore et al. [39], who assessed age based on mandibular pulp volume measurements in an Indian population. Their findings highlight the necessity of incorporating sex-specific regression parameters to improve accuracy in forensic applications.

Limitations

Despite the strengths of the present study, several limitations must be acknowledged. First, the sample consisted exclusively of mandibular canine teeth, which may limit the generalizability of the findings to other tooth types. While mandibular canines were selected for their structural stability and single-rooted anatomy, including additional teeth in future studies may enhance the robustness of the model. Second, the retrospective design may introduce selection bias, and although care was taken to ensure high-quality imaging, the radiographs were not originally acquired for research purposes. Third, although pulp and tooth volumes were measured using semi-automatic segmentation in 3D Slicer, some degree of manual correction was required, potentially introducing operator-dependent variability. Fourth, the study was conducted at a single center, which may limit the external validity of the findings. Therefore, multi-center studies involving more ethnically and geographically diverse populations are warranted to confirm the generalizability of the results. Fifth, although CBCT images were obtained using standardized device settings, minor voxel-size variations across scans may still have influenced the precision of volumetric measurements. Future studies should assess and control for this potential source of variability. Finally, although the regression model demonstrated moderate predictive power, its applicability across populations with different ethnic, genetic, and environmental backgrounds has yet to be validated.

Conclusion

This study demonstrated that the pulp/tooth volume ratio (PV/TV), derived from CBCT images of mandibular canines, can serve as a moderately (± 7–11 years error) reliable indicator for chronological age estimation. The method revealed a consistent inverse relationship between pulp volume and age, with moderate prediction accuracy.

Given the non-invasive nature of CBCT imaging and the acceptable error margins observed, the PV/TV method may be used as a complementary tool in forensic and anthropological contexts. While it may not replace more precise age estimation techniques in all cases, it offers a practical and efficient alternative—particularly when access to biological specimens is limited. Future studies should explore the use of multi-tooth or whole-jaw models, as well as fully automated segmentation models to reduce observer dependency and improve reproducibility. Moreover, external validation in larger, ethnically diverse populations is crucial to assess the model’s generalizability and practical applicability across different settings.

Abbreviations

CBCT

Cone-beam computed tomography

MAE

Mean absolute error

RMSE

Root mean square error

PV/TV

Pulp volume/Tooth volume

Author contributions

AA, and İBY designed the study. İBY collected the data. Statistical analyses were performed by AA. Data interpretation and manuscript preparation were carried out by İBY, and AA. İBY, and AA also conducted the literature search. The study was supervised by AA. All authors reviewed the manuscript.

Funding

None. The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

Data is available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

The study received ethical approval from the Research Ethics Committee of the Faculty of Dentistry at Necmettin Erbakan University, and the study conducted in accordance with the Helsinki Declaration of Human Rights guidelines (Approval Date: 29/07/2021, and Approval Number:2021/08–77). Informed consents were obtained from all patients, and all data were processed anonymously.

Informed consent

Informed consent was obtained from all patients involved in the study.

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.