Teeth selection
This study was approved by the Peking University Ethics Committee and Competent Authority (No. PKUSSIRB-202168154).
Periapical radiographs and CBCT results of the upper premolar were collected from the previous periodontitis case database of the oral radiology department in our hospital [Peking University School and Hospital of Stomatology] for analysis. Intraoral periapical radiographs were obtained using the bisecting-angle projection technique. Scans of premolars in vivo were obtained using a CBCT scanner (NewTom VG, QR s.r.l.) at 110 kV and 5 mA. The field of view was 12 cm × 8 cm, and the layer thickness was 0.3 mm. CBCT data of all patients were taken for therapeutic (such as prosthodontic and orthodontic) purposes and were not related to the study.
The inclusion criteria were as follows: patients (1) aged 18–40 years with systemic health data, (2) with > 20 remaining teeth, (3) with clear periapical radiographs and CBCT of the upper premolar, and (4) with single and intact roots of the upper premolar.
The exclusion criteria were as follows: patients (1) who were systemically unhealthy with tumour or metabolic disease; (2) with oral and maxillofacial acute inflammation or tumour; (3) with an orthodontics history; (4) with root resorption, filling in the root, or root fracture of the premolar; (5) with crowding or malposition in the premolar region; (6) with curved-rooted, syncretic-rooted, or short-rooted teeth; and (7) with no periodontal ligament on radiographs.
Conical root inspection
The conical root was identified visually on periapical radiographs by three experienced periodontists. All teeth were divided into two groups, which were referred to as the cone-rooted teeth (CRT) and normal-rooted teeth (NRT) groups.
Measurements on periapical radiographs
Root width parameters
The parameters of the upper premolars measured on periapical radiographs (Fig. 1) included the following: parameters of root width (PRW): the median points of the lines joining the mesial or distal cemento-enamel junction (CEJ) point and apex were referred to as B or C. The intersection point of the line and the mesial or distal margin of the root was defined as A or D: root width parameter = (AD − BC)/2.
Measurements on cone-beam computed tomography
Root length
Mimics (Quotation Mimics 21.0, Materialise Dental) was used for the analysis. After 3D reconstruction, the images were oriented in three perpendicular planes to measure the RL on the buccal-lingual plane (Fig. 1). RL was defined as the distance between the apex and median point of the line that joins the buccal and lingual CEJ points.
Root surface area and root volume
After 3D reconstruction, the margin of the premolar was determined manually from every two layers on the cross-sectional and sagittal plane images. The crown and root portions were segmented separately to the CEJ manually from each layer on the sagittal plane images. After segmentation, crown, root, and tooth masks were exported in STereoLithography format and parameters such as the RSA and RV were measured using Mimics (Fig. 2). The RSA was calculated using the formula RSA = (SMT + SMR − SMC)/2 [15]. The RV can be calculated automatically using software. RSA/RL and RV/RL were calculated manually.
Raw colour map
All 3D original surface models from the same teeth (upper first premolar or upper second premolar) of different patients were unified in a common coordinate system and loaded in Geomagic (Geomagic Wrap 2017, 3D System). The average 3D surface models, which were the average of original models of the same teeth, were created by achieving the position of the minimum distance of the corresponding marker points through coordinate transformation under the principle of the least squares method. Furthermore, 3D differences between normal and conical root teeth were calculated to assess the morphological differences.
To help visualise the differences, a 3D, colour-coded map for the differences between the normal and conical root teeth models was generated, where the normal root teeth were displayed as smaller (negative, blue), the same (0 surface distances, green), or larger (positive, red) than the conical root teeth. The blue to red colour-coded scale was standardised, allowing a proper comparison, and pure blue and red were set at − 1.145 mm and 1.145 mm, respectively.
The raw colour map of the average models for NRT and CRT showed regions of statistically significant surface-to-surface differences (Fig. 3).
Sample size
Based on the following formula, the sample size of this study was calculated:
$$n_{1} = \frac{k + 1}{k}\left[ {\frac{{\left( {z_{\alpha /2} + z_{\beta } } \right)\sigma }}{\delta }} \right]^{2}$$
Based on the results of the preliminary trial, if \(k,\) α, and β values were 0.75, 0.05, and 0.20, respectively, 54 teeth and 41 teeth were required for the NRT and CRT groups, respectively.
Statistical analyses
All statistical analyses were performed using Statistical Package for the Social Sciences Statistics version 25 software (IBM Corp., Armonk, NY, USA). Every measurement of the periapical radiograph and CBCT scan was repeated after 1 week by the same researcher, and the intra-examiner error was tested using a paired t-test. The intraclass correlation coefficient (ICC) was used to evaluate the magnitude of the measurement error. The mean of the two measurements was used in this study. The group t-test and nonparametric test were used for statistical analysis. P < 0.05 was considered significant for all tests.
The RA diagnosed visually on periapical radiographs was referred to as the clinical gold standard for comparison, and the diagnostic criteria remained stable in the study. Receiver operator characteristic (ROC) curves were generated, and areas under the curves (AUCs) were calculated to evaluate the diagnostic value of PRW, RV, RSA, RV/RL, and RSA/RL. After performing statistical analyses, the cut-off values were calculated for PRW, RV, RSA, RV/RL, and RSA/RL. The accuracy of the CBCT method in detecting root abnormalities was re-evaluated using the cut-off values. Subsequently, the sensitivity (Se), specificity (Sp), positive and negative predictive values, Youden index (YI), and positive and negative likelihood ratios were calculated.