Ex vivo detection of mandibular incisors’ root canal morphology using cone-beam computed tomography with 4 different voxel sizes and micro-computed tomography

CBCT is a valuable diagnostic tool in endodontics. The selection of an appropriate scanning protocol is critically important to balance between maintenance of essential image quality and the reduction of radiation dose [30]. The voxel size is one of the most important scanning parameters, and in most of the cases, an optimal, rather than the minimum voxel size is desirable for achieving the best diagnostic efficiency, as a smaller voxel size is associated with more radiation exposure, increased noise level and decreased contrast-to-noise ratio, while might provide a same diagnostic outcome as lower resolution images [16, 21, 31, 32].

In the current ex vivo study, to identify the presence of a double root canal system and ACs in mandibular incisors, extracted teeth were fixed in the tooth sockets of dry mandibles with wax to simulate the in vivo condition. All observers assessed CBCT image sets taken at 4 different voxel sizes as our clinical routine, while the micro-CT images at 9 μm voxel size were used as the gold standard. During ROC curve analysis, a high AUC value closing 1.0 indicated an ideal diagnostic test, while a low AUC value closing 0.5 indicated no diagnostic effectiveness. The data in Table 3 reveals that all 4 voxel sizes are reliable in identifying the double root canal system (the AUC ranged from 0.765 to 0.967, all p < 0.0001), and for all 4 observers, a voxel size of 125 μm could achieve high diagnostic accuracy (all AUC > 0.9, and all p < 0.0001), while a voxel size of 300 μm could only achieve moderate diagnostic accuracy (the AUC ranged from 0.765 to 0.889). Professional experience is also an influencing factor, as for Observer 1 (a researcher with 9 years of working experience in radiology) and Observer 2 (an experienced endodontic specialist), the AUC values were greater than those of Observer 3 and 4 (both were graduate students of dental school) at each voxel size. Statistical differences were detected between Observer 1 and 4, and Observer 2 and 4 at 300 μm voxel size, and between Observer 2 and 4 at 200 μm voxel size (all p < 0.05). These findings also indicate that sharp CBCT images with high resolution (125 μm voxel size) are more essential for an inexperienced clinician; while for an experienced specialist, CBCT with a lower resolution (200 μm voxel size) is only slightly less accurate than the high-resolution modality (125 μm voxel size), which has an advantage of lower exposure dosage. This is in contrast to the report of Bauman et al. [22], who noted that the accuracy of MB2 canal detection couldn’t be correlated with the observers’ level of clinical experience. Previous scholars demonstrated that the presence of artifacts can also affect the diagnostic ability of CBCT [33]. Vizzotto et al. [23] found that the root-filling materials in MB1 may affect the detection of MB2, and the degree of influence was dependent on the CBCT voxel size during the examination. Recently, Lee et al. [34] retrospectively examined CBCT images of a Korean population and reported that the mandibular first molar showed a high incidence (27.0%) of separate disto-lingual (DL) roots, and this root variation was highly associated with the presence of two‑canaled mandibular incisors (left central incisors: p = 0.001, odds ratio = 4.25; left lateral incisors: p < 0.001, odds ratio = 3.8; right central incisors: p = 0.003, odds ratio = 3.86; right lateral incisors: p = 0.001, odds ratio = 3.44). Similar finding has been reported in the Taiwanese population [10, 11], which indicates that when the voxel size and quality of CBCT images are inadequate for accurate detection of two-canaled mandibular incisors, the root number of the mandibular first molars may provide clinicians some clues. The DL root can more easily be diagnosed, and its presence indicates a high possibility of two canals in mandibular incisors on the same side.

Literature review shows only very few scholars have previously assessed the diagnostic potential of CBCT in identifying ACs [16, 17]. To evaluate the effectiveness of CBCT imaging (using a voxel size of 80 μm) for the detection of lateral canals in endodontically treated premolars, a similar ex vivo study was carried out by Sousa et al. [16]. They reported that the AUC value was only 0.58 and 0.49 (all p < 0.001) before and after root canal treatment, respectively, and CBCT imaging was not an effective diagnostic tool for the detection of ACs. The current study found that CBCT images taken at 125 μm voxel size have a low (AUC: 0.5–0.7) diagnostic potential in the detection of ACs in mandibular incisors for all 4 observers. The AUC values ranged from 0.638 to 0.678 (all p < 0.05), and the sensitivity was at a low level ranging between 36.8% and 57.9%, while the specialty was at a relatively high level ranging between 73.6% and 98.8%. The higher specificity value can be due to the observers’ difficulty in detecting the accessory canals and they tended to score “4 = probably absent” or “5 = absent”, which resulted in a large number of “negative” answers, consequently increasing the number of “true-negative” cases [16]. This finding is similar to that of Ji et al. [17], who reported that partial lateral accessory canals could be detected using CBCT at high-resolution scan mode (voxel size = 125 μm). Theoretically, CBCT cannot identify an anatomic structure with a size below the voxel size. Moreover, the relationship between CBCT spatial resolution and voxel size indicated that the former is at least twice greater than the latter [35]. The odontometric measurement by micro-CT showed that the majority of the ACs had a canal size below 250 μm (Fig. 3), which means even using the smallest voxel size of 125 μm, the spatial resolution (which is often 2-fold greater than the voxel size) is inadequate to identify the majority of ACs. The low sensitivity also indicated that the majority of ACs were missed diagnoses. Theoretically, more advanced modulates with 75–80 μm voxel size can improve diagnostic accuracy.

This ex vivo study has several limitations. First, to simulate the in vivo dental anatomy, the extracted sample teeth were fixed in the tooth sockets of dry mandibles, which were stationary during CBCT scanning. While in the clinical scenario, patient movement is also an influential factor, with the magnitudes even exceeding voxel size and scatter effects, and this may lead to overestimated diagnostic accuracy of the CBCT. Second, in the current study, the majority of the sample teeth were collected from elderly people, and the small size of the main canals or ACs was associated with a lower CBCT detection rate. Finally, this study only used one CBCT machine (3D eXam I, KaVo) with the smallest voxel size of 125 μm, which was not the most advanced machine currently in the CBCT market. Therefore, further studies are essential based on other brands of CBCT units, especially those with a minimum voxel size of 75–80 μm.