Chairs:
dorothea dagassan
raluca roman

123: ACCURACY OF PERIAPICAL RADIOGRAPHS AND CBCT FOR DIAGNOSING APICAL PERIODONTITIS. AN EX VIVO HISTOLOGICAL STUDY ON HUMAN CADAVERS

L.-L. Kirkevang1, C. Kruse1, R. Spin-Neto1, M. Væth2

1Aarhus University, Department of Dentistry and Oral Health, Section for Oral Radiology and Endodontics, Aarhus, Denmark, 2Aarhus University, Department of Public Health, Aarhus, Denmark

Aim: To assess the diagnostic accuracy of periapical radiographs (PA) and CBCT to diagnose apical periodontitis (AP) using histology as reference standard.

Material and Methods: PA and CBCT were performed on 221 teeth from human specimens. Teeth from all tooth groups, with different disease and treatment status were included: root filled teeth (n=71), non-root filled teeth (n=150). PA were acquired using a ‘Gendex 1000 DC’ X-ray unit (Gendex Corporation, Milwaukee, WI, USA), and phosphor plate system (Dürr Dental VistaScan Plus, Dürr Dental AG, Bietigheim-Bissingen, Germany). CBCT was performed using Cranex® 3Dx (Soredex Oy, Tuusula, Finland, FOV 5×5 cm, and resolution 0.08 mm). The presence of AP was assessed in PA using Periapical Index, and in CBCT using a probability index. Histology was used as a reference standard to compute sensitivity, specificity, positive predictive values (PPV), and negative predictive values (NPV)

Results: On PA non-root-filled teeth produced estimates of 73%, 89%, 53%, and 95% regarding sensitivity, specificity, PPV, and NPV, respectively. For root-filled teeth, the estimates were 73%, 68%, 63%, and 78%, respectively. On CBCT non-root-filled teeth produced estimates of 96%, 92%, 75%, and 99% regarding sensitivity, specificity, PPV, and NPV, respectively. For root-filled teeth, the estimates were 67%, 70%, 71%, and 76%, respectively.

Conclusion: The diagnostic accuracy for diagnosis of AP differs whether the tooth is root filled or not, and it is lower in root-filled teeth for both PA and CBCT. In particular, PPV-values were low, indicating a high risk of over-diagnosis in both diagnostic modalities.

148: INTERPRETATION OF INCIDENTAL HYPERDENSE SOFT TISSUE FINDINGS. A RETROSPECTIVE STUDY ON 511 CONE BEAM CT DATASETS

A. Zoukos1, S. Damaskos1, C. Angelopoulos1

1School of Dentistry, National and Kapodistrian University of Athens (NKUA), Oral Diagnosis and Radiology, Athens, Greece

Aim: To retrospectively evaluate and record hyperdense extra-gnathic findings in anatomic regions wherein soft tissue structures are located using a series of large field of view (FOV) CBCT datasets.

Materials & Methods: 511 anonymized consecutive large FOV CBCT datasets (voxel size 0.3mm) were obtained from the archive of the Department of Oral Diagnosis & Radiology. These were systematically evaluated and analyzed, by consensus, by a panel of two (2) oral and maxillofacial radiologists. Each finding in the scan volume was recorded separately as well as the patient‘s age and gender.

Results: A wide spectrum of hyperdense soft tissue findings was finally recorded to those of intermediate-high clinical significance such as intracranial carotid artery calcified atheromas (42.1%), extracranial carotid artery calcified atheromas (23.5%), and degenerative changes in the craniocervical joint, in particular, findings consistent with osteophytic changes located at the attachment sites of the anterior atlantoccipital membrane (40.5%), apical ligament of dens (36.4%), and anterior atlantoaxial joint (20%) being the most prevalent.

Conclusions: While apparently benign, hyperdense soft tissue findings can be of intermediate-high clinical significance. At any rate, thorough evaluation of large FOV CBCT datasets is highly recommended.

216: CBCT OF THE JAW AND ITS USEFULNESS IN FORENSIC DENTISTRY: SCOPE REVIEW

E. Parraguez López1, M.P. Cancino Villarroel1, C. González Elgueta1

1Universidad Mayor, Oral Radiology, Santiago, Chile

Aim: Determine anatomical structures of the jaw through CBCT, contributing to individual forensic identification.

Material & methods: Digital search in PubMed and Google Scholar databases with keywords in English with respective Boolean operators performed. Inclusion criteria: articles published between 2013-2023; full text; in English; related to human beings, all ethnic groups and ages. Exclusion criteria: studies performed with radiographic examinations other than CBCT; odontometric parameters; bone structures analyzed other than jaw; less than 50 patients; systematic review type studies; age an identification characteristic.

Results: 14 articles included for this scoping review. Two studies related to individual‘s own identity through mandibular lingual canals showed significant morphological variability in each individual. Precision rates revealed values between 96%-100% of agreement among records from the same jaw. The other 12 were carried out in 6 countries with 2,666 CBCT. From 93 parameters analyzed, 73 presented dysmorphic characteristics gender dependent. Four most common were: ramus height, minimum mandibular ramus width, bigonial width and angle mandibular. Sex determination accuracy rates between 60% and 95.1% ( 77.55%). Males presented higher average values.

Conclusion: Mandibular anatomical structures contribute to identify individuals, both for gender and corroborating identity. However, structures show variations between various populations and ethnicities and is not possible to establish universal dysmorphic characteristics to identify gender. Rather, it can be specified what mandibular parameters contribute in a specific population. An exception to this observation is found in the middle mandibular region, specifically in the mandibular lingual canals, which are distinctive and unique for each person, determining their identity.

235: AGE ESTIMATION AND SEX DETERMINATION BY ARTIFICIAL INTELLIGENCE IN A GREEK POPULATION STUDY: A PILOT STUDY

A. Mitsea1, M. Vodanović2, N. Christoloukas1, C. Angelopoulos1

1Dental School, NKUA, Oral Diagnosis and Radiology, Athens, Greece, 2School of Dental Medicine, University of Zagreb, Croatia, Department of Dental Anthropology, Zagreb, Croatia

Objective: This study aims to evaluate the efficacy of an artificial intelligence system to estimate the age and determine the sex in a Greek population sample.

Material and Methods: This study‘s sample comprised panoramic radiographs obtained from 110 adult subjects. Sex distribution was evenly balanced, (55 males/55 females) in the sample. Each subject‘s dental status varied. The sample’s mean age was 48,87 years (±16,14 yrs), ranging from 20 to 84 years. The methodology employed a beta version of a convolutional neural network (CNN) software developed by the University of Zagreb, leveraging ReLU and SoftMax functions.

Results: In only one case, concerning sex determination, and in 21 instances regarding age estimation, results could not be provided. The accuracy of sex determination approached 96.09%. Regarding dental age estimation, the accuracy appeared to be at 60%.

Conclusions: This fully automated analysis of panoramic images, requires no specific criteria for images and imposes no restrictions on the ethnic background of the subjects. With an accuracy of 96.09%, the results for sex determination were considered particularly satisfactory. It seems that the image quality, the number of teeth, and the number and kind of dental treatments may affect the accuracy of this software. The current limitations of this software are that it was trained on orthopantomograms of the Croatian population and that all images used for training were made on only one type of X-ray device. For even better results in the training process, orthopantomograms from different devices and populations should be used.

239: ARTIFICIAL INTELLIGENCE IN FORENSIC DENTISTRY IDENTIFICATION. A SYSTEMATIC REVIEW

K. Briamatou1, N. Christoloukas1, A. Mitsea1, C. Angelopoulos1

1NKUA, Oral Diagnosis and Radiology, Athens, Greece

Aim: The aim of this systematic review is to evaluate the scientific evidence concerning the implementation of Artificial Intelligence in Forensic Dentistry Identification.

Material and Methods: An electronic systematic literature search was conducted according to the PRISMA guidelines by using four electronic databases; PubMed, Web of Science, Scopus, and Cochrane Library for articles published from January 2014 to March 2024. The essential information was recorded using a predefined data extraction form. The Cochrane method of bias: ROBINS-I tool was applied to assess the quality of evidence.

Results: Initially, 579 articles were retrieved, after the duplicate removal of 350 articles, the remaining 229 were assessed for their relevance to our research question. Studies were evaluated based on inclusion and exclusion criteria, by which 83 articles remained by title and 59 articles remained by abstract. Finally, 50 articles were eligible for this systematic review via full text reading. The following AI methods are included in this systematic review: convolutional neural networks, machine learning algorithms, and hybrid models. AI methods demonstrate high accuracy rates: the highest accuracy rate was 95% for age estimation, 95.2% for gender estimation, 100% for dental chart identification, 98.27% for segmentation, and 98.9% for identification based on dental implants.

Conclusion: This systematic review supports the importance of AI in Forensic Dentistry Identification. The implementation of AI techniques in forensic dentistry introduces a new era of innovation, offering a synergy of human expertise and technological capabilities. Further high-quality research is needed to verify the outcomes of this systematic review.