Maxillofacial Radiology Research Articles

Evaluation of an artificial intelligence system for the diagnosis of apical periodontitis on digital panoramic images.

The aim of the present study was to evaluate the effectiveness of an artificial intelligence (AI) system in the detection of roots with apical periodontitis (AP) on digital panoramic radiographs. Three hundred and six panoramic radiographs containing 400 roots with AP (an equal number for both jaws) were used to test the diagnostic performance of an AI system. Panoramic radiographs of the patients were selected with the terms 'apical lesion' and 'apical periodontitis' from the archive and then with the agreement of two oral and maxillofacial radiologists. The radiologists also carried out the grouping and determination of the lesion borders. A deep learning (DL) model was built and the diagnostic performance of the model was evaluated by using recall, precision, and F measure. The recall, precision, and F-measure scores were 0.98, 0.56, and 0.71, respectively. While the number of roots with AP detected correctly in the mandible was 169 of 200 roots, it was only 56 of 200 roots in the maxilla. Only four roots without AP were incorrectly identified as those with AP. The DL method developed for the automatic detection of AP on digital panoramic radiographs showed high recall, precision, and F measure values for the mandible, but low values for the maxilla, especially for the widened periodontal ligament (PL)/uncertain AP.

Radiographic Evaluation of Medication-Related Osteonecrosis of the Jaw (MRONJ) With Different Primary Cancers and Medication Therapies.

Medication-related osteonecrosis of the jaw (MRONJ) is a condition that affects the jaws and is characterized by exposed bone in the oral cavity that persists for more than eight weeks despite treatment. Additional criteria include that the patient should have a current or past history of antiresorptive drugsand/or in combination with antiangiogenic drugs, absence of metastasis, and no previous radiotherapy to the affected area. The radiographic features of MRONJ in most instances do not have any specific radiographic features. This is because standard radiographs usually show no stark abnormalities in the early stages of the disease. The study aimed to evaluate if any specific radiographic patterns are associated with primary cancers and between medications. The study is an observational case series. A total of 50 cases of possible osteonecrosisfrom June 2010 to June 2013 archives of the Department of Oral and Maxillofacial Radiology were assessed. Based on the history, 12 cases that had a history of medication use that could lead to medication-related osteonecrosis of the jaw (MRONJ) were selected. Cone beam computed tomography (CBCT) scans of these 12 cases were evaluated using the CBCT reconstruction program InVivo Dental version 6 (Anatomage Inc., San Jose, CA, USA). The number of areas showing sequestration, the pattern of osteonecrosis, and the extent were assessed. Primary cancer and the type of medication were also assessed to identify if certain cancers or drugs showed any distinctive pattern of osteonecrosis. Reconstructed panoramic images and true three-dimensional (3D) multi-planar images were assessed to study the condition. An oral and maxillofacial radiology resident in training and a board-certified oral and maxillofacial radiologist assessed the images. Radiographic findings varied among the 12 cases and included generalized sclerosis, osteosclerosis with widened periodontal ligament (PDL) space, bony sequestra, and bony necrosis. However, no specific or distinctive radiographic patterns were observed in any of the cases, regardless of the type of primary cancer or medication used. It is challenging to radiographically distinguish between MRONJ cases with different primary cancer and/or medication. Future studies should include evaluating larger samples with varying primary cancers and medications and combination drug therapies. Cases in an advanced stage of MRONJ do not have distinctive features due to extensive destruction and superimposed infection;it may be valuable to evaluate patients in the early stages of MRONJ to better understand distinguishing radiographic patterns specific to certain primary cancers or medications.

The Potential Usefulness of ChatGPT in Oral and Maxillofacial Radiology.

Aim This study aimed to evaluate the potential usefulness of Chat Generated Pre-Trained Transformer-3 (ChatGPT-3) in oral and maxillofacial radiology for report writing by identifying radiographic anatomical landmarksand learning about oral and maxillofacial pathologiesand their radiographic features. The study also aimed to evaluate the performance of ChatGPT-3 and its usage in oral and maxillofacial radiology training. Materials and methods A questionnaire consisting of 80 questions was queried on the OpenAI app ChatGPT-3. The questions were stratified based on three categories. The categorization was based on random anatomical landmarks, oral and maxillofacial pathologies, and the radiographic features of some of these pathologies. One oral and maxillofacial radiologist evaluated queries that were answered by the ChatGPT-3 model and rated them on a 4-point, modified Likert scale. The post-survey analysis for the performance of ChatGPT-3 was based on the Strengths, Weaknesses, Opportunities, and Threats (SWOT) analysis,its application in oral and maxillofacial radiology training, and its recommended use. Results In order of efficiency, Chat GPT-3 gave 100%accuracy in describing radiographic landmarks. However, the content of the oral and maxillofacial pathologies was limited to major or characteristic radiographic features. The mean scores for the queries related to the anatomic landmarks, oral and maxillofacial pathologies, and radiographic features of the oral and maxillofacial pathologies were 3.94, 3.85, and 3.96, respectively. However, the median and mode scores were 4 and were similar to all categories. The data for the oral and maxillofacial pathologies when the questions were not specifically included in the format of the introduction of the pathology, causes, symptoms, and treatment. Out of two abbreviations, one was not answered correctly. Conclusion The study showed that ChatGPT-3 is efficient in describing the pathology, characteristic radiographic features, and describing anatomical landmarks. ChatGPT-3 can be used as an adjunct when an oral radiologist needs additional information on any pathology, however, it cannot be the mainstay for reference. ChatGPT-3 is less detail-oriented, and the data has a risk of infodemics and the possibility of medical errors. However, Chat GPT-3 can be an excellent tool in helping the community in increasing the knowledge and awareness of various pathologies and decreasing the anxiety of the patients while dental healthcare professionals formulate an appropriate treatment plan.

Root Dilaceration Using Deep Learning: A Diagnostic Approach

Understanding usual anatomical structures and unusual root formations is crucial for root canal treatment and surgical treatments. Root dilaceration is a tooth formation with sharp bends or curves, which causes dental treatments to fail, especially root canal treatments. The aim of the study was to apply recent deep learning models to develop an artificial intelligence-based computer-aided detection system for root dilaceration in panoramic radiographs. A total of 983 objects in 636 anonymized panoramic radiographs were initially labelled by an oral and maxillofacial radiologist and were then used to detect root dilacerations. A total of 19 state-of-the-art deep learning models with distinct backbones or feature extractors were used with the integration of alternative frameworks. Evaluation was carried out using Common Objects in Context (COCO) detection evaluation metrics, mean average precision (mAP), accuracy, precision, recall, F1 score and area under precision-recall curve (AUC). The duration of training was also noted for each model. Considering the detection performance of all models, mAP, accuracy, precision, recall, and F1 scores of up to 0.92, 0.72, 0.91, 0.87 and 0.83, respectively, were obtained. AUC were also analyzed to better understand where errors originated. It was seen that background confusion limited performance. The proposed system can facilitate root dilaceration assessment and alleviate the burden of clinicians, especially for endodontists and surgeons.

Prevalence of bifid variations of the mandibular canal in an Iranian population using cone-beam computed tomography.

Bifid mandibular canal (MC) is an anatomical variation of the MC. This study aimed to assess the prevalence and shape of bifid MC in an Iranian population. A total of 681 patients who had undergone cone-beam computed tomography (CBCT) for different purposes between 2018 and 2020 were evaluated. After detection, bifid MCs were classified into four types forward, buccolingual, dental, and retromolar. CBCT images were assessed by two oral and maxillofacial radiologists. Data were analyzed by SPSS using an independent t-test and Chi-square test. Bifid MC was found in 23 (3.4%) out of 681 patients, with a mean age of 32.21years. Ten patients (1.5%) had a bifid MC on the right side, 6 (0.9%) on the left side, and 7 (1%) bilaterally. However, no significant correlation was found between laterality and the prevalence of bifid MC (P > 0.05). Bifid MC was found in 8 males (34.8%) and 15 females (65.2%). Gender had no significant correlation with the prevalence of bifid MC (P > 0.05). Forward type was the most common (n = 8, 1.2%) followed by buccolingual (n = 5, 0.73%), dental (n = 2, 0.3%), and retromolar (n = 1, 0.14%) types. According to the present results, bifid MC was not uncommon in the Iranian population of the present study, and forward type was the most common, followed by buccal and then dental bifid MCs. There was no significant correlation between sex and age with bifid MC but bifid MC was detected more frequently in females than males, and it was seen unilaterally in a higher percentage of the cases.

Cone Beam CT Features and Oral Radiologist's Decision-making of Arrested Pneumatization of the Sphenoid Sinus.

To assess the demographic and radiographic features of arrested pneumatization of the sphenoid sinus (APS) and their influence on the confidence of oral and maxillofacial radiologists (OMFRs) in diagnosing APS. Reports of cone beam computed tomography (CBCT) APS were retrieved, and the demographic and radiographic features were retrospectively analyzed. Five OMFRs assessed the CBCT images and their confidence in diagnosing APS. The OMFRs' experience (years), expertise (skull-base CBCT cases/month) and diagnostic confidence level were analyzed for agreement and associations with demographic or radiographic features. Of 29 APS cases, 17 (58.6%) were females, and the mean age was 29.9±19 years. Twenty cases (69.0%) presented unilaterally, and 27 (93.1%) involved the sphenoid body. The most common accessory site was the pterygoid process (19, 65.5%). The vidian canal and foramen rotundum were involved in 27 (93.1%) and 17 (58.6%) cases, respectively. Most cases (28, 96.6%) were well-defined, corticated, and showed mixed attenuation. APS diagnostic confidence was higher among the expert OMFRs (72.4%-82.8% vs. 58.6%-62.1%). Radiographic features differentiating APS from skull-base tumors were shown on CBCT. The confidence of OMFRs with similar experience in years depended on their frequency of examining CBCT cases involving the skull base.

CBCT Evaluation of Lingual Foramen and its Anatomic Variations in Northeast Indian Population.

The identification of nearby neurovascular structures is a vital part of dental implant treatment planning. Commonly two implants are placed in the mandibular inter foraminal region to support the overdenture. Placing dental implants in this region is considered relatively safe. The purpose of this study was to employ cone beam computed tomography (CBCT) on a representative sample of Northeast Indians to find out the location and distribution of lingual foramen in the anterior mandible. A total of 104 CBCT scans were performed using the Cranex 3D CBCT system. Three Oral and Maxillofacial Radiologists agreed on a visibility grade for assessing the lingual foramen using a 3-point scale. The number and presence of foramina were noted. On cross-sectional pictures, the distance in millimeters between the crest of the alveolar ridge and the superior cortical margin of the foramen was determined. Furthermore, the analysis of the distribution of lingual foramina in terms of frequency and percentages was performed. Scan results showed that the lingual foramen was clearly visible in 65.4% of cases, somewhat visible in 12.5% of scans, and not visible at all in 22.1% of cases. Half of the people in the study had two foramina, whereas 27.9 percent had one and 22.1 percent had none. The mean distance from the crest of the alveolar ridge was 10.84 mm in males whereas 9.91 mm in females. The Northeast Indian population exhibits an unusually high prevalence of lingual foramen in the anterior portion of the mandible. This research will aid in the precise pre-operative identification of the position and number of lingual foramina prior to surgical treatments involving the midline of the mandible.

Case – Calcification of the epiglottis

This 70-year-old male patient presented to the Department of Oral and Maxillofacial Radiology for CBCT imaging prior to implant therapy (Figure 1)

Panoramik Radyograflar İnterproksimal Çürük Tanısında Ne Kadar Kullanışlıdır? Diş Hekimliği Öğrencileri ve Diş Hekimleriyle Yapılan Bir Çalışma

Purpose: It was aimed to investigate the usability of panoramic radiographs without bitewing radiographs in the diagnosis of interproximal caries and to compare the performance of trainee dentists and dentists in the diagnosis of interproximal caries on panoramic radiographs.
 Material and Method: 20 4th grade, 20 5th grade students studying at Erciyes University Faculty of Dentistry and 20 general dentists were included in the study. In the study, images of 11 individuals who had both panoramic and bitewing taken on the same day within the indication in 2020 were used. Initially, radiographs were evaluated by three Oral, Dental and Maxillofacial Radiology research assistants, and caries on the interproximal surfaces of posterior teeth were recorded with a consensus on bitewing radiographs. Second, the participants who agreed to participate in the study evaluated only the panoramic radiographs and coded the lesions at the interfaces of the premolar and molar teeth as “0”, “1”, “2”, and “3” according to their depth. Statistical analyzes were performed with SPSS v.22 software. Results: Dentists were the most successful in terms of correct evaluation of teeth in terms of the presence or absence of caries (80.52%). This was followed by class 5 (67.29%) and class 4 (60.12%) students. (p

Classification of temporomandibular joint osteoarthritis on cone beam computed tomography images using artificial intelligence system.

The use of artificial intelligence has many advantages, especially in the field of oral and maxillofacial radiology. Early diagnosis of temporomandibular joint osteoarthritis by artificial intelligence may improve prognosis. The aim of this study is to perform the classification of temporomandibular joint (TMJ) osteoarthritis and TMJ segmentation on cone beam computed tomography (CBCT) sagittal images with artificial intelligence. In this study, the success of YOLOv5 architecture, an artificial intelligence model, in TMJ segmentation and osteoarthritis classification was evaluated on 2000 sagittal sections (500 healthy, 500 erosion, 500 osteophyte, 500 flattening images) obtained from CBCT DICOM images of 290 patients. The sensitivity, precision and F1 scores of the model for TMJ osteoarthritis classification are 1, 0.7678 and 0.8686, respectively. The accuracy value for classification is 0.7678. The prediction values of the classification model are 88% for healthy joints, 70% for flattened joints, 95% for joints with erosion and 86% for joints with osteophytes. The sensitivity, precision and F1 score of the YOLOv5 model for TMJ segmentation are 1, 0.9953 and 0.9976, respectively. The AUC value of the model for TMJ segmentation is 0.9723. In addition, the accuracy value of the model for TMJ segmentation was found to be 0.9953. Artificial intelligence model applied in this study can be a support method that will save time and convenience for physicians in the diagnosis of the disease with successful results in TMJ segmentation and osteoarthritis classification.

The Readiness of ChatGPT to Write Scientific Case Reports Independently: A Comparative Evaluation Between Human and Artificial Intelligence.

The use of artificial intelligence (AI) in scientific publishing has gained increased attention, and one AI tool that has been the subject of much discussion isChatGPT. It is a large languagemodel(LLM) built on the OpenAI platform that aims to emulate human-like writing and continually improves through user interactions. In this paper, ChatGPT's performance was assessed in medical publishing by comparing it to a case report written by oral and maxillofacial radiologists. ChatGPT was tasked with writing a case report based on a drafted report written by the authors in five different prompts.Thefindings of this study highlight issues related to the accuracy, completeness, and readability of the generated text. These results have significant implications for the future use of AI in scientific publishing and suggestthat in the current iteration of ChatGPT, scientific information must be revised by an expert.

Oral and Maxillofacial Radiology

Clinically (Figure 1) a unilateral swelling, proptosis and obliteration of the nasolabial fold was noted. Intraoral examination revealed normal-appearing overlying mucosa. A pantomograph (Figure 2) demonstrates a mixed diffuse expansile lesion and thinning of the cortices affecting both jaws. 3D reconstruction (Figure 3) overview the lesions' extent. CBCT interpretation (Figures 4 and 5) indicated engrossment of the frontal, parietal, temporal, sphenoid, ethmoid, maxillary, palatine, zygomatic, and mastoid bones. T1-weighted gadolinium-enhanced MRI image (Figure 6) of a patient with a similar lesion in the right maxilla demonstrates a heterogeneous appearance.

Cone-Beam Computed Tomography Assessment of Quality of Endodontic Treatment and Prevalence of Procedural Errors in Mandibular Molars.

This study assessed the quality of endodontic treatment and the prevalence of procedural errors in permanent mandibular molars using cone-beam computed tomography (CBCT). This cross-sectional study was conducted on 328 CBCT scans (182 females and 146 males) of endodontically treated mandibular molars retrieved from the archives of two radiology centers in Ardabil city, Iran, in 2019. Mandibular molars were evaluated on sagittal, coronal, and axial sections regarding obturation length, obturation density (voids), missed canals, broken instruments, apical perforation, strip perforation, ledge formation, transportation, root fracture, root resorption, and periapical lesions by a senior dental student under the supervision of an oral and maxillofacial radiologist and an endodontist. Differences between the frequency of procedural errors and tooth type and gender were analyzed by the chi-square test. The frequency of underfilling, missed canals, overfilling, voids, apical perforation, transportation, ledge formation, broken instruments, root fracture, strip perforation, root resorption, and periapical lesions was 34.8%, 17.4%, 16.8%, 14.3%, 7.3%, 6.1%, 4.3%, 3%, 1.2%, 0.6%, 5.5%, and 46%, respectively. The frequency of root fracture was significantly higher in females than in males (P < 0.05). The prevalence of underfilling was the highest in right second molars (47.2%), followed by right first molars, left second molars, and left first molars (P < 0.005). The frequency of transportation was maximum in right first molars (10%), followed by right second molars, left first molars, and left second molars (P < 0.04). Underfilling, missed canals, and overfilling were the most prevalent procedural errors in mandibular molars in our study population.

Morphological Assessment of Nasopalatine Canal Using Cone Beam Computed Tomography: A Retrospective Study of 124 Consecutive Patients.

This study aimed to assess and analyze the morphology and dimensions of the nasopalatine canal (NPC), as well as the adjacent buccal osseous plate (BOP), and to investigate the effect of gender, edentulism, NPC types, absence of maxillary central incisors (ACI) and age on the NPC and BOP, using cone beam computed tomography (CBCT). A total of 124 CBCT examinations (67 female and 57 male patients) were retrospectively included and evaluated. The assessment of the dimensions of the NPC, as well as the dimensions of the adjacent BOP, was performed by three Oral and Maxillofacial Radiologists on reconstructed sagittal and coronal CBCT sections under standardized conditions. Regarding the dimensions of the NPC and the adjacent BOP, the mean values were significantly higher among males than females. Furthermore, edentulous patients showed a significant reduction in BOP dimensions. Additionally, NPC types showed a significant effect on the length of the NPC, and the ACI had a significant effect on reducing BOP dimensions. Age had a significant effect on the diameter of the incisive foramen, with the mean values generally increasing with an increasing age. CBCT imaging of this anatomical structure contributes significantly to its full assessment.

Deep learning for detection and 3D segmentation of maxillofacial bone lesions in cone beam CT.

To develop an automated deep-learning algorithm for detection and 3D segmentation of incidental bone lesions in maxillofacial CBCT scans. The dataset included 82 cone beam CT (CBCT) scans, 41 with histologically confirmed benign bone lesions (BL) and 41 control scans (without lesions), obtained using three CBCT devices with diverse imaging protocols. Lesions were marked in all axial slices by experienced maxillofacial radiologists. All cases were divided into sub-datasets: training (20,214 axial images), validation (4530 axial images), and testing (6795 axial images). A Mask-RCNN algorithm segmented the bone lesions in each axial slice. Analysis of sequential slices was used for improving the Mask-RCNN performance and classifying each CBCT scan as containing bone lesions or not. Finally, the algorithm generated 3D segmentations of the lesions and calculated their volumes. The algorithm correctly classified all CBCT cases as containing bone lesions or not, with an accuracy of 100%. The algorithm detected the bone lesion in axial images with high sensitivity (95.9%) and high precision (98.9%) with an average dice coefficient of 83.5%. The developed algorithm detected and segmented bone lesions in CBCT scans with high accuracy and may serve as a computerized tool for detecting incidental bone lesions in CBCT imaging. Our novel deep-learning algorithm detects incidental hypodense bone lesions in cone beam CT scans, using various imaging devices and protocols. This algorithm may reduce patients' morbidity and mortality, particularly since currently, cone beam CT interpretation is not always preformed. • A deep learning algorithm was developed for automatic detection and 3D segmentation of various maxillofacial bone lesions in CBCT scans, irrespective of the CBCT device or the scanning protocol. • The developed algorithm can detect incidental jaw lesions with high accuracy, generates a 3D segmentation of the lesion, and calculates the lesion volume.

Evaluation of The Prevalence of Systemic Disease in Patients Admitted To The Oral And Maxillofacial Radiology Clinic

Amaç: Bu çalışmanın amacı, diş hekimliği fakültesi hastanesine başvuran hastaların sistemik hastalık prevalansının belirlenmesi ve medikal profilinin değerlendirilmesidir. Gereç ve Yöntem: Bu retrospektif çalışmaya, Diş Hekimliği Fakültesi Ağız, Diş ve Çene Radyoloji kliniğine Mart-Eylül 2022 tarihleri arasında çeşitli nedenlerle muayene için başvuran hastalar dahil edildi. Hastaların yaş, cinsiyet gibi demografik bilgileri ve sistemik hastalıkları belirlendi. Sistemik hastalık görülme sıklığı ile cinsiyet ve yaş arasındaki farklılık ki-kare testi ile değerlendirildi. Bulgular: Çalışmada, 14-89 yaş aralığında, 2007 kadın ve 1284 erkek olmak üzere toplam 3291 hasta değerlendirildi. Çalışmaya dahil edilen bireylerden %71.77’sinde (n=2362) sistemik hastalık gözlenmezken, %28.23’ünde (n=929) en az bir tane sistemik hastalık belirlendi. Sistemik hastalık görülme sıklığı, cinsiyetler ve yaş grupları arasında istatiksel anlamda farklılık gösterdi (p&amp;lt;0.05). Sonuç: Çalışma sonuçlarına göre en sık gözlenen hastalık kardiyo-vasküler ve endokrin sistem hastalıklarıdır. Diş hekimleri, bu hastalıklarla ilgili oluşabilecek komplikasyonları önlemek için muayene ve dental tedavi öncesinde ayrıntılı bir anamnez almalıdır.

Refinement of image quality in panoramic radiography using a generative adversarial network.

We aimed to develop and assess the clinical usefulness of a generative adversarial network (GAN) model for improving image quality in panoramic radiography. Panoramic radiographs obtained at Yonsei University Dental Hospital were randomly selected for study inclusion (n = 100). Datasets with degraded image quality (n = 400) were prepared using four different processing methods: blur, noise, blur with noise, and blur in the anterior teeth region. The images were distributed to the training and test datasets in a ratio of 9:1 for each group. The Pix2Pix GAN model was trained using pairs of the original and degraded image datasets for 100 epochs. The peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM) were obtained for the test dataset, and two oral and maxillofacial radiologists rated the quality of clinical images. Among the degraded images, the GAN model enabled the greatest improvement in those with blur in the region of the anterior teeth but was least effective in improving images exhibiting blur with noise (PSNR, 36.27 > 32.74; SSIM, 0.90 > 0.82). While the mean clinical image quality score of the original radiographs was 44.6 out of 46.0, the highest and lowest predicted scores were observed in the blur (45.2) and noise (36.0) groups. The GAN model developed in this study has the potential to improve panoramic radiographs with degraded image quality, both quantitatively and qualitatively. As the model performs better in refining blurred images, further research is required to identify the most effective methods for handling noisy images.

Norman John David Smith (1931-2010): radiologist, master mariner and instigator of dental distance-learning courses.

David Smith was a major influence on the world of dental radiology, helping to found the European Academy of Dental Radiology. He was a president of both the British Society of Dental Radiology and the British Society of Dental and Maxillofacial Radiology, and an honourary life member of the European Academy of Dentomaxillofacial Radiology. David was also a master mariner, politician and fighter for the introduction of distance-learning programmes in dental education.

Https://scholarhub.ui.ac.id/jdi/vol30/iss1/6/

Ramsay Hunt syndrome (RHS) is a disease that is caused by the varicella-zoster virus and is characterized by severe ear pain, auricular vesicular eruptions, and peripheral facial paralysis. Objective: The aim of this case report is to provide information about the clinical findings and treatment process of RHS, which is a rare case and may have oral findings and stress the importance of early diagnosis. Case Report: A 60-year-old male patient had previously consulted an otolaryngologist and a family physician with complaints of vesicular eruptions in the right ear auricle and on the mandible. The patient in whom a diagnosis could not be established presented to the Department of Oral and Maxillofacial Radiology after exacerbated lesions. White plaque-like and ruptured vesicular lesions were observed in the intraoral examination. All vesicular lesions were on one side of the face, and the patient was referred to the dermatology clinic with the diagnosis of RHS. Facial paralysis fully recovered in a short time after early diagnosis and treatment. It should be kept in mind that there may also be oral findings in RHS, and a patient’s intraoral and extraoral examination findings should be evaluated together. Conclusion: Early diagnosis and treatment are highly important in preventing complications such as permanent facial paralysis, vestibulocochlear dysfunction, and hearing loss

Detecting Pulp Stones with Automatic Deep Learning in Bitewing Radiographs: A Pilot Study of Artificial Intelligence

Purpose: This study aims to examine the diagnostic performance of detecting pulp stones with a deep learning model on bite-wing radiographs. Material and Methods: 2203 radiographs were scanned retrospectively. 1745 pulp stones were marked on 1269 bite-wing radiographs with the CranioCatch labeling program (CranioCatch, Eskişehir, Turkey) in patients over 16 years old after the consensus of two experts of Maxillofacial Radiologists. This dataset was divided into 3 grou as training (n = 1017 (1396 labels), validation (n = 126 (174 labels)) and test (n = 126) (175 labels) sets, respectively. The deep learning model was developed using Mask R-CNN architecture. A confusion matrix was used to evaluate the success of the model. Results: The results of precision, sensitivity, and F1 obtained using the Mask R-CNN architecture in the test dataset were found to be 0.9115, 0.8879, and 0.8995, respectively. Discussion- Conclusion: Deep learning algorithms can detect pulp stones. With this, clinicians can use software systems based on artificial intelligence as a diagnostic support system. Mask R-CNN architecture can be used for pulp stone detection with approximately 90% sensitivity. The larger data sets increase the accuracy of deep learning systems. More studies are needed to increase the success rates of deep learning models.