Pseudo labelling workflow, margin losses, hard triplet mining, and PENViT backbone for explainable age and biological gender estimation using dental panoramic radiographs

SN Applied Sciences, Journal Year: 2023, Volume and Issue: 5(11)

Published: Oct. 5, 2023

Abstract This study aimed to estimate human age and gender from panoramic radiographs using various deep learning techniques while explainability have a novel hybrid unsupervised model explain the decision-making process. The classification task involved training neural networks vision transformers on 706 different loss functions backbone architectures namely ArcFace, triplet network named TriplePENViT, subsequently developed called PENViT. Pseudo labeling were applied train models unlabeled data. FullGrad Explainable AI was used gain insights into process of PENViT model. ViT Large 32 achieved validation accuracy 68.21% without demonstrating its effectiveness in task. outperformed other backbones, achieving same ArcFace an improved 70.54% with ArcFace. TriplePENViT 67.44% hard mining techniques. yielded poor performance, 64.34%. Validation established at for Age 84.49% gender. considered developing tooth buds, proximity mandibular shape estimating within deciduous mixed dentitions. For ages 20–29, it factored permanent dentition, alveolar bone density, root apices, third molars. Above 30, notes occlusal deformity resulting missing dentition temporomandibular joint complex as predictors estimation radiographs. Graphical abstract

Language: Английский

---

Clinical Annotation and Segmentation Tool (CAST) Implementation for Dental Diagnostics

Cureus, Journal Year: 2023, Volume and Issue: unknown

Published: Nov. 13, 2023

Purpose This study aims to document the early stages of development an unsupervised, deep learning-based clinical annotation and segmentation tool (CAST) capable isolating clinically significant teeth in both intraoral photographs their corresponding oral radiographs. Methods The dataset consisted 172 424 dental radiographs, manually annotated by two operators, augmented yield 6258 images for training, 183 validation, 98 testing. training involved use object detection model ('YOLOv8') combined with a feature extraction system ('Segment Anything Model'). combination enabled auto-annotation tooth-related features lesions types without operator intervention. Outputs were further processed using data relabelling ('X-AnyLabeling') enabling option reannotate erroneous outputs through reinforcement learning. Results trained achieved mean average precision (mAP) 77.4%, recall rates 75.0% 72.1%, respectively. was able segment from polygonal boundaries better than radiological bounding boxes. Conclusion showed initial promise automating image labelling process Further work is required address limitations.

Language: Английский

---

A 3D printed electronic wearable device to generate vertical, horizontal and phono-articulatory jaw movement parameters: A concept implementation

PLoS ONE, Journal Year: 2023, Volume and Issue: 18(9), P. e0290497 - e0290497

Published: Sept. 13, 2023

The current research aimed to develop a concept open-source 3D printable, electronic wearable head gear record jaw movement parameters.A printed device was designed and manufactured then fitted with sensors vertical, horizontal phono-articulatory motions. Mean deviation relative error were measured invitro. implemented on two volunteers for the parameters of maximum anterior protrusion (MAP), lateral excursion (MLE), normal (NMO), (MMO) mouth opening fricative phono-articulation. Raw data normalized using z-score root mean squared (RMSE) values used evaluate differences in readings across participants.RMSE left right piezoresistive demonstrated near similar bilateral movements during (0.12) maximal (0.09) participant 1, while varying greatly 2 (0.25 0.14, respectively). There larger RMSE accelerometric motion different axes MAP, MLE Fricatives.The implementation that sensor technology can horizontal, maxillomandibular participants. However, future efforts must be made overcome limitations documented within experiment.

Language: Английский

---

Penggunaan smartphone photography pada mobile teledentistry untuk pemeriksaan karies pada survey epidemiologi: systematic review

Padjadjaran Journal of Dental Researchers and Students, Journal Year: 2023, Volume and Issue: 7(2), P. 183 - 183

Published: Aug. 11, 2023

ABSTRAK Pendahuluan: Pemeriksaan klinis secara langsung telah menjadi gold standard untuk skrining karies; namun, karena keterbatasannya, teledentistry mulai dikembangkan sebagai metode alternatif pemeriksaan karies jarak jauh. Mobile merupakan salah satu pengembangan dari telemedicine yang memanfaatkan teknologi smartphone photography mudah digunakan, relatif murah, dan dapat dibawa kemana saja. Tujuan ulasan sistematis ini adalah melihat bagaimana penggunaan pada mobile alat survei epidemiologi, serta apakah akurat reliable survey epidemiologi. Metode: Pengumpulan data dalam penulisan sesuai dengan pedoman PRISMA statement . Penjaringan artikel berbagai literatur diterbitkan 10 tahun terakhir, dicari melalui database PubMed, ProQuest, Science Direct Pemilihan disesuaikan formula PICOS. Uji kelayakan kualitas studi penelitian diinklusikan dilakukan menggunakan kriteria CEBM Diagnostic Accuracy Hasil: Terjaring 6 total 242 didapatkan terkait gigi mulut. Penelitian menunjukkan tingkat sensitivitas, spesifisitas, akurasi tinggi (>80%) reliabilitas cukup baik (Kappa>0,81). memiliki kelemahan diantaranya variasi foto intra oral ketidakmampuan mendeteksi lokasi tertentu. Simpulan: Smartphone sensitif diandalkan Beberapa faktor, seperti teknik pengambilan foto, pencahayaan kemampuan operator perlu diperhatikan mempengaruhi hasil karies. KATA KUNCI: , karies, The use of on for caries examination in epidemiological surveys: a systematic review ABSTRACT Introduction: clinical has been the screening dental caries; however, due to its various limitations, developed as an alternative method remote examination. is one developments that employ technology because it user-friendly, reasonably affordable, and portable. purpose this see how used tool surveys, whether accurate survey. Methods: Data collection was according guidelines. Articles were screened from literature published last years, which searched through databases. selection articles adjusted PICOS formula. feasibility quality tests included research carried out using criteria research. Results: There six publications search results related screening. studies showed high levels sensitivity, specificity accuracy well fairly good reliability (Kappa>0.81). Caries weaknesses, such variations intra-oral photos inability detect certain locations. Conclusion: sensitive surveys. However, several factors technique, light camera operator’s skills need be taken into consideration they can affect KEY WORDS : photography, teledentistry, examination,

---

A Basic Study for Predicting Dysphagia in Panoramic X-ray Images Using Artificial Intelligence (AI) Part 2: Analysis of the Position of the Hyoid Bone on Panoramic Radiographs

Eng—Advances in Engineering, Journal Year: 2023, Volume and Issue: 4(4), P. 2542 - 2552

Published: Oct. 10, 2023

Background: Oral frailty is associated with systemic frailty. The vertical position of the hyoid bone important when considering risk dysphagia. However, dentists usually do not focus on this position. Purpose: To create an AI model for detection bone. Methods: In study, 1830 images from 915 panoramic radiographs were used learning. was classified into six types (Types 0, 1, 2, 3, 4, and 5) based same criteria as in our previous study. Plan 1 learned all types. five other than Type 0 learned. reduce number groupings, three classes formed using combinations two each class. 3 learning classes, 4 Class A 1). Precision, recall, f-values, accuracy, areas under precision–recall curves (PR-AUCs) calculated comparatively evaluated. Results: showed highest accuracy PR-AUC values, 0.93 0.97, respectively. Conclusions: By reducing cases which anatomical structure partially invisible, correctly detected.

Language: Английский

---

Informatic tools for diagnosis in dentistry. A compilation review

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: May 6, 2024

ABSTRACT The objective of this study was to compile the computer tools available in scientific literature aimed at diagnosis dentistry. A scoping review conducted using PubMed , Scopus and Web Science . Were include, original researches type articles, Articles that reported usefulness a computer/technological tool helps dental practice, published last 20 years (period 2004-2024) written English Spanish. Online Rayyan® used establish homogeneity authors on single online platform where they had access could centralize results. Variables were extracted from articles included study. In total, 12648 records retrieved database. After decantation, 39 reports described 36 for More informatic related "Restorative Dentistry’ have been developed than rest specialties 14 (40%). Python predominant programming language, 83.3% validated, 27.8% free. Informatics dentistry enhance treatment planning. However, robust regulatory framework is required validation prior clinical implementation. Continuous training professionals these technologies crucial maximize their benefits ensure optimal patient care. research needed explore potential informatics applications dentistry, integration into existing health systems, accessibility resource-limited areas.

Language: Английский

---

Automated detection and segmentation of dental caries using a novel cascaded learning approach

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 102, P. 107344 - 107344

Published: Dec. 12, 2024

Language: Английский

---

Artificial Intelligence Platforms in Dental Caries Detection: A Systematic Review and Meta-Analysis

Journal of Evidence Based Dental Practice, Journal Year: 2024, Volume and Issue: 25(1), P. 102077 - 102077

Published: Dec. 12, 2024

To assess Artificial Intelligence (AI) platforms, machine learning methodologies and associated accuracies used in detecting dental caries from clinical images radiographs. A systematic search of 8 distinct electronic databases: Scopus, Web Science, MEDLINE, Educational Resources Information Centre, Institute Electrical Electronics Engineers Explore, Science Direct, Directory Open Access Journals JSTOR, was conducted January 2000 to March 2024. AI studies using for detection were extracted along with essential study characteristics. The quality included assessed QUADAS-2 the CLAIM checklist. Meta-analysis performed obtain a quantitative estimate accuracy. Of 2538 identified, 45 met inclusion criteria underwent qualitative synthesis. studies, 33 radiographs, 12 as datasets. total 21 different platforms reported. accuracy ranged 41.5% 98.6% across reported platforms. meta-analysis 7 mean sensitivity 76% [95% CI (65% - 85%)] specificity 91% [(95% (86% 95%)]. area under curve (AUC) 92% (89% 94%)], high heterogeneity studies. Significant variability exists performance demonstrates that has superior equal compared bitewing radiography. Although is promising detection, further refinement necessary achieve consistent reliable varying imaging modalities.

Language: Английский

---

Examination of new clinical dental caries in school children using real intra oral photos with artificial intelligence model YOLO-V8x

Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 23, 2024

Background Dental caries is one of the most common chronic diseases in school-aged children, with a prevalence above 80% Indonesia. Traditional diagnostic practices are time-consuming and dependent on number healthcare professionals available, as such, have led to creation AI-based alternatives, such HI Bogi application. This research used YOLO-v8 model assist detection dental caries, which faster, more efficient, highly accurate, can be developed enhance existing health programs schools IndonesiaMaterials Methods A dataset 3,221 JPG images labeled using ICDAS D0 D1 method was prepared processed Roboflow labeling software, were resized 640×640 pixels standardize input for training. The divided into training (2,266 images), validation (635 testing (320 images) subsets. YOLOv8x algorithm deep learning performance evaluated confusion matrix analysis calculate True Positive (TP), False (FP), Negative (FN) values. Statistical Mann–Whitney tests conducted compare classification accuracy between AI dentists across categories, whereas speed test assessed efficiency relative dentists.Results showed encouraging results, recall 41.1%, 72.6%, mAP 45.8%. P-values 0.301 D1, 0.690 D2, 0.621 D3, 0.693 D4, 0.634 D5, 0.302 D6 obtained from comparative dentists. These results no significant variations sensitivity, specificity, PPV, or NPV among categories (p > 0.05). Furthermore, performed much faster than during examinations = 0.000). may improve efficacy early diagnosis demonstrated by these findings.Conclusion integrated application promising detection, comparable those all criteria. significantly outperformed terms examination completed tasks four times faster. Future should explore transformer-based models expand datasets ability identify diverse including rare lesions.

Language: Английский

---

Using the Case-Based Reasoning-YOLO Approach for Rapid and Effective Identification of Bruises

Published: Sept. 12, 2023

Bruises can appear when blood vessels rupture, which lead to the risk of leakage into surrounding tissues.Evaluation and detection these symptoms, especially those related health problems or accidents, are very important in medical environments.Bruises also serve as an alert sign that a evaluation is recommended might be urgently needed.Unfortunately, it challenging for practitioners appropriately identify categorize bruises due complexity situations many types bruises.The main goal this study promote use Artificial Intelligence (AI) healthcare systems.It aims help improve computer-aided practices by making open-source algorithm such YOLOv8 incorporate case-based reasoning (CBR) approach fast precise identification bruises.In study, we introduce problem using CBR-YOLO approach.The support decision-making practice.Although have same appearance, still provide recommendations commentary on bruises.This method useful diagnosing patients timely manner.

Language: Английский

---