FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4229053969> ?p ?o ?g. }

• W4229053969 abstract "Full AccessLetter to the Editor-Article CommentaryDeep learning for automated detection and numbering of permanent teeth on panoramic imagesBaris Oguz Gurses, Elif Sener and Pelin GuneriBaris Oguz GursesDepartment of Mechanical Engineering, Faculty of Engineering, Ege University, Izmir, TurkeySearch for more papers by this author, Elif SenerDepartment of Oral and Maxillofacial Radiology, School of Dentistry, Ege University, Izmir, TurkeySearch for more papers by this author and Pelin GuneriDepartment of Oral and Maxillofacial Radiology, School of Dentistry, Ege University, Izmir, TurkeySearch for more papers by this authorPublished Online:4 May 2022https://doi.org/10.1259/dmfr.20220128SectionsPDF/EPUB ToolsAdd to favoritesDownload CitationsTrack Citations ShareShare onFacebookTwitterLinked InEmail AboutDear Editor,We have read the interesting article which is cited as “Estai M, Tennant M, Gebauer D, Brostek A, Vignarajan J, Mehdizadeh M, et al. Deep learning for automated detection and numbering of permanent teeth on panoramic images. Dentomaxillofac Radiol 2022 Feb 1;51(2):20210296.” In this paper, the efficacy of combining three different convolutional neural networks (CNN), namely U-Net, Faster R-CNN and VGG-16 on tooth detection and numbering according to FDI system was investigated after three observers cropped 17,135 teeth on 591 digital orthopanthomographs (OPG). The proposed three-step method has reached a recall of 0.99 and a precision of 0.99 for the tooth detection module, and a recall, precision and F1 score of 0.98 for the tooth numbering module. The study revealed the applicability of their method for automatic tooth detection and numbering for general dentistry and forensic medicine purposes using OPGs.1 The relatively high performance of the proposed combined method was impressive, since it revealed that the efficacy of artificial Intelligence (AI) has been increasing very rapidly in recent decade. As a contribution to the authors study, it is known that AI has been developed to mimic human brain decision-making process via using algorithms,2 and machine learning (ML) and its subset Deep Learning (DL) or convolutional neural networks (CNN) are employed to form AI. Task and data dependence of machine-learning applications on dental diagnosis are frequently criticized amongst dental experts. For task-specific applications, perhaps the focus should be “how we train dental experts” since this is the first and the most important phase of structure segmentation as observed in the referred paper,1 because the detection of the anomaly or the alteration is manageable if there is a reference. Unfortunately, definition of anatomical landmarks or delineation of oral pathologies is not an easy task, because every patient’s tooth or oral cavity is unique and includes variations which are not considered as pathologies. Therefore, training of the experts to appropriately and adequately to define the landmarks on radiographic images of oral cavity is a compulsory step in order to provide a proper environment for the algorithm to detect for anomalies. During the algorithm developing process, overfitting which can be defined as “memorizing the dataset” is the main concern about the applications. Each medical application developer shall consider whether their data cover the whole diagnostic space or it is converged on a single, specific case because the probability of an outliner that escapes the dataset in training is usually concerning. Whole dental machine-learning literature is crowded by “task specific developed classifiers” such as artificial neural networks (ANNs), support vector machines (SVMs), k-nearest neighbor (KNN), gray-level co-occurrence matrix (GLCM), gray-level run length matrix (GLRLM), and gray-level size zone matrix (GLSZM), logistic regression (LR), random forest (RF), decision tree (DT), U-Net, Faster R-CNN and VGG-16 and XGBoost.1,3 Although many these well written and delicately applied classifiers have led continuously increasing success in classifying the proposed conditions, their performances in actual health conditions still need further assessment.4 Prior to starting such machine classifier projects, the regular research question is “will this classifier succeed in performing this definite dental task (detection of dental decays, bone sclerosis or necrosis, TMJ disorders, etc.)?” However, it shall better be noticed that when the sets of healthy and abnormal conditions are statistically distinguishable (for example, because of contrast differentiation in caries/filling detection), the answer is already obvious: the well-known classifiers affirm their talents to cluster highly nonlinear datasets in subsets in the mathematical domain and when there is a significant difference in the subsets, the classifiers tend to find the non-linear surface that separate them and they succeed. In this context, perhaps the research question should be redefined as “could we find explicit features to solve our problem?”. Obviously, this is not the case in kernel-based classifiers such as CNNs. Whole data set is fed into CNNs without any prior feature extraction, however, the size of the data set needed to train CNNs is generally larger than that needed for SVMs or shallow ANNs. Due to the large number of parameters to be tuned, the training time is longer, and the process is costly, as well.2,5 When these constraints are managed, CNNs would probably do the job with success.In short, AI has at least two points to consider prior to accept its’ value in dental clinic tasks: 1) the need to train the experts to define the features of the “diseased/test” and “healthy/control” structure which is the most important step for diagnostic tasks, 2) the shortage of actual data to train the algorithm and to assess its’ applicability in real life/actual conditions. It is believed that the combined models, as reported in the in the referred paper, have contributed to the dental care with respect to smoother and faster dental workflow4 and higher patient satisfaction due to high-quality diagnosis, treatment planning, and predictable treatment outcomes.1,6 However, before accepting AI as the future magical tool in dental care, it shall be considered that AI is still under development to reach to that ultimate glorious title.REFERENCES1. Estai M, , Tennant M, , Gebauer D, , Brostek A, , Vignarajan J, , Mehdizadeh M, , et al.. Deep learning for automated detection and numbering of permanent teeth on panoramic images. Dentomaxillofac Radiol 2022; 51(2): 20210296. doi: https://doi.org/10.1259/dmfr.20210296 Link, Google Scholar2. Adnan N, , Umer F. Understanding deep learning - challenges and prospects. J Pak Med Assoc 2022; 72(Suppl 1): S59–63. doi: https://doi.org/10.47391/JPMA.AKU-12 Medline, Google Scholar3. Orhan K, , Driesen L, , Shujaat S, , Jacobs R, , Chai X. Development and validation of a magnetic resonance imaging-based machine learning model for TMJ pathologies. Biomed Res Int 2021; 2021: 6656773. doi: https://doi.org/10.1155/2021/6656773 Crossref ISI, Google Scholar4. Ahmed N, , Abbasi MS, , Zuberi F, , Qamar W, , Halim MSB, , Maqsood A, , et al.. Artificial intelligence techniques: analysis, application, and outcome in dentistry-A systematic review. Biomed Res Int 2021; 2021: 9751564. doi: https://doi.org/10.1155/2021/9751564 Crossref, Google Scholar5. Umer F, , Habib S, , Adnan N. Application of deep learning in teeth identification tasks on panoramic radiographs. Dentomaxillofac Radiol 2022; 2: 20210504. doi: https://doi.org/10.1259/dmfr.20210504 Google Scholar6. Carrillo-Perez F, , Pecho OE, , Morales JC, , Paravina RD, , Della Bona A, , Ghinea R, , et al.. Applications of artificial intelligence in dentistry: A comprehensive review. J Esthet Restor Dent 2022; 34: 259–80. doi: https://doi.org/10.1111/jerd.12844 Crossref Medline ISI, Google Scholar Previous article FiguresReferencesRelatedDetails Volume 51, Issue 5July 2022 © 2022 The Authors. Published by the British Institute of Radiology History ReceivedApril 07,2022AcceptedApril 13,2022Published onlineMay 04,2022 Metrics Download PDF" @default.
• W4229053969 created "2022-05-08" @default.
• W4229053969 creator A5009719341 @default.
• W4229053969 creator A5038775691 @default.
• W4229053969 creator A5057718462 @default.
• W4229053969 date "2022-05-04" @default.
• W4229053969 modified "2023-09-25" @default.
• W4229053969 title "Deep learning for automated detection and numbering of permanent teeth on panoramic images" @default.
• W4229053969 cites W3176562962 @default.
• W4229053969 cites W3178036972 @default.
• W4229053969 cites W3205574907 @default.
• W4229053969 cites W3215605706 @default.
• W4229053969 cites W4211120445 @default.
• W4229053969 cites W4220995888 @default.
• W4229053969 doi "https://doi.org/10.1259/dmfr.20220128" @default.
• W4229053969 hasPubMedId "https://pubmed.ncbi.nlm.nih.gov/35507741" @default.
• W4229053969 hasPublicationYear "2022" @default.
• W4229053969 type Work @default.
• W4229053969 citedByCount "0" @default.
• W4229053969 crossrefType "journal-article" @default.
• W4229053969 hasAuthorship W4229053969A5009719341 @default.
• W4229053969 hasAuthorship W4229053969A5038775691 @default.
• W4229053969 hasAuthorship W4229053969A5057718462 @default.
• W4229053969 hasConcept C154945302 @default.
• W4229053969 hasConcept C191916993 @default.
• W4229053969 hasConcept C199343813 @default.
• W4229053969 hasConcept C199360897 @default.
• W4229053969 hasConcept C29694066 @default.
• W4229053969 hasConcept C31972630 @default.
• W4229053969 hasConcept C41008148 @default.
• W4229053969 hasConcept C71924100 @default.
• W4229053969 hasConceptScore W4229053969C154945302 @default.
• W4229053969 hasConceptScore W4229053969C191916993 @default.
• W4229053969 hasConceptScore W4229053969C199343813 @default.
• W4229053969 hasConceptScore W4229053969C199360897 @default.
• W4229053969 hasConceptScore W4229053969C29694066 @default.
• W4229053969 hasConceptScore W4229053969C31972630 @default.
• W4229053969 hasConceptScore W4229053969C41008148 @default.
• W4229053969 hasConceptScore W4229053969C71924100 @default.
• W4229053969 hasLocation W42290539691 @default.
• W4229053969 hasLocation W42290539692 @default.
• W4229053969 hasOpenAccess W4229053969 @default.
• W4229053969 hasPrimaryLocation W42290539691 @default.
• W4229053969 hasRelatedWork W1891287906 @default.
• W4229053969 hasRelatedWork W1969923398 @default.
• W4229053969 hasRelatedWork W2036807459 @default.
• W4229053969 hasRelatedWork W2058170566 @default.
• W4229053969 hasRelatedWork W2166024367 @default.
• W4229053969 hasRelatedWork W2229312674 @default.
• W4229053969 hasRelatedWork W258625772 @default.
• W4229053969 hasRelatedWork W2755342338 @default.
• W4229053969 hasRelatedWork W2772917594 @default.
• W4229053969 hasRelatedWork W3116076068 @default.
• W4229053969 isParatext "false" @default.
• W4229053969 isRetracted "false" @default.
• W4229053969 workType "article" @default.