Taseef Hasan Farook, Saif Ahmed, Farah Rashid, Faisal Ahmed Sifat, Preena Sidhu, Pravinkumar Patil, Sumaya Yousuf Zai, Nafij Bin Jamayet, James Dudley, Umer Daood
{"title":"应用三维神经网络和可解释人工智能对下颌磨牙的 ICDAS 检测系统进行分类。","authors":"Taseef Hasan Farook, Saif Ahmed, Farah Rashid, Faisal Ahmed Sifat, Preena Sidhu, Pravinkumar Patil, Sumaya Yousuf Zai, Nafij Bin Jamayet, James Dudley, Umer Daood","doi":"10.1016/j.prosdent.2024.09.014","DOIUrl":null,"url":null,"abstract":"<p><strong>Statement of problem: </strong>Considerable variations exist in cavity preparation methods and approaches. Whether the extent and depth of cavity preparation because of the extent of caries affects the overall accuracy of training deep learning models remains unexplored.</p><p><strong>Purpose: </strong>The purpose of this study was to investigate the difference in 3-dimensionsal (3D) model cavity preparations after International Caries Detection and Assessment System (ICDAS) classification performed by different practitioners and the subsequent influence on the ability of a deep learning model to predict cavity classification.</p><p><strong>Material and methods: </strong>Two operators prepared 56 restorative cavities on simulated mandibular first molars according to 4 ICDAS classifications, followed by 3D scanning and computer-aided design processing. The surface area, virtual volume, Hausdorff distance (HD), and Dice Similarity Coefficients were computed. Multivariate analysis of variance was used to assess cavity size and operator proficiency interactions, and 1-way ANOVA was used to evaluate HD differences across 4 cavity classifications (α=.05). The 3D convolutional neural network (CNN) predicted the ICDAS class, and Saliency Maps explained the decisions of the models.</p><p><strong>Results: </strong>Operator 1 exhibited a cavity preparation surface area of 360.55 ±15.39 mm<sup>2</sup>, and operator 2 recorded 355.24 ±10.79 mm<sup>2</sup>. Volumetric differences showed operator 1 with 440.41 ±35.29 mm<sup>3</sup> and operator 2 with 441.01 ±35.37 mm<sup>3</sup>. Significant interactions (F=2.31, P=.01) between cavity size and operator proficiency were observed. A minimal 0.13 ±0.097 mm variation was noted in overlapping preparations by the 2 operators. The 3D CNN model achieved an accuracy of 94.44% in classifying the ICDAS classes with a 66.67% accuracy when differentiating cavities prepared by the 2 operators.</p><p><strong>Conclusions: </strong>Operator performance discrepancies were evident in the occlusal cavity floor, primarily due to varying cavity depths. Deep learning effectively classified cavity depths from 3D intraoral scans and was less affected by preparation quality or operator skills.</p>","PeriodicalId":16866,"journal":{"name":"Journal of Prosthetic Dentistry","volume":null,"pages":null},"PeriodicalIF":4.3000,"publicationDate":"2024-10-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Application of 3D neural networks and explainable AI to classify ICDAS detection system on mandibular molars.\",\"authors\":\"Taseef Hasan Farook, Saif Ahmed, Farah Rashid, Faisal Ahmed Sifat, Preena Sidhu, Pravinkumar Patil, Sumaya Yousuf Zai, Nafij Bin Jamayet, James Dudley, Umer Daood\",\"doi\":\"10.1016/j.prosdent.2024.09.014\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><strong>Statement of problem: </strong>Considerable variations exist in cavity preparation methods and approaches. Whether the extent and depth of cavity preparation because of the extent of caries affects the overall accuracy of training deep learning models remains unexplored.</p><p><strong>Purpose: </strong>The purpose of this study was to investigate the difference in 3-dimensionsal (3D) model cavity preparations after International Caries Detection and Assessment System (ICDAS) classification performed by different practitioners and the subsequent influence on the ability of a deep learning model to predict cavity classification.</p><p><strong>Material and methods: </strong>Two operators prepared 56 restorative cavities on simulated mandibular first molars according to 4 ICDAS classifications, followed by 3D scanning and computer-aided design processing. The surface area, virtual volume, Hausdorff distance (HD), and Dice Similarity Coefficients were computed. Multivariate analysis of variance was used to assess cavity size and operator proficiency interactions, and 1-way ANOVA was used to evaluate HD differences across 4 cavity classifications (α=.05). The 3D convolutional neural network (CNN) predicted the ICDAS class, and Saliency Maps explained the decisions of the models.</p><p><strong>Results: </strong>Operator 1 exhibited a cavity preparation surface area of 360.55 ±15.39 mm<sup>2</sup>, and operator 2 recorded 355.24 ±10.79 mm<sup>2</sup>. Volumetric differences showed operator 1 with 440.41 ±35.29 mm<sup>3</sup> and operator 2 with 441.01 ±35.37 mm<sup>3</sup>. Significant interactions (F=2.31, P=.01) between cavity size and operator proficiency were observed. A minimal 0.13 ±0.097 mm variation was noted in overlapping preparations by the 2 operators. The 3D CNN model achieved an accuracy of 94.44% in classifying the ICDAS classes with a 66.67% accuracy when differentiating cavities prepared by the 2 operators.</p><p><strong>Conclusions: </strong>Operator performance discrepancies were evident in the occlusal cavity floor, primarily due to varying cavity depths. Deep learning effectively classified cavity depths from 3D intraoral scans and was less affected by preparation quality or operator skills.</p>\",\"PeriodicalId\":16866,\"journal\":{\"name\":\"Journal of Prosthetic Dentistry\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":4.3000,\"publicationDate\":\"2024-10-21\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of Prosthetic Dentistry\",\"FirstCategoryId\":\"3\",\"ListUrlMain\":\"https://doi.org/10.1016/j.prosdent.2024.09.014\",\"RegionNum\":2,\"RegionCategory\":\"医学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"DENTISTRY, ORAL SURGERY & MEDICINE\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Prosthetic Dentistry","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.1016/j.prosdent.2024.09.014","RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"DENTISTRY, ORAL SURGERY & MEDICINE","Score":null,"Total":0}
Application of 3D neural networks and explainable AI to classify ICDAS detection system on mandibular molars.
Statement of problem: Considerable variations exist in cavity preparation methods and approaches. Whether the extent and depth of cavity preparation because of the extent of caries affects the overall accuracy of training deep learning models remains unexplored.
Purpose: The purpose of this study was to investigate the difference in 3-dimensionsal (3D) model cavity preparations after International Caries Detection and Assessment System (ICDAS) classification performed by different practitioners and the subsequent influence on the ability of a deep learning model to predict cavity classification.
Material and methods: Two operators prepared 56 restorative cavities on simulated mandibular first molars according to 4 ICDAS classifications, followed by 3D scanning and computer-aided design processing. The surface area, virtual volume, Hausdorff distance (HD), and Dice Similarity Coefficients were computed. Multivariate analysis of variance was used to assess cavity size and operator proficiency interactions, and 1-way ANOVA was used to evaluate HD differences across 4 cavity classifications (α=.05). The 3D convolutional neural network (CNN) predicted the ICDAS class, and Saliency Maps explained the decisions of the models.
Results: Operator 1 exhibited a cavity preparation surface area of 360.55 ±15.39 mm2, and operator 2 recorded 355.24 ±10.79 mm2. Volumetric differences showed operator 1 with 440.41 ±35.29 mm3 and operator 2 with 441.01 ±35.37 mm3. Significant interactions (F=2.31, P=.01) between cavity size and operator proficiency were observed. A minimal 0.13 ±0.097 mm variation was noted in overlapping preparations by the 2 operators. The 3D CNN model achieved an accuracy of 94.44% in classifying the ICDAS classes with a 66.67% accuracy when differentiating cavities prepared by the 2 operators.
Conclusions: Operator performance discrepancies were evident in the occlusal cavity floor, primarily due to varying cavity depths. Deep learning effectively classified cavity depths from 3D intraoral scans and was less affected by preparation quality or operator skills.
期刊介绍:
The Journal of Prosthetic Dentistry is the leading professional journal devoted exclusively to prosthetic and restorative dentistry. The Journal is the official publication for 24 leading U.S. international prosthodontic organizations. The monthly publication features timely, original peer-reviewed articles on the newest techniques, dental materials, and research findings. The Journal serves prosthodontists and dentists in advanced practice, and features color photos that illustrate many step-by-step procedures. The Journal of Prosthetic Dentistry is included in Index Medicus and CINAHL.