Panoramic radiographic features for machine learning based detection of mandibular third molar root and inferior alveolar canal contact
Abstract This study uses machine learning (ML) to elucidate the contact relationship between the mandibular third molar (M3M) and the inferior alveolar canal (IAC), leading to three major contributions; (1) The first publicly accessible PR image dataset with semantic annotations for 1,478 IACs and M...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-02-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-024-82915-5 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1823862329791479808 |
|---|---|
| author | A. Canberk Ulusoy Tuğçe Toprak M. Alper Selver Pelin Güneri Betül İlhan |
| author_facet | A. Canberk Ulusoy Tuğçe Toprak M. Alper Selver Pelin Güneri Betül İlhan |
| author_sort | A. Canberk Ulusoy |
| collection | DOAJ |
| description | Abstract This study uses machine learning (ML) to elucidate the contact relationship between the mandibular third molar (M3M) and the inferior alveolar canal (IAC), leading to three major contributions; (1) The first publicly accessible PR image dataset with semantic annotations for 1,478 IACs and M3Ms from 1,010 patients is introduced, which includes challenging cases, such as false positive contacts, with CBCT images as the gold standard, (2) Established radiological indicators for M3M-IAC contact were extracted as features using digital image processing, and these features were used as inputs for various ML methods. Eligibility was assessed through statistical analysis and radiologists evaluations. Clinical feedback from radiologists on these features provides insights for future improvements. (3) ANNs, two custom CNNs, seven established DL models, and their combinations were used for automatic M3M-IAC contact determination with extracted features, semantic annotations, and ROIs. The ANN configuration surpassed both radiologists and DL models in specificity (82%), F1 score (92%), and accuracy (85%), while maintaining a comparable sensitivity (86%) to the DL models. This indicates that ANNs can effectively predict M3M-IAC contact relations and are particularly effective at identifying cases with no contact relation between M3M and IAC compared to other ML methods. Future work should focus on developing automated segmentation algorithms for M3M and IAC on PRs, to identify relevant anatomical structures, thereby improving clinical usability. The dataset, feature extraction, and ML codes are available through the CONTACT grand challenge. |
| format | Article |
| id | doaj-art-ac5ac89c75684b2288cce9de31a67abe |
| institution | Kabale University |
| issn | 2045-2322 |
| language | English |
| publishDate | 2025-02-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj-art-ac5ac89c75684b2288cce9de31a67abe2025-02-09T12:35:04ZengNature PortfolioScientific Reports2045-23222025-02-0115111410.1038/s41598-024-82915-5Panoramic radiographic features for machine learning based detection of mandibular third molar root and inferior alveolar canal contactA. Canberk Ulusoy0Tuğçe Toprak1M. Alper Selver2Pelin Güneri3Betül İlhan4Department of Oral and Maxillofacial Radiology, School of Dentistry, Ege UniversityThe Graduate School of Natural and Applied Sciences and Izmir Vocational School (IMYO), Dokuz Eylül UniversityElectrical and Electronics Engineering Department and Izmir Health Technologies Development and Accelerator (BioIzmir), Dokuz Eylül UniversityDepartment of Oral and Maxillofacial Radiology, School of Dentistry, Ege UniversityDepartment of Oral and Maxillofacial Radiology, School of Dentistry, Ege UniversityAbstract This study uses machine learning (ML) to elucidate the contact relationship between the mandibular third molar (M3M) and the inferior alveolar canal (IAC), leading to three major contributions; (1) The first publicly accessible PR image dataset with semantic annotations for 1,478 IACs and M3Ms from 1,010 patients is introduced, which includes challenging cases, such as false positive contacts, with CBCT images as the gold standard, (2) Established radiological indicators for M3M-IAC contact were extracted as features using digital image processing, and these features were used as inputs for various ML methods. Eligibility was assessed through statistical analysis and radiologists evaluations. Clinical feedback from radiologists on these features provides insights for future improvements. (3) ANNs, two custom CNNs, seven established DL models, and their combinations were used for automatic M3M-IAC contact determination with extracted features, semantic annotations, and ROIs. The ANN configuration surpassed both radiologists and DL models in specificity (82%), F1 score (92%), and accuracy (85%), while maintaining a comparable sensitivity (86%) to the DL models. This indicates that ANNs can effectively predict M3M-IAC contact relations and are particularly effective at identifying cases with no contact relation between M3M and IAC compared to other ML methods. Future work should focus on developing automated segmentation algorithms for M3M and IAC on PRs, to identify relevant anatomical structures, thereby improving clinical usability. The dataset, feature extraction, and ML codes are available through the CONTACT grand challenge.https://doi.org/10.1038/s41598-024-82915-5Machine learningdeep learningCNNANNmandibuler third molarinferior alveolar canal |
| spellingShingle | A. Canberk Ulusoy Tuğçe Toprak M. Alper Selver Pelin Güneri Betül İlhan Panoramic radiographic features for machine learning based detection of mandibular third molar root and inferior alveolar canal contact Scientific Reports Machine learning deep learning CNN ANN mandibuler third molar inferior alveolar canal |
| title | Panoramic radiographic features for machine learning based detection of mandibular third molar root and inferior alveolar canal contact |
| title_full | Panoramic radiographic features for machine learning based detection of mandibular third molar root and inferior alveolar canal contact |
| title_fullStr | Panoramic radiographic features for machine learning based detection of mandibular third molar root and inferior alveolar canal contact |
| title_full_unstemmed | Panoramic radiographic features for machine learning based detection of mandibular third molar root and inferior alveolar canal contact |
| title_short | Panoramic radiographic features for machine learning based detection of mandibular third molar root and inferior alveolar canal contact |
| title_sort | panoramic radiographic features for machine learning based detection of mandibular third molar root and inferior alveolar canal contact |
| topic | Machine learning deep learning CNN ANN mandibuler third molar inferior alveolar canal |
| url | https://doi.org/10.1038/s41598-024-82915-5 |
| work_keys_str_mv | AT acanberkulusoy panoramicradiographicfeaturesformachinelearningbaseddetectionofmandibularthirdmolarrootandinferioralveolarcanalcontact AT tugcetoprak panoramicradiographicfeaturesformachinelearningbaseddetectionofmandibularthirdmolarrootandinferioralveolarcanalcontact AT malperselver panoramicradiographicfeaturesformachinelearningbaseddetectionofmandibularthirdmolarrootandinferioralveolarcanalcontact AT pelinguneri panoramicradiographicfeaturesformachinelearningbaseddetectionofmandibularthirdmolarrootandinferioralveolarcanalcontact AT betulilhan panoramicradiographicfeaturesformachinelearningbaseddetectionofmandibularthirdmolarrootandinferioralveolarcanalcontact |