ISSN :2582-9793
Tara Azimi, Mahmood Dashti, Farshad Khosraviani, Mohammad Hosein Amirzade-Iranaq, Neda Tajbakhs, Seyede Fateme Rezaei Taleshi and Sofia Ani
Adv. Artif. Intell. Mach. Learn., 4 (4):3173-3185
Tara Azimi : Ucla school of dentistry, Orofacial Pain and Dysfunction department
Mahmood Dashti : Researcher, Dentofacial Deformities Research Center, Research Institute of Dental Sciences, Shahid Beheshti University of Medical Sciences, Tehran, Iran
Farshad Khosraviani : Researcher, UCLA School of Dentistry, CA, USA
Mohammad Hosein Amirzade-Iranaq : Post Graduate Student, Department of Oral and Maxillofacial Medicine, Isfahan University of Medical Sciences, Isfahan, Iran.
Neda Tajbakhs : Researcher, School of Dentistry, Islamic Azad University Tehran, Dental Branch, Tehran, Iran.
Seyede Fateme Rezaei Taleshi : Researcher, School of Dentistry, Mazandaran University of Medical Sciences, Sari, Iran.
Sofia Ani : International foundation program, Biomedical science, London, UK.
Article History: Received on: 27-Oct-24, Accepted on: 23-Dec-24, Published on: 30-Dec-24
Corresponding Author: Tara Azimi
Email: taraazimi93@g.ucla.edu
Citation: Mahmood Dashti, Farshad Khosraviani , Mohammad Hosein Amirzade-Iranaq, Neda Tajbakhs, Seyede Fateme Rezaei Taleshi , Sofi Ani, Tara Azimi. (USA) (2024). Application of deep learning algorithms in segmentation of mandibular nerve canal in orthopantomogram (panoramic) radiographs: a State-of-Art systematic review. Adv. Artif. Intell. Mach. Learn., 4 (4 ):3173-3185.
Objectives: To assess the current landscape and
efficacy of artificial intelligence and deep learning (DL) algorithms in detecting
and segmenting mandibular canals in orthopantomogram (panoramic) radiographs.
Methods: Research on the detection and
segmentation of the mandibular canal for developing AI models was conducted by
searching five major electronic databases. The PICO question was, “Are 2D
radiographic images suitable for utilizing deep learning algorithms to identify
the infra-alveolar nerve?” The included studies adapted
customized assessment criteria based on QUADAS-2 for quality assessments.
Results: 255 records were identified during
the initial electronic search. After a thorough evaluation, six studies
specifically addressing the detection and segmentation of mandibular canals
were selected for inclusion. Various outcome metrics were reported. The dice
coefficient varies between 0.78 and 0.97 between models. Also, sensitivity
(recall) varies from 0.83 to 0.99, indicating high performance in various DL
models.
Conclusion: The AI models discussed in the
included studies vary in performance. Additionally, the outcome metrics
reported were not consistent, making it difficult to compare all the deep
learning (DL) models comprehensively. The impressive performance of these DL models
should be evaluated using external datasets to compare their effectiveness and
train them to achieve better results.