{"id":{"repo_id":"uic","oai_identifier":"oai:figshare.com:article/32991902"},"canonical_url":"https://search.dev.ndltd.org/etd/uic/oai:figshare.com:article/32991902","repository":{"repo_id":"uic","name":"University of Illinois - Chicago","base_url":"https://api.figshare.com/v2/oai"},"display":{"title":"Development of a Fully Automated Al System for Skeletal Maturity Assessment Using Cone Beam CT","abstract":"Skeletal maturity assessment plays a critical role in orthodontic diagnosis and treatment planning, particularly in optimizing the timing of interventions such as myofunctional appliances and orthognathic surgery. Clinicians look to biological markers, such as the cervical vertebrae and the spheno-occipital synchondrosis to perform such assessment, and automation of this process would greatly assist clinicians with both accuracy and efficiency in their daily practice. Our study seeks to develop, test, and validate automated, interpretable deep learning algorithms for assessing and classifying the spheno-occipital synchondrosis (SOS) fusion and cervical vertebrae maturity (CVM) from cone beam computed tomography scans. The total data set consisted of 1200 CBCT scans from orthodontic patients in the Midwest. For SOS, images with indistinct liminal stages were removed, leaving 723 scans remaining. For CVM, scans failing to capture the fourth vertebra or exhibiting ambiguous transitional phases were also removed, leaving 364 scans remaining. Classification of both sites was performed by two orthodontists and an oral and maxillofacial radiologist. The primary deep learning models employed were ResNet, EffNet, and ConvNeXt (with and without attention mechanisms). In addition, several adjunctive models and techniques—including YOLO, SAM, Grad-CAM, knowledge distillation, and demographic conditioning—were incorporated into the development of several frameworks. The highest-performing, fully automated pipeline for SOS fusion staging was ConvNeXt with self-attention, YOLO, and knowledge distillation, achieving an accuracy of 83.05%. For CVM assessment, the highest performing pipeline was ResNet50 with knowledge distillation, which achieved an accuracy of 88.54%. Our results demonstrate the development of the first deep learning framework for SOS classification. In addition, this study introduces a novel pipeline for CVM classification, representing the first fully automated approach capable of staging all six classes without the need for an intermediate landmarking step. These pipelines improve the efficiency and reliability of skeletal maturity assessment while minimizing the labor required from clinicians.","abstract_html":"Skeletal maturity assessment plays a critical role in orthodontic diagnosis and treatment planning, particularly in optimizing the timing of interventions such as myofunctional appliances and orthognathic surgery. Clinicians look to biological markers, such as the cervical vertebrae and the spheno-occipital synchondrosis to perform such assessment, and automation of this process would greatly assist clinicians with both accuracy and efficiency in their daily practice. Our study seeks to develop, test, and validate automated, interpretable deep learning algorithms for assessing and classifying the spheno-occipital synchondrosis (SOS) fusion and cervical vertebrae maturity (CVM) from cone beam computed tomography scans. The total data set consisted of 1200 CBCT scans from orthodontic patients in the Midwest. For SOS, images with indistinct liminal stages were removed, leaving 723 scans remaining. For CVM, scans failing to capture the fourth vertebra or exhibiting ambiguous transitional phases were also removed, leaving 364 scans remaining. Classification of both sites was performed by two orthodontists and an oral and maxillofacial radiologist. The primary deep learning models employed were ResNet, EffNet, and ConvNeXt (with and without attention mechanisms). In addition, several adjunctive models and techniques—including YOLO, SAM, Grad-CAM, knowledge distillation, and demographic conditioning—were incorporated into the development of several frameworks. The highest-performing, fully automated pipeline for SOS fusion staging was ConvNeXt with self-attention, YOLO, and knowledge distillation, achieving an accuracy of 83.05%. For CVM assessment, the highest performing pipeline was ResNet50 with knowledge distillation, which achieved an accuracy of 88.54%. Our results demonstrate the development of the first deep learning framework for SOS classification. In addition, this study introduces a novel pipeline for CVM classification, representing the first fully automated approach capable of staging all six classes without the need for an intermediate landmarking step. These pipelines improve the efficiency and reliability of skeletal maturity assessment while minimizing the labor required from clinicians.","abstract_has_math":false,"creators":["Amanda Nicole Nikho (24399017)"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026-07-15T12:03:11Z","date_published":"2026-07-15T12:03:11Z","updated_at":"2026-07-27T21:33:06Z","subjects":["Orthodontics","Artificial Intelligence","Medical Imaging"],"languages":[],"rights":["In Copyright"],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://doi.org/10.25417/uic.32991902.v1","outbound_label":"DOI","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["Amanda Nicole Nikho (24399017)"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date","label":"Dc Date","values":["2026-07-15T12:03:11Z"]},{"key":"dc:relation","label":"Dc Relation","values":["https://figshare.com/articles/thesis/Development_of_a_Fully_Automated_Al_System_for_Skeletal_Maturity_Assessment_Using_Cone_Beam_CT/32991902"]},{"key":"dc:type","label":"Dc Type","values":["Text","Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Orthodontics","Artificial Intelligence","Medical Imaging"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["In Copyright"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["10.25417/uic.32991902.v1"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["Skeletal maturity assessment plays a critical role in orthodontic diagnosis and treatment planning, particularly in optimizing the timing of interventions such as myofunctional appliances and orthognathic surgery. Clinicians look to biological markers, such as the cervical vertebrae and the spheno-occipital synchondrosis to perform such assessment, and automation of this process would greatly assist clinicians with both accuracy and efficiency in their daily practice. Our study seeks to develop, test, and validate automated, interpretable deep learning algorithms for assessing and classifying the spheno-occipital synchondrosis (SOS) fusion and cervical vertebrae maturity (CVM) from cone beam computed tomography scans. The total data set consisted of 1200 CBCT scans from orthodontic patients in the Midwest. For SOS, images with indistinct liminal stages were removed, leaving 723 scans remaining. For CVM, scans failing to capture the fourth vertebra or exhibiting ambiguous transitional phases were also removed, leaving 364 scans remaining. Classification of both sites was performed by two orthodontists and an oral and maxillofacial radiologist. The primary deep learning models employed were ResNet, EffNet, and ConvNeXt (with and without attention mechanisms). In addition, several adjunctive models and techniques—including YOLO, SAM, Grad-CAM, knowledge distillation, and demographic conditioning—were incorporated into the development of several frameworks. The highest-performing, fully automated pipeline for SOS fusion staging was ConvNeXt with self-attention, YOLO, and knowledge distillation, achieving an accuracy of 83.05%. For CVM assessment, the highest performing pipeline was ResNet50 with knowledge distillation, which achieved an accuracy of 88.54%. Our results demonstrate the development of the first deep learning framework for SOS classification. In addition, this study introduces a novel pipeline for CVM classification, representing the first fully automated approach capable of staging all six classes without the need for an intermediate landmarking step. These pipelines improve the efficiency and reliability of skeletal maturity assessment while minimizing the labor required from clinicians."]},{"key":"dc:title","label":"Title","values":["Development of a Fully Automated Al System for Skeletal Maturity Assessment Using Cone Beam CT"]}]}],"canonical_facts":{"dc:creator":["Amanda Nicole Nikho (24399017)"],"dc:date":["2026-07-15T12:03:11Z"],"dc:description":["Skeletal maturity assessment plays a critical role in orthodontic diagnosis and treatment planning, particularly in optimizing the timing of interventions such as myofunctional appliances and orthognathic surgery. Clinicians look to biological markers, such as the cervical vertebrae and the spheno-occipital synchondrosis to perform such assessment, and automation of this process would greatly assist clinicians with both accuracy and efficiency in their daily practice. Our study seeks to develop, test, and validate automated, interpretable deep learning algorithms for assessing and classifying the spheno-occipital synchondrosis (SOS) fusion and cervical vertebrae maturity (CVM) from cone beam computed tomography scans. The total data set consisted of 1200 CBCT scans from orthodontic patients in the Midwest. For SOS, images with indistinct liminal stages were removed, leaving 723 scans remaining. For CVM, scans failing to capture the fourth vertebra or exhibiting ambiguous transitional phases were also removed, leaving 364 scans remaining. Classification of both sites was performed by two orthodontists and an oral and maxillofacial radiologist. The primary deep learning models employed were ResNet, EffNet, and ConvNeXt (with and without attention mechanisms). In addition, several adjunctive models and techniques—including YOLO, SAM, Grad-CAM, knowledge distillation, and demographic conditioning—were incorporated into the development of several frameworks. The highest-performing, fully automated pipeline for SOS fusion staging was ConvNeXt with self-attention, YOLO, and knowledge distillation, achieving an accuracy of 83.05%. For CVM assessment, the highest performing pipeline was ResNet50 with knowledge distillation, which achieved an accuracy of 88.54%. Our results demonstrate the development of the first deep learning framework for SOS classification. In addition, this study introduces a novel pipeline for CVM classification, representing the first fully automated approach capable of staging all six classes without the need for an intermediate landmarking step. These pipelines improve the efficiency and reliability of skeletal maturity assessment while minimizing the labor required from clinicians."],"dc:identifier":["10.25417/uic.32991902.v1"],"dc:relation":["https://figshare.com/articles/thesis/Development_of_a_Fully_Automated_Al_System_for_Skeletal_Maturity_Assessment_Using_Cone_Beam_CT/32991902"],"dc:rights":["In Copyright"],"dc:subject":["Orthodontics","Artificial Intelligence","Medical Imaging"],"dc:title":["Development of a Fully Automated Al System for Skeletal Maturity Assessment Using Cone Beam CT"],"dc:type":["Text","Thesis"]},"updated_at":"2026-07-27T21:33:06Z"}