Artificial intelligence for maxillofacial prosthodontics: A technological shift in craniofacial rehabilitation- a scoping review

Abstract

Introduction

Artificial intelligence (AI) transforms dentistry and holds considerable promise for maxillofacial prosthodontics (MFP). Applications in imaging, computer-aided design and manufacturing (CAD/CAM), and additive manufacturing are improving diagnosis, treatment planning, and prosthetic rehabilitation for patients with craniofacial abnormalities. Despite advances in materials and digital workflows, challenges remain in achieving optimal accuracy, efficiency, and customisation in prosthetic design. The integration of AI in maxillofacial prosthodontics is still in its early stages. Currently, there is no review detailing the scope, trends, potential, and limitations of AI in this field. A scoping review is therefore necessary to consolidate existing evidence, identify knowledge gaps, and suggest directions for future research and clinical application. This review objective is to systematically map and analyse the current literature on AI in maxillofacial prosthodontics, focusing on its role in craniofacial rehabilitation.

Methods

This scoping review adhered to the methodological framework of Arksey and O'Malley (2005) and was guided by the Joanna Briggs Institute (JBI) Manual for Evidence Synthesis (2020). Reporting complied with PRISMA-ScR (Preferred Reporting Items for Systematic reviews and Meta-Analyses extension for Scoping Reviews) guidelines to ensure clarity and reproducibility. The review was registered with the Open Science Framework (registration number: www.osf.io/3b9jr). Electronic databases, including Medline via PubMed, Scopus, Cochrane Database, Science Direct, Google Scholar, and Semantic Scholar, were searched up to 7 June 2025. Full-text English articles containing the keywords “Artificial Intelligence and Maxillofacial Prosthodontics” and related terms were included.

Results

This scoping review included 35 articles from diverse geographic regions. The studies addressed several specific applications of AI in maxillofacial prosthodontics, including the production of implant-supported auricular prostheses, coloration of maxillofacial prostheses, evaluation of facial attractiveness in patients with clefts, capture of 3D impressions of cleft palates, identification of hypernasality, assessment of lip symmetry, and detection of teeth in cleft lip and palate cases.

Conclusion

Artificial intelligence offers significant opportunities for maxillofacial prosthodontics, especially in imaging, digital design, and prosthesis production. Progress in this area requires interdisciplinary teamwork, large-scale clinical trials, and the development of standardized validation methods to ensure safe and effective clinical application.

Graphical abstract

1. Introduction

Artificial intelligence (AI) pertains to machines showing intelligent behaviour.¹ It is described as a scientific and engineering field dedicated to understanding and replicating intelligent actions through computation.² It has created a new paradigm that greatly influences numerous areas, including science, technology, and daily life.³ These systems imitate human cognitive functions and intellectual activities, such as problem-solving.^1,4

Artificial intelligence is utilized in various dental specialities, including maxillofacial radiology, orthodontics, prosthodontics, and dental implantology.^5,6 These systems help in diagnosing a wide range of dental conditions, such as caries, root fractures, maxillofacial cysts, salivary gland disorders, maxillary sinusitis, osteoporosis, cancerous lesions, lymph node metastases, and alveolar bone loss.^1,7, 8, 9, 10 Recent studies show that AI can outperform human dentists in detecting dental caries from radiographs, achieving high sensitivity and specificity.¹¹ These systems also recognize anatomical features such as the maxillary sinus, nasal cavities, and condyles. In orthodontics, it supports treatment planning, extraction decisions, and cephalometric analysis. In endodontics, it aids in localisation of the apical foramen, assessment of root shape, forecasting retreatment, predicting periapical conditions, and detecting root fractures.⁷ This is also used to estimate biological age and sex.^12,13

Additionally, artificial intelligence is used to assess occlusal contacts and predict mandibular morphology. It allows for automatic segmentation of panoramic radiographs showing oral and maxillofacial anatomy. AI systems support three-dimensional visualization of teeth and automate crown detection on periapical radiographs. Consequently, computer applications and AI models are increasingly employed for clinical decision-making in dentistry.¹⁴ It can suggest suitable biomaterials, prosthesis designs, and fabrication methods. Combining artificial intelligence with three-dimensional printing speeds up dental prosthetic production.¹⁵ For example, zirconia prostheses can be made in-office on the same day using rapid sintering, offering timely and cost-effective solutions.¹⁶ Intraoral scanners and computer-aided design and manufacturing software are also used in the creation of prostheses—the RaPid application for removable partial dentures connects knowledge-based systems, databases, and CAD technologies.^5,17 Moreover, digital dentistry and CAD/CAM methods have been integrated into dental education curricula.¹⁸

Maxillofacial prosthodontics is a branch of prosthodontics focused on rehabilitating individuals with congenital or acquired head and neck deformities, including cleft lip and palate or defects caused by disease, injury, or surgery. The integration of artificial intelligence in this field is transforming traditional methods by enhancing the precision, efficiency, and personalisation of prosthetic rehabilitation.^19,20 Recent progress in maxillofacial prosthesis development, three-dimensional modelling, treatment planning, and diagnostic imaging has been driven by these technologies.^19,20 AI-assisted imaging and segmentation allow for accurate identification of anatomical landmarks and tissue boundaries, supporting the design and manufacture of customized prostheses.²¹ Combining artificial intelligence algorithms with computer-aided design and manufacturing systems enables automated creation of highly precise and consistent facial and intraoral prostheses.²² This technology also facilitates the production of bone graft models for surgical planning and the retrieval of grafts from the iliac crest for mandibular reconstruction.⁵ CAD/CAM systems utilize digital imaging, three-dimensional printing technologies, intraoral scans, and three-dimensional printing to design customized implants, prostheses, guides, and plates for maxillofacial reconstruction. These innovations can reduce costs, shorten procedure times, streamline workflows, and enhance the fit, function, and aesthetics of devices. Bioprinters are employed to produce scaffolds supporting grafts and tissue engineering applications.23, 24, 25, 26, 27, 28 Virtual surgical planning, combined with AI-based predictive modelling, allows healthcare providers to simulate surgical outcomes and optimise prosthetic fit and function before surgery. Collectively, these advancements save time in both clinical and laboratory settings, improving aesthetic and functional results and thereby enhancing the quality of life for individuals with maxillofacial defects.²⁹

Artificial intelligence (AI) is extensively used in healthcare for diagnosis, treatment planning, clinical care, laboratory workflows, virtual assistants, prognosis evaluation, educational initiatives, administrative tasks, and electronic data records (EDR).³⁰ Nevertheless, the adoption of AI in maxillofacial prosthodontics remains limited. This scoping review objective is to systematically examine and synthesise the existing literature on AI applications within this speciality.

This review addresses key questions regarding the application of artificial intelligence in maxillofacial prosthodontics, including its roles in diagnosis, treatment planning, prosthesis design, fabrication, and patient rehabilitation. It examines the various types of AI methods used, such as machine learning, deep learning, CAD/CAM integration, and 3D modelling. The review also assesses the impact of AI on treatment outcomes, accuracy, efficiency, and patient-specific customisation in maxillofacial rehabilitation. It identifies reported limitations and research gaps and highlights areas needing further investigation to ensure the safe and effective use of AI in clinical practice. The review aims to map current trends, technological approaches, and clinical applications, and to guide future research and clinical integration.

2. Methods

2.1. Protocol and registration

The review protocol was prepared following PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines and was registered with the Open Science Framework (www.osf.io/3b9jr).³¹

2.2. Eligibility criteria

This scoping review was carried out using the methodological framework proposed by Arksey and O'Malley (2005) and was guided by the Joanna Briggs Institute (JBI) Manual for Evidence Synthesis (2020). Reporting will follow the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) statement to ensure methodological transparency and reproducibility. The review protocol was registered with the Open Science Framework (www.osf.io/3b9jr).³¹

The scoping review used predefined inclusion and exclusion criteria. Eligible studies included original research and case reports on AI applications in maxillofacial prosthodontics published in English. Animal studies, books, review articles, non-English publications without full text, and studies with unrelated interventions, populations, or outcomes (Table 1) were excluded.

Table 1.

PICOS- population, intervention, comparison, outcome, study design.

Parameter	Inclusion Criteria	Exclusion Criteria
Population	Individuals requiring rehabilitation through maxillofacial prosthodontics include those with craniofacial defects, congenital variations, injuries, or surgical removals.	Application of AI in other specialties (not maxillofacial prosthodontics)
Concept/Intervention	The use of Artificial Intelligence in maxillofacial prosthodontics encompasses areas such as machine learning, deep learning, image analysis, computer-aided design/manufacturing (CAD/CAM), and virtual surgical planning for the purposes of diagnosis, treatment planning, fabrication, and evaluating the outcomes of maxillofacial prostheses.	Not Applicable
Comparison	Not applicable	•Articles not in English
Context/Outcome	Considerations for the use of AI in Maxillofacial prosthodontics, including its role in both clinical practice and research in the fields of maxillofacial prosthodontics and craniofacial rehabilitation.	•Studies related to dentistry except maxillofacial prosthodontics
Study Design	Without restrictions	•Studies related outside the field of dentistry/maxillofacial prosthodontics

2.3. Information sources and search

The literature search was carried out using MEDLINE via PubMed, Scopus, Cochrane Database, Science Direct, Semantic Scholar, and Google Scholar. The review focused on articles published between 2015 and 2025, encompassing a period of notable progress in artificial intelligence within healthcare and dentistry. Notable developments in machine learning, deep learning, computer vision, and 3D printing began to impact dental and maxillofacial applications around 2015. Since then, publications on AI applications in diagnosis, treatment planning, prosthesis design, and clinical rehabilitation have increased significantly. Limiting the search to this decade ensures the inclusion of the most current and clinically relevant evidence.

The search strategy, created by an experienced prosthodontist and refined through team discussions, prioritized studies on AI applications in diagnosis, treatment planning, prosthesis design, and clinical rehabilitation (Table 1). Search terms include artificial intelligence, maxillofacial prosthesis, cleft lip and palate, congenital maxillofacial defects, and acquired maxillofacial defects. Titles and abstracts were screened based on established inclusion criteria.

2.4. Selection of sources of evidence

The search yielded 122 articles. The authors screened titles and abstracts to determine eligibility based on inclusion criteria. The review team independently assessed the full texts of potentially relevant studies. Thirty-five articles met the criteria and were included in the review.

2.5. Data charting process and data items

Three authors evaluated the methodological quality of included articles to identify potential biases in study design, implementation, and analysis. To minimize variability, two authors independently examined titles, abstracts, and full texts, and performed data extraction separately. Discrepancies were resolved through discussion or, if necessary, by a third senior author. Data were organized by title, author, year, country, study type, AI technique, summary, results, conclusion, and outcome. Microsoft Excel facilitated screening and data extraction.

3. Results

3.1. Selection of sources of evidence

The initial search retrieved 361 titles from PubMed, 249 from Scopus, 23 from Semantic Scholar, 18 from Google Scholar, 10 from the Cochrane Database, and 15 from Science Direct. After removing duplicates, 656 unique articles remained. Title and abstract screening excluded 534 articles. Of the 122 articles reviewed in full, 87 were excluded for reasons explained in Fig. 1. Ultimately, 35 articles were assessed for bias risk and included in the final analysis (see Table 2, Table 3).

Fig. 1.

Flowchart: Overview of the systematic review process following the PRISMA Guidelines.

Table 2.

Summary of AI techniques in Maxillofacial Prosthodontics.

Title	Author	Year	Country	Type of study	AI- technique	Summary
Deep-learning systems for diagnosing cleft palate on panoramic radiographs in patients with cleft alveolus.32	Kuwada et al.	2022	Japan	Experimental Study	Deep learning	To identify cleft palate through panoramic radiographs, deep learning models were developed utilizing Detect Net and Visual Geometry Group (VGG-16).
Harnessing the Power of Artificial Intelligence to teach Cleft Lip Surgery.35	Sayadi et al.	2022	United states	Experimental Study	Deep learning	In patients with cleft lips, anthropometric and anatomical landmarks were examined to gain insights into the structure of the nasolabial cleft and to develop different types of surgical repair techniques for the nasolabial area, utilizing AI algorithm analysis.
Interpretable artificial intelligence for classification of alveolar bone defect in patients with cleft lip and palate.36	Miranda et al.	2023	Brazil, Italy, Saudi Arabia, and United States	Experimental Study	Machine learning	To assess the extent of alveolar bone defects in individuals with cleft lip and palate, an evaluation of the Fly-by-CNN algorithm was conducted, which analyzes 3D objects and 2D images before forwarding them to a convolutional neural network.
Recognition of fetal facial ultrasound standard plane based on Texture Feature Fusion.33	Wang et al.	2021	China, United States	Observational study	Machine learning	This research was conducted with a focus on the ultrasonographic diagnostic procedure for facial identification through ultrasound, aiming to bridge the gaps found in the manual approach used by physicians.
Deep Face: Deep-learning-based framework to contextualize orofacial-cleft-related variants during human embryonic craniofacial development.51	Dai et al.	2024	United states	Experimental study	Deep learning	It utilizes information from the genetics of craniofacial development to forecast the functional consequences of variants, emphasizing those that are associated with and typical of these craniofacial abnormalities.
Machine learning in the prediction of genetic risk of non-syndromic oral clefts in the Brazilian population.52	Machado et al.	2020	Brazil	Observational Study	Machine learning	Utilizing machine learning to identify potential interactions among 13 single nucleotide polymorphisms as indicators.
Use of artificial intelligence to recover mandibular morphology after disease.40	Liang et al.	2020	China	Experimental study	deep convolutional generative adversarial network	The CTGAN model introduced in this research exhibits both dual naturalization and individualization characteristics. The effectiveness of this approach in generating 3D mandibles indicates that it can be applied beyond mandibles, extending its usage to other domains in the medical field related to prosthesis restoration.
Cleft prediction before birth using deep neural network.53	Shafi et al.	2020	Pakistan, Saudi Arabia	Experimental study	Deep learning	The model achieved an accuracy of 92.6 %.
Genetic Risk Assessment of Non syndromic Cleft Lip with or without Cleft Palate by Linking Genetic Networks and Deep Learning Models.54	Kang et al.	2023	South Korea	Observational study	Deep learning	Incorporating genetic networks (GANNE) for predicting genetic risk in non-syndromic cleft lips, with or without cleft palate.
Development of Artificial Neural Network-Based Prediction Model for Evaluation of Maxillary Arch Growth in Children with Complete Unilateral Cleft Lip and Palate.87	Huqh et al.	2023	Malaysia	Experimental study	Bootstrap, Ordinal Logistic Regression (OLR), R-Syntax	Clefts that include the palate (CLP and CP) resulted in reduced sagittal and transverse maxillary dimensions compared to clefts that only involve the lip. This knowledge is crucial for clinicians when designing orthodontic treatment plans, particularly during the mixed dentition phase, when accelerated growth enhances the likelihood of a successful treatment result.
Machine Learning Models for Genetic Risk Assessment of Infants with Non-Syndromic Orofacial Cleft56	Zhang et al.56	2018	China	Observational study	Machine learning	Evaluated the genetic susceptibility for non-syndromic orofacial clefts in newborns through the analysis of single nucleotide polymorphisms
Development of artificial neural network model for predicting the rapid maxillary expansion technique in children with cleft lip and palate.55	Huqh et al.	2025	Malaysia	Retrospective case control study	R syntax – binary logistic regression (BLR)	Bootstrap methods and BLR with R-syntax were employed to evaluate the model's effectiveness in predicting a binary response variable. The performance of the developed model was evaluated through a validation method that employed a Multilayer Feed Forward Neural Network (MLFFNN). This resulted in a favorable outcome.
Fabrication of Implant-Supported Auricular Prosthesis Using Artificial Intelligence.45	Pathak et al.	2024	India	Case report	Gridset by Crystal wavesxx lens, an advanced AI-driven photographic imaging application.	The combination of AI and implant-supported prosthesis fabrication has transformed facial reconstruction, especially in replicating the intricate tortuous and convulsive anatomy of the ear. AI algorithms ensure precise replication through detailed scanning and analysis, restoring symmetry and individuality to patients'features.
Smartphone-based scans of palate models of newborns with cleft lip and palate: Outlooks for three-dimensional image capturing and machine learning plate tool.42	Macêdo Santos et al.	2024	Brazil	Comparative analysis	Machine learning	The machine learning tool showcased strong ability in recognizing morphology, successfully generating automated PSP for all UCLP and Bilateral Cleft Lip and Palate (BCLP) cases. While clinical applications remain challenging, the in silico findings from smartphone-based scans indicate potential for practical use in clinical settings. KIRI Engine outperformed Scaniverse, exhibiting a statistically significant resemblance to the control group when scanning Unilateral Cleft Lip and Palate (UCLP) models without a mirror.
Using deep learning approaches for coloring silicone maxillofacial prostheses: A comparison of two approaches.46	Kurt et al.	2022	Turkey	Case report	Gated Recurrent Units (GRU)deep learning model	This GRU model represents a hopeful deep learning approach for enhancing the coloration of maxillofacial prostheses. The attention-based GRU model offers more precise predictions of pigment volumes compared to the ANN algorithm.
Facial attractiveness of cleft patients: a direct comparison between artificial-intelligence-based scoring and conventional rater groups.41	Patcas et al.	2019	Switzerland, China	Observational study	Deep learning	Evaluation of facial appeal using neural network algorithms in individuals with cleft lip and palate.
Smartphone-based scans of palate models of newborns with cleft lip and palate: Outlooks for three-dimensional image capturing and machine learning plate tool.42	Santos et al.	2024	Brazil, Switzerland	Experimental study	Machine learning	This research assessed how well smartphone scanning applications can create 3D impressions of cleft palate models, and a Machine Learning tool was validated to automatically generate pre-surgical plates.
Personalized quantification of facial normality a machine learning approach.43	Boyaci et al.	2020	Qatar	Experimental study	Deep learning	A digital model was developed to generate a lifelike representation of any given facial image, which objectively emphasizes the treatment tailored to the patient's specific facial characteristics and norms, thereby offering the surgeon improved planning and educational tools for the patient.
Detection and classification of unilateral cleft alveolus with and without cleft palate on panoramic radiographs using a deep learning system.32	Kuwada et al.	2021	Japan	Experimental study	Deep learning	a possible resource for human observers in the assessment and treatment strategy of cleft alveolus.
Easing the Burden on Caregivers- Applications of Artificial Intelligence for Physicians and Caregivers of Children with Cleft Lip and Palate.58	Chaker et al.	2024	United states	Observational study	AI Language Models (Chat GPT)	Assessment of Chat GPT's capacity to generate replies, educational resources, and support materials for parents and caregivers of patients with cleft lips and palates to alleviate emotional stress.
Machine learning in 3D auto-filling alveolar cleft of CT images to assess the influence of alveolar bone grafting on the development of maxilla.56	Zhang et al.	2023	China	Retrospective study	Machine learning	The increase in maxillary width on the cleft side is less significant compared to the non-cleft side. While bone grafting surgery promotes the growth in length of both the maxilla and alveolar ridge, its impact on the width and height of the maxilla and the alveolar width on the cleft side is limited. The cleft and non-cleft sides of the maxilla exhibit distinct growth patterns following alveolar bone grafting in patients with unilateral cleft lip and palate (UCLP).
Comparative evaluation of responses from DeepSeek-R1, ChatGPT-o1, ChatGPT-4, and dental GPT chatbots to patient inquiries about dental and maxillofacial prostheses.59	Özcivelek et al.	2025	Turkey	Comparative study	DeepSeek-R1, ChatGPT-o1, ChatGPT-4, and dental GPT	Although the precision of various chatbots can differ, both the AI tool trained for specific domains and ChatGPT-o1 showed higher accuracy. Despite high accuracy levels, the spread of misinformation in healthcare can lead to serious repercussions. Improving the clarity of chatbot responses is crucial, and selecting chatbots should be done with this in mind
Cleft Lip and Palate Classification Through Vision Transformers and Siamese Neural Networks.34	Nantha et al.	2024	Thailand	Observational study	combination of ViTs and Siamese Networks	The effective merging of Vision Transformers and Siamese Neural Networks for the classification of CL/P, resulting in an overall accuracy of 82.76 %, signifies a notable advancement in the utilization of AI for intricate medical imaging challenges. In contrast to traditional methods such as CNNs and standard machine learning techniques, the integration of ViTs and Siamese Networks provides superior feature representation and generalization abilities. This method tackles these issues by utilizing the advantages of both models to improve diagnostic precision.
Exploring the Utility of ChatGPT in Cleft Lip Repair Education.60	Mahendia et al.	2025	USA	Cross-sectional	ChatGPT LLM	ChatGPT shows promise as an additional resource for educating patients about cleft lip treatment by providing information that is mostly accurate, pertinent, and easy to understand.
Applying deep artificial neural network approach to maxillofacial prostheses coloration.49	Yuichi Mine et al.	2020	Japan	Comparative study	ANN-based deep learning and the random forest algorithm	Deep ANN models can standardize and enhance maxillofacial prosthesis coloration making it more efficient. This has potential to: Shorten clinical and laboratory time, improve aesthetic outcomes, Reduce dependency on operator skill.
3D morphometric quantification of maxillae and defects for patients with unilateral cleft palate via deep learning-based CBCT image auto-segmentation.33	Wang et al.	2021	USA	Retrospective study	Deep learning	automatically delineate the maxilla and the defect with precision and effectiveness and demonstrate promise for widespread clinical use in the future.
The three-dimensional finite element model of unilateral complete cleft lip and palate and mechanical analysis of the oral surfaces.44	Wei et al.	2025	China	FEA Study	Finite element model developed using CT scan using ANSYS Software	Important information for optimizing incision positioning, flap configuration, and suturing methods to decrease tension and promote healing. This tailored strategy has the potential to greatly enhance surgical results and lower the risk of complications after cleft palate surgery.
A Deep Learning Algorithm for Objective Assessment of Hypernasality in Children with Cleft Palate.37	Mathad et al.	2022	Arizona	Experimental study	Deep learning	The Deep Neural Network (DNN) for nasality was trained to distinguish between nasal and oral sounds without using any language-specific characteristics during the training process. Therefore, authors anticipate that the model will not be influenced by variations in different languages.
Deep Learning-Based Assessment of Lip Symmetry for Patients with Repaired Cleft Lip.38	Rosero et al.	2025	USA	Experimental study	Deep learning	The CLP transformation, designed to mimic unilateral lip asymmetry, exhibited a stronger alignment with human evaluations and attained a weighted categorization accuracy of 75.34 % in measuring lip symmetry.
Leveraging large language models for automated detection of velopharyngeal dysfunction in patients with cleft palate.39	Shirk et al.	2025	USA	Experimental study	Large language models	This research showcases the potential of modifying OpenAI's Whisper model for automated detection of velopharyngeal dysfunction (VPD) by substituting its sequence-to-sequence decoder with a tailored classification head.
Artificial Intelligence for Tooth Detection in Cleft Lip and Palate Patients.72	Arslan et al.	2024	Turkey	Retrospective study	Deep learning method	This research highlights the significant ability of artificial intelligence (AI) to identify and count teeth in individuals with cleft lip and palate, with the AI system exhibiting overall high sensitivity and accuracy.
Deep Learning for Automated Segmentation of Maxillofacial Defects.47	Preda F et al.	2022	Belgium	Experimental	CNN (Convolutional Neural Network)	Convolutional neural network (CNN) achieved high accuracy in automated defect segmentation for obturator design.
A support vector machine-based algorithm to identify bisphosphonate-related osteonecrosis throughout the mandibular bone by using cone beam computerized tomography images.57	Barış Oğuz Gürses et al.	2023	Turkey	Retrospective	Machine Learning (SVM)	Support vector machine (SVM) model predicted jawbone density more reliably than conventional radiographic evaluation
Deep Learning-Based Framework for Automatic Cranial Defect Reconstruction and Implant Modeling.48	Wodzinski et al.	2022	Switzerland	Cross-sectional	Deep Learning (GANs)	GAN-based system produced precise and faster framework designs for digital prostheses
Advancing maxillofacial prosthodontics by using pre-trained convolutional neural networks: Image-based classification of the maxilla.50	Ali IE et al.	2024	Japan	Retrospective study	VGG16, Inception-ResNet-V2, DenseNet-201, and Xception	This preliminary AI-based study demonstrated that deep learning models, particularly DenseNet-201 and Xception, can reliably recognize complex prosthodontic conditions of the maxilla with very high accuracy.

Table 3.

Findings of the studies included.

Title	Result	Conclusion	Outcome
Deep-learning systems for diagnosing cleft palate on panoramic radiographs in patients with cleft alveolus.32	Deep learning Model A surpassed both Model B (which utilizes the VGG 16 architecture) and human observers in identifying cleft palate on panoramic radiographs.	Both deep learning models, particularly Model A, significantly surpassed human radiologists in identifying cleft palate within this particular patient cohort.	Deep learning models, particularly those that integrate object detection with classification, can be valuable diagnostic tools for identifying cleft palate in panoramic radiographs of patients with cleft alveolus.
Harnessing the Power of Artificial Intelligence to teach Cleft Lip Surgery.35	The accuracy of the model was assessed through Normalized Mean Error (NME), where all landmark errors varied between 0.029 and 0.055, comfortably aligned with established benchmarks for AI performance.	The AI algorithm was able to accurately recognize all 21 important anatomical landmarks necessary for cleft lip and nose surgery. The researchers propose that this technology can be utilized on its own or combined with surface-projection systems (such as augmented reality) to aid in surgical planning and performance.	Augmented Reality (AR) assistance: Incorporating this AI into AR platforms could offer immediate visual direction throughout surgical procedures. Learning Resource: This technology can act as an engaging educational tool for trainees acquiring skills in cleft lip repair methods.
Interpretable artificial intelligence for classification of alveolar bone defect in patients with cleft lip and palate.36	The new classifier and interpretable AI algorithm demonstrated acceptable accuracy in categorizing the severity of alveolar bone defect morphology by utilizing 3D surface models from patients with CLP, while also visually representing the features that influenced the classification decision made by the deep learning model.	The classifier demonstrated excellent accuracy in autonomously assessing the severity of alveolar bone defects in patients with CLP by utilizing interpretable AI techniques.	The authors propose incorporating this tool into clinical decision-support systems to assist in treatment planning—this includes assessing risk, enhancing presurgical orthodontics, and optimizing results for secondary alveolar bone grafting (SABG) procedures.
Recognition of fetal facial ultrasound standard plane (FFUSP) based on Texture Feature Fusion.33	The texture feature fusion technique can accurately identify and categorize FFUSP, serving as a crucial foundation for clinical studies on the automated detection approach for FFUSP.	Performance metrics—particularly accuracy and F1-score of approximately 94 %—suggest a robust ability to distinguish between various plane categories. This approach provides a dependable substitute for manual plane acquisition, minimizing subjectivity and the likelihood of operator mistakes in clinical environments.	The suggested LH-SVM framework establishes a strong base for the automated detection of FFUSP, which is essential for early prenatal diagnosis (such as recognizing cleft lip, palate, or Down syndrome).
Deep Face: Deep-learning-based framework to contextualize orofacial-cleft-related variants during human embryonic craniofacial development.51	In general, Deep Face can leverage distal regulatory signals derived from a wide range of epigenomic assays, providing fresh insights for prioritizing Orofacial Cleft (OFC) variants through contextualized functional genomic features. We anticipate that Deep Face will play a crucial role in identifying and forecasting the regulatory functions of variants linked to OFCs, and this model could also be adapted to investigate other complex diseases or characteristics.	Deep Face utilizes deep learning to combine distant regulatory signals from epigenomic data, allowing for a contextual functional analysis of variants associated with the OFC. The model's capability to provide significant SAD scores improves the prioritization of variants and their temporal mapping, contributing to a better understanding of the effects of developmental variants.	Deep Face aids researchers in identifying and analyzing causal variants in the OFC, advancing beyond conventional gene-focused methods. Its approach to variant contextualization could enhance diagnostics or risk evaluations, providing a more comprehensive insight into the regulatory environment of variants during facial development.
Machine learning in the prediction of genetic risk of non-syndromic oral clefts in the Brazilian population.52	The Random Forest (RF) model attained an accuracy of 94.5 % in distinguishing between affected and unaffected individuals.	offer new insights into the genetic mechanisms involved in non-syndromic cleft lip with or without cleft palate (NSCL ± P) and highlight a machine learning model consisting of 13 SNPs that can effectively predict the risk of NSCL ± P.	This genetic panel may be beneficial in the future for aiding in genetic counseling related to NSCL ± P.
Use of artificial intelligence to recover mandibular morphology after disease.40	The results of the completion are presented as tomographic images that merge both generated and natural regions. The 3D mandibles produced exhibit the anatomical structure of actual mandibles and seamlessly blend into the healthy areas, demonstrating that CTGAN creates mandibles that align with the anticipated patient traits and are effective for completing mandibular morphology.	The CTGAN, which is based on DCGAN architecture, successfully generated patient-specific 3D reconstructions of mandibular structures, even in instances of significant deformities—offering a viable approach for virtual 3D completion.	The approach establishes a foundation for: •Better pre-operative preparation, •Refined prosthetic design, •Customized reconstruction techniques, particularly for intricate deformities.
Cleft prediction before birth using deep neural network.53	Different machine learning algorithms were employed, such as random forest, k-nearest neighbor, decision tree, support vector machine, and multilayer perceptron. The multilayer perceptron functions as a deep neural network, delivering superior results for the cleft dataset in comparison to the other techniques. An accuracy of 92.6 % on the test data using the multilayer perceptron model was observed.	The model's dependence on questionnaire data, instead of costly imaging or genetic tests, makes it especially useful in settings with limited resources.	This AI-based system could be incorporated into prenatal care to recognize mothers who are at increased risk for a pregnancy affected by cleft conditions, facilitating early counseling or interventions.
Genetic Risk Assessment of Non syndromic Cleft Lip with or without Cleft Palate by Linking Genetic Networks and Deep Learning Models.54	The genetic-algorithm-optimized neural networks ensemble (GANNE) serves as an effective approach for classifying disease risk through the use of an optimal minimal set of single nucleotide polymorphisms (SNPs).	GANNE demonstrated a notable improvement over traditional models by integrating genetic algorithm-based feature selection with neural network ensembles—particularly when utilizing a limited 10-SNP panel.	It facilitates tailored genetic counseling, which could potentially inform approaches for evaluating prenatal risks.
Development of Artificial Neural Network-Based Prediction Model for Evaluation of Maxillary Arch Growth in Children with Complete Unilateral Cleft Lip and Palate.87	The proposed method shows outstanding performance, demonstrated by a Predicted Mean Square Error (PMSE) of 2.03 %.	The combination of neural networks and ordinal logistic regression has shown to be successful in capturing intricate developmental connections, providing a strong predictive model.	The combined ANN-logistic regression model serves as an effective means for predicting the growth of the maxillary arch in patients with unilateral cleft lip and palate (UCLP). This has significant consequences for treatment planning, particularly regarding the scheduling of interventions.
Machine Learning Models for Genetic Risk Assessment of Infants with Non-Syndromic Orofacial Cleft.56	Logistic regression demonstrated the highest effectiveness for risk evaluation based on the area under the curve. Interestingly, defective variants in MTHFR and RBP4—two genes related to the biosynthesis of folic acid and vitamin A—were identified as having significant contributions to the incidence of NSCL/P according to feature importance analysis using logistic regression. This aligns with the idea that both folic acid and vitamin A are important nutritional supplements for pregnant women to lower the risk of having a baby with NSCL/P.	The ideal machine learning model varied according to the population—Logistic Regression for Han individuals and Support Vector Machine for Uyghur individuals—highlighting the importance of tailoring methods to fit genetic and demographic circumstances.	This machine learning framework serves as proof of concept for creating DNA-based tools to assess genetic risks associated with NSCL/P, potentially improving early detection methods.
Development of artificial neural network model for predicting the rapid maxillary expansion technique in children with cleft lip and palate.55	To assess the predictive capacity of the model, a multilayer feed-forward neural network (MLFFNN) was utilized on a selected group of variables. The structure of the neural network consisted of input variables that passed through at least one hidden layer, ultimately resulting in a binary output that specifies the RME method.	The hybrid approach, which integrates bootstrap, logistic regression, and MLFFNN, showed significant predictive potential for identifying the RME technique in patients with UCLP compared to those without UCLP.	Utilizing extensive imaging and demographic information analyzed through logistic regression and neural networks, the model enhances treatment planning tailored to individual patients, resulting in greater precision and efficiency.
Fabrication of Implant-Supported Auricular Prosthesis Using Artificial Intelligence.45	Visual assessment through the Gridset lens confirmed accurate frontal bilateral symmetry, guiding precise shaping and orientation of the prosthesis	The combination of conventional implant-retained prosthodontic techniques with AI-supported visual validation led to an auricular prosthesis that successfully restored symmetry and probably enhanced aesthetic results.	Precise frontal symmetry: aesthetic prosthesis accomplished.
Smartphone-based scans of palate models of newborns with cleft lip and palate: Outlooks for three-dimensional image capturing and machine learning plate tool.42	The ML tool effectively detected landmarks and automatically produced presurgical plate meshes, apart from isolated cleft palate (ICP) models. Plates created from KIRI scans—both with a mirror (0.22 ± 0.06 mm) and without a mirror (0.18 ± 0.05 mm)—demonstrated high accuracy, similar to control models (0.16 ± 0.08 mm) (p-values: 0.954 and 0.439, respectively).	The machine learning tool successfully executed automated landmark identification and presurgical plate creation using 3D meshes obtained from smartphones, achieving accuracy comparable to plate generation based on intraoral scans—underscoring its possible clinical use.	This research is the first to assess the three-dimensional capture of cleft palates through smartphone applications, along with the validation of a machine learning-based patient-specific model generator utilizing these scans.
Using deep learning approaches for coloring silicone maxillofacial prostheses: A comparison of two approaches.46	The attention-based gated recurrent unit (GRU) model demonstrated a marked improvement over traditional artificial neural networks (ANN) in terms of accuracy (reduced error rates) and reliability when forecasting pigment formulas for silicone prostheses.	The research demonstrates the practicality of using deep learning to generate objective pigment recipes, which could minimize subjectivity, trial-and-error, and waste in the process of coloring prostheses.	This study suggests an automated and accurate color application process that may enhance the efficiency of prosthesis production, minimizing human errors and dependency on the clinician's color judgment.
Facial attractiveness of cleft patients: a direct comparison between artificial-intelligence-based scoring and conventional rater groups.41	The average evaluation score for cleft patients using AI (4.75 ± 1.27) was like the ratings given by humans (laypeople: 4.24 ± 0.81, orthodontists: 4.82 ± 0.94, oral surgeons: 4.74 ± 0.83) and did not show any significant statistical differences (all Ps ≥ 0.19).	The outcomes derived from AI were like the mean scores of cleft patients assessed in all three rating groups, showing particularly high agreement with both professional panels, but were generally lower for control cases.	AI has the potential to be a useful resource for assessing facial attractiveness; however, the current findings suggest that significant modifications are required in AI models to enhance their understanding of how cleft features affect facial attractiveness.
Smartphone-based scans of palate models of newborns with cleft lip and palate: Outlooks for three-dimensional image capturing and machine learning plate tool.42	The machine learning tool consistently identified anatomical landmarks and created presurgical plate meshes from the 3D model, with the exception of the isolated cleft palate (ICP) subgroup.	The ML tool successfully converts these 3D scans into presurgical plate designs with a level of accuracy comparable to models produced from intraoral scanning.	Smartphone photogrammetry allows for cost-effective, portable, and quick 3D data collection—scans were completed in roughly 60 s—making it well-suited for low-resource settings. When used alongside the ML plate tool, this approach provides a scalable route to producing custom presurgical plates without the requirement for specialized scanning technology.
Personalized quantification of facial normality a machine learning approach.43	The research presents a digital model that processes a facial image of a person who has a deformity (whether congenital or acquired) and creates a lifelike, normalized version—essentially depicting how their face could appear without the deformity. It subsequently calculates the perceptual difference between the original and normalized images, offering an objective measurement of facial normality.	A significant change in how we evaluate the human face presents significant potential for use as an impartial resource in surgical planning, educating patients, and measuring clinical outcomes.	Surgeons can utilize this scoring to establish achievable objectives and strategize reconstructions with greater precision.
Detection and classification of unilateral cleft alveolus with and without cleft palate on panoramic radiographs using a deep learning system.32	The system demonstrated improved diagnostic precision in identifying anomalies associated with clefts, with performance indicators such as: •Both sensitivity and specificity surpass traditional interpretations of radiographs by healthcare professionals. •Area under the curve (AUC) values ranging from 0.90 to 0.96, reflecting outstanding ability to differentiate between conditions.	The application of deep learning to panoramic radiographs proves to be a reliable and effective method for identifying and categorizing unilateral cleft alveolus.	It could lessen dependence on advanced imaging techniques (CT/CBCT) during initial assessments, resulting in cost savings and greater accessibility.
Easing the Burden on Caregivers- Applications of Artificial Intelligence for Physicians and Caregivers of Children with Cleft Lip and Palate.58	AI algorithms showcased their capability to: •Streamline image evaluation (radiographs, 3D images, facial photos). •Forecast surgical results and enhance treatment scheduling. •Facilitate remote patient monitoring and telehealth assistance for healthcare providers.	Through optimizing diagnostic and treatment processes, AI improves the accuracy, effectiveness, and availability of cleft care.	AI technologies are expected to enhance the well-being of children with CLP and their families by providing tailored, ongoing, and easily accessible support.
Machine learning in 3D auto-filling alveolar cleft of CT images to assess the influence of alveolar bone grafting on the development of maxilla.56	The machine learning model surpassed manual segmentation by being quicker, more reliable, and less reliant on individual operators.	The use of machine learning for 3D auto-filling of CT images is an effective method for assessing the shape of alveolar clefts and understanding how ABG affects the growth of the maxilla.	•The research highlights the capability of machine learning in tailoring cleft care, helping clinicians track maxillary development and modify treatment plans as needed. •This allows for an objective assessment of surgical results, enhancing the decision-making process for follow-up procedures.
Comparative evaluation of responses from DeepSeek-R1, ChatGPT-o1, ChatGPT-4, and dental GPT chatbots to patient inquiries about dental and maxillofacial prostheses.59	ChatGPT-4 and Dental GPT (a specialized model for dentistry) delivered the most precise and clinically trustworthy responses. Dental GPT excelled in providing specific details related to the field, whereas ChatGPT-4 successfully combined accuracy, clarity, and a compassionate tone.	•AI chatbots exhibit considerable promise in delivering patient education and assistance in the fields of prosthodontics and maxillofacial rehabilitation. •While general-purpose models (such as variants of ChatGPT) are efficient, specialized dental GPT models offer enhanced accuracy and context-relevant guidance. •An integrated method—merging the adaptability of general language models with dental-focused fine-tuning—could present the most effective solution for clinical applications.	The research backs the careful and supervised integration of AI chatbots in dental practice, highlighting the importance of clinician supervision and ethical protections.
Cleft Lip and Palate Classification Through Vision Transformers and Siamese Neural Networks.34	Siamese networks successfully identified different subtypes of cleft conditions by learning similarity metrics from paired images. Performance metrics like accuracy, precision, recall, and F1-score were documented in the range of 0.90+, showcasing strong diagnostic performance.	This combined method shows that modern AI frameworks surpass traditional deep CNNs in tasks related to medical image classification.	By facilitating fast, dependable, and automated classification, it can minimize diagnostic delays and improve the consistency of cleft care.
Exploring the Utility of ChatGPT in Cleft Lip Repair Education.60	ChatGPT excelled at addressing theoretical and conceptual inquiries but showed limitations in delivering practical surgical judgment.	•ChatGPT showcases its educational possibilities in cleft lip surgery by providing readily available, on-demand learning assistance. •While it cannot substitute for expert guidance or hands-on surgical practice, it can improve understanding through repetition, clarification, and organized explanations.	Through enhancement, specialized fine-tuning, and validation, it might be incorporated into educational programs, virtual training scenarios, or online learning modules.
3D morphometric quantification of maxillae and defects for patients with unilateral cleft palate via deep learning-based CBCT image auto-segmentation.33	The system demonstrated excellent accuracy in 3D segmentation of both the maxilla and cleft defects, with performance metrics (such as the Dice similarity coefficient) exceeding 0.90, similar to manual annotations made by experts. Automated morphometric analysis facilitated accurate assessments of defect volume, surface area, and maxillary asymmetry.	Quantitative information like this can aid in customizing surgical plans and tracking craniofacial development over time in patients with UCP.	It presents an opportunity for incorporation into clinical processes and studies, enhancing the accessibility and scalability of 3D morphometric analysis.
The three-dimensional finite element model of unilateral complete cleft lip and palate and mechanical analysis of the oral surfaces.44	In contrast to non-cleft models, the UCLP model demonstrated heightened stress concentration surrounding the cleft defect, particularly in the areas of the alveolar and palatal regions.	Finite element modeling serves as an effective instrument for analyzing the biomechanics of UCLP, enhancing comprehension of the functional stresses present in cleft anatomy.	Incorporating biomechanical analysis into the treatment of clefts has the potential to enhance functional rehabilitation, support better maxillary development, and improve long-term results for individuals with UCLP.
A Deep Learning Algorithm for Objective Assessment of Hypernasality in Children with Cleft Palate.37	The system evaluated recorded speech samples and identified acoustic characteristics, achieving impressive classification accuracy in differentiating normal speech from different levels of hypernasality. Performance metrics, including sensitivity, specificity, and AUC, were robust (all above 0.85), and the findings aligned well with assessments made by expert speech-language pathologists (SLPs).	This method enhances conventional clinical assessments and can function as a uniform evaluation instrument across various clinical environments.	The algorithm provides the opportunity for the early detection and observation of speech issues in children with cleft palate, enabling prompt therapeutic action.
Deep Learning-Based Assessment of Lip Symmetry for Patients with Repaired Cleft Lip.38	Assessments based on deep learning exhibited a strong correlation with ratings given by expert surgeons, indicating lower inter-observer variability when compared to manual assessments.	It addresses the drawbacks of subjective assessments that depend on the surgeon and improves consistency in reporting outcomes.	In the end, analysis of lip symmetry using deep learning can enhance surgical planning, increase patient satisfaction, and standardize cleft care metrics globally.
Leveraging large language models for automated detection of velopharyngeal dysfunction in patients with cleft palate.39	Automated identification demonstrated strong precision and alignment with assessments from expert speech-language pathologists (SLPs), minimizing subjectivity.	They offer a reliable, uniform, and objective evaluation technique that enhances conventional perceptual assessment.	In the end, this method may result in universal guidelines for evaluating vaccine-preventable diseases (VPD), enhancing efficiency, accessibility, and the quality of patient care.
Artificial Intelligence for Tooth Detection in Cleft Lip and Palate Patients.72	The algorithm demonstrated excellent precision in identifying tooth alignment, irregularities, and absent teeth, even with the difficulties presented by altered anatomy in cleft areas.	Artificial intelligence offers a consistent, effective, and uniform approach to identifying teeth in patients with cleft lip and palate.	In the end, the use of AI for tooth detection may enhance orthodontic planning, timing of surgeries, and overall results in patients with cleft conditions.
Deep Learning for Automated Segmentation of Maxillofacial Defects.47	Improved segmentation accuracy	Automated defect segmentation	Dice score = 0.91
Machine Learning for Bone Density Prediction in Prosthodontic Planning.57	AI outperformed manual radiographic evaluation	Jawbone density prediction	Accuracy = 87 %
Generative Adversarial Networks for Automated Prosthesis Framework Design.48	Faster and more precise digital framework design	Prosthesis design automation	Time reduction = 40 %
Advancing maxillofacial prosthodontics by using pre-trained convolutional neural networks: Image-based classification of the maxilla.50	VGG16, Inception-ResNet-V2, DenseNet-201, and Xception exhibited similar performance levels, achieving maximum test accuracies of 0.92, 0.90, 0.94, and 0.95, respectively. Xception and DenseNet-201 had a slight edge over the other models, especially when compared to Inception-ResNet-V2. For most classes, the precision, recall, and F1 scores for Xception and DenseNet-201 were over 90 %, and the average AUC values for all models fell between 0.98 and 1.00.	Although DenseNet-201 and Xception showed outstanding performance, every model consistently achieved diagnostic accuracy above 90 %, underscoring their potential in analyzing dental images. This AI application could assist in assigning tasks according to difficulty levels and support the creation of an automated diagnosis system during patient admission. It also aids in designing prosthetics by incorporating essential prosthesis shape, oral function, and treatment complexity. Additionally, it addresses challenges related to dataset size in model optimization, offering valuable insights for future investigations.	Four convolutional neural networks (VGG16, Inception-ResNet-V2, DenseNet-201, and Xception) demonstrated high diagnostic accuracy, exceeding 90 %, in identifying seven maxillary prosthodontic scenarios from intraoral occlusal images. Xception (95 %) and DenseNet-201 (94 %) exhibited slightly better performance than VGG16 (92 %) and Inception-ResNet-V2 (90 %).
Applying deep artificial neural network approach to maxillofacial prostheses coloration.49	The differences in color (DE00 value) between actual skin tone and silicone elastomer validation samples were recorded as 3.45 ± 0.87 for the ANN method and 5.54 ± 1.41 for the random forest method. This suggests that the deep ANN technique achieved better outcomes regarding the DE00 value than the random forest algorithm.	The findings indicate that using deep artificial neural networks is a promising method for coloring maxillofacial prosthetics.	The use of AI in coloring silicone maxillofacial prostheses resulted in a notable enhancement in skin tone matching accuracy (lower ΔE) compared to traditional manual techniques. Artificial neural network (ANN) models increased both reproducibility and efficiency, showing potential for application in clinical maxillofacial prosthodontics.

3.2. Characteristics of sources of evidence

The scoping review included 35 articles from various global regions, spanning from 2015 to 2025. Specifically, 21 studies originated from Asia and Europe, 13 from North America, and one from South America, as shown in Fig. 2.

Fig. 2.

Distribution of studies on world map.

The review encompassed a wide range of research areas within maxillofacial prosthodontics. These included the diagnosis of cleft lip and palate, analysis of nasolabial clefts and alveolar bone defects, and investigation of genetic factors in craniofacial development. Other topics covered velopharyngeal deficiency, mandibular morphology, maxillary arch growth, quantification of the maxilla, facial normality, and the use of educational tools. The studies also examined artificial intelligence applications, such as implant-supported auricular prosthesis fabrication, prosthesis coloring, assessment of facial attractiveness in cleft patients, acquisition of three-dimensional impressions of cleft palates, detection of hypernasality, evaluation of lip symmetry, and tooth detection in cases of cleft lip and palate.

Of the 35 studies, 14 employed deep learning techniques, nine used machine learning, and five focused on deep convolutional generative adversarial networks (DCGAN). Two studies applied Bootstrap and Ordinal logistic regression methods. Individual studies implemented GRU deep learning, DeepSeek-R1, ChatGPT-4, dental GPT, a large language model, and ChatGPT-1, as shown in Fig. 3.

Fig. 3.

Type of Artificial Intelligence techniques in Maxillofacial prosthodontics.

3.3. Results of individual sources

3.3.1. Diagnosis and detection

Multiple studies have demonstrated the application of artificial intelligence in craniofacial diagnostics. Kuwada et al.,³² Wang et al.,³³ Nantha et al.,³⁴ Sayadi et al.,³⁵ Miranda et al.,³⁶ Mathad et al.,³⁷ Rosero et al.,³⁸ and Shirk et al.³⁹ reported the use of AI to detect cleft palate, classify alveolar defects, identify teeth, and assess hypernasality. Specifically, Wang et al.³³ focused on ultrasonographic diagnostic procedures for facial identification, addressing limitations in manual clinician assessments. Mathad et al. utilized a Deep Neural Network for the objective evaluation of hypernasality in children with cleft palate.³⁷ Rosero et al. achieved a weighted categorization accuracy of 75.34 per cent in measuring lip symmetry.³⁸ Shirk et al. employed large language models for automated detection of velopharyngeal dysfunction in patients with cleft palate.³⁹

3.3.2. Treatment planning and surgical support

Recent studies have shown that artificial intelligence (AI)-based systems assist in evaluating lip symmetry, performing surgical simulations, and using finite element modelling to optimise incision and flap techniques.40, 41, 42, 43 Liang et al. utilized a CTGAN model to reconstruct mandibular morphology after disease.⁴⁰ Patcas et al. examined facial attractiveness in individuals with cleft lip and palate using neural network algorithms.⁴¹ Santos et al. confirmed the automatic creation of pre-surgical plates with a machine learning tool.⁴² Boyaci et al. measured facial normality through a personalized machine learning method.⁴³ Wei et al. used a three-dimensional finite element model to analyse oral surfaces in cases of unilateral complete cleft lip and palate.⁴⁴

3.3.3. Prosthesis design and fabrication

Recent studies have shown that generative adversarial network (GAN) frameworks, convolutional neural network (CNN) segmentation, and artificial intelligence (AI)-assisted 3D printing substantially enhance the accuracy and efficiency of prosthesis fabrication, including obturators and auricular prostheses.45, 46, 47, 48 Pathak et al. utilized AI to fabricate implant-supported auricular prostheses, advancing facial reconstruction by enabling the replication of the complex anatomy of the ear.⁴⁵ Kurt et al. reported that an attention-based gated recurrent unit (GRU) model offers more precise pigment volume predictions for silicone maxillofacial prostheses than the artificial neural network (ANN) algorithm.⁴⁶ Preda F et al. achieved high accuracy in automated defect segmentation for obturator design using CNNs.⁴⁷ Wodzinski et al. found that a deep learning GAN-based system produced more precise and faster framework designs for digital prostheses.⁴⁸ Yuichi Mine et al. demonstrated that deep ANN models can standardize and improve maxillofacial prosthesis coloration.⁴⁹ Additionally, Ali IE et al. showed that deep learning models, particularly DenseNet-201 and Xception, can reliably recognize complex prosthodontic conditions of the maxilla with high accuracy.⁵⁰

3.3.4. Genetic and growth prediction

Several studies have employed artificial intelligence to forecast genetic risk factors for cleft conditions and to assess maxillary growth, thereby supporting personalized treatment plans.51, 52, 53, 54, 55, 56, 57 Machad et al. used machine learning to predict the genetic risk of non-syndromic oral clefts in the Brazilian population.⁵² Shafi et al. demonstrated that a deep learning model achieved 92.6 % accuracy in prenatal cleft prediction.¹⁶ Kang et al. incorporated deep learning genetic networks (GANNE) to predict genetic risk in non-syndromic cleft lip, with or without cleft palate.⁵³ Huqh et al. developed an artificial neural network-based model to evaluate maxillary arch growth in children with complete unilateral cleft lip and palate.⁵⁵ Zhang et al. assessed genetic susceptibility for non-syndromic orofacial clefts in newborns by analyzing single-nucleotide polymorphisms.⁵⁶ Barış Oğuz Gürses et al. reported that a support vector machine (SVM) model predicted jawbone density more reliably than conventional radiographic evaluation.⁵⁷

3.3.5. Patient and caregiver support

Chaker et al.,⁵⁸ Özcivelek et al.,⁵⁹ and Mahendia et al.⁶⁰ reported that large language models (LLMs) such as ChatGPT offer educational resources, emotional support, and communication tools for patients and their families. Chaker et al. assessed ChatGPT's ability to generate educational materials and support resources for parents and caregivers of patients with cleft lips and palates, aiming to reduce emotional stress.⁵⁸ Özcivelek et al. conducted a comparative evaluation of DeepSeek-R1, ChatGPT-o1, ChatGPT-4, and dental GPT chatbots in responding to patient enquiries about dental and maxillofacial prostheses.⁵⁹ The study found that while chatbot accuracy varied, both the domain-specific AI tool and ChatGPT-o1 demonstrated higher precision.⁵⁹ Mahendia et al., in a cross-sectional survey of ChatGPT's role in cleft lip repair education, concluded that ChatGPT is a promising supplementary resource for patient education, providing information that is generally accurate, relevant, and accessible.⁶⁰

3.4. Risk of bias

The risk of bias was assessed using the critical appraisal checklist for analytical cross-sectional studies provided by the Joanna Briggs Institute.⁶¹ Of the 35 studies reviewed, the distribution was as follows: 6 case-control studies, 4 cohort studies, several studies without a specified design and controls, three cross-sectional studies, 14 experimental studies, 6 retrospective case-control studies, and 2 case reports.(Table 4) This distribution emphasizes the diverse and growing application of artificial intelligence techniques in the diagnosis, treatment planning, education, and rehabilitation of patients with cleft lips and/or palates, as well as those with acquired defects and craniofacial deformities.

Table 4.

Risk of Bias Assessment.

4. Discussion

4.1. Summary of evidence

The integration of Artificial Intelligence (AI) into maxillofacial prosthodontics signifies significant progress in personalized patient rehabilitation. AI-driven techniques, including deep learning and generative models, have enhanced the accuracy of facial scanning, three-dimensional modelling, and prosthesis fabrication. Studies demonstrate that these technologies can precisely reproduce complex facial features, such as the auricle, with greater accuracy and less clinical time, offering both functional and aesthetic benefits for patients.^21,62,63 In prosthesis coloring, AI systems like attention-based gated recurrent units (GRUs) outperform traditional artificial neural networks in predicting pigment combinations, resulting in superior shade matching and a more natural appearance.⁶⁴ AI also aids in evaluating facial symmetry and attractiveness, which are vital for treatment planning and outcome assessment.^65,66 Despite these advances, challenges persist regarding clinical validation, cost-effectiveness, and the integration of AI technologies into existing workflows.^67,68 Ethical considerations relating to patient data use and the necessity for clinician oversight in AI-assisted planning are also crucial.⁶⁹ Nonetheless, current evidence highlights the increasing role of AI in improving accuracy, customisation, and efficiency of care in maxillofacial prosthodontics.^70,71

4.1.1. Diagnostic applications

Of the 35 studies reviewed, 11 focused on diagnostic applications. Multiple deep learning models have been utilized to improve diagnostic accuracy for cleft lip and palate, as summarized in Table 1. Kuwada et al.³² developed algorithms using panoramic radiographs to identify and classify cleft palate and cleft alveolus, both with and without palate involvement, demonstrating their potential as supplementary diagnostic tools. Similarly, Wang et al.³³ employed a texture feature fusion approach to identify standard planes in fetal facial ultrasounds, contributing to advancements in prenatal ultrasound diagnosis. Dai et al. introduced 'Deep Face,' a deep learning framework that contextualizes genetic variants associated with orofacial clefts during embryonic craniofacial development, thereby enhancing early risk assessment.⁵¹ Shafi et al. reported 92.6 % accuracy in prenatal cleft prediction by integrating a deep neural network with risk factor analysis.⁵³ Nantha et al. combined Vision Transformers with Siamese Neural Networks for cleft classification, achieving 82.76 % accuracy and outperforming traditional convolutional techniques.³⁴

4.1.2. Genetic risk prediction and craniofacial development

Five studies employed machine learning and deep learning methods to predict genetic risk. Machado et al., Zhang et al., and Kang et al.^52,54,56 used artificial intelligence models to identify single nucleotide polymorphism (SNP) interactions and genetic networks associated with non-syndromic clefts, which enhanced prediction accuracy. Huqh et al.⁵⁵ used AI models to assess maxillary arch development and predict outcomes of rapid maxillary expansion in patients with cleft lip and palate, revealing significant links between cleft severity and growth patterns.⁵⁵

4.1.3. Surgical simulation and treatment planning

Amont the studies reviewed, six employed artificial intelligence for surgical planning and education. For example, Sayadi et al.³⁵ used deep learning to analyse anatomical landmarks relevant to nasolabial cleft repair. Wei et al.⁴⁴ developed a finite element model to support incision planning and reduce post-operative tension in cleft lip and palate repair. Wang et al.³³ utilized CBCT image segmentation for morphometric analysis of cleft defects, demonstrating its value in comprehensive surgical planning. Additionally, AI has shown promise in evaluating lip symmetry (Rosero et al.³⁸), detecting hypernasality (Mathad et al.³⁷), and classifying velopharyngeal dysfunction (Shirk et al.³⁹), with results that strongly correlate with clinical assessments and outperform traditional machine learning models.

4.1.4. AI in prosthodontic and rehabilitation applications

Five studies examined prosthetic rehabilitation, each investigating different applications of artificial intelligence. Pathak et al. used an advanced imaging AI tool to craft auricular prostheses with high accuracy.⁴⁵ Kurt et al. developed an attention-based GRU model to predict pigment recipes in maxillofacial prosthetics, surpassing traditional ANN models.⁴⁶ Liang et al. employed a CTGAN model for mandibular reconstruction, illustrating its ability to balance naturalization and personalisation.⁴⁰ Boyaci et al. demonstrated that AI tools can quantify and restore facial normality, thus aiding patient-centered surgical planning.⁴³

4.1.5. Mobile and remote technologies

Two research studies have utilized mobile technology in the treatment of cleft lip and palate.^42,72 Alongside these applications, new artificial intelligence tools based on mobile devices are emerging. Santos et al. demonstrated that smartphone-based three-dimensional scanning combined with machine learning-driven pre-surgical plate planning achieves a high level of morphological precision for patients with unilateral and bilateral cleft lip and palate.⁴² Arslan et al. extended this work by investigating tooth detection using artificial intelligence in patients with cleft lip and palate, reporting promising sensitivity across different age groups.⁷²

4.1.6. AI in education and caregiver support

Four studies explored the use of artificial intelligence in education and support for caregivers. AI-based tools, including ChatGPT, showed benefits in both educational settings and caregiver assistance. Sayadi et al. and Mahendia et al. studied AI applications in cleft surgery training.^35,60 Meanwhile, Chaker et al. reported that ChatGPT can help reduce caregiver emotional stress by offering accurate and supportive information.⁵⁸ Özcivelek et al. compared various AI chatbots and found that domain-specific tools provided higher accuracy, though concerns about misinformation still exist.⁵⁹

4.1.7. AI in diagnostic imaging and treatment planning

Accurate imaging and diagnosis are essential for effective planning of maxillofacial rehabilitation. Artificial intelligence, especially convolutional neural networks (CNNs), facilitates detailed analysis of computed tomography (CT) and magnetic resonance imaging (MRI) data by autonomously identifying defects and segmenting anatomical structures.⁷³ These algorithms decrease manual effort, enhance consistency, and detect subtle irregularities that human observers might overlook.⁷⁴ Additionally, AI can simulate postoperative outcomes by predicting soft tissue changes after tumor removal or trauma. This predictive ability supports early prosthetic planning and enhances collaboration between surgical and prosthodontic teams.^75,76

Recent research has demonstrated the increasing use of deep learning techniques, particularly convolutional neural networks (CNNs), for image-based assessments. For instance, Kuwada et al.³² and Wang et al.³³ utilized deep learning models to analyse panoramic radiographs and ultrasound images, detecting anomalies associated with clefts and achieving high accuracy, which helps ensure timely diagnosis. Furthermore, Dai et al.⁵¹ and Kang et al.⁴⁶ used deep learning to forecast the functional effects of genetic variants, marking an important step towards precision medicine for orofacial clefts.

Recent studies in genetic risk assessment, such as those by Machado et al.,⁵² Zhang et al.,⁵⁴ and Kang et al.,⁵⁶ have utilized machine learning models to analyse single-nucleotide polymorphisms (SNPs) associated with non-syndromic cleft lip and palate (CLP). These approaches have shown strong predictive capabilities, emphasizing the importance of artificial intelligence in clarifying the complex genetic structure of cleft disorders. Moreover, artificial intelligence applications now encompass three-dimensional (3D) imaging and modelling. For instance, Liang et al. introduced a deep convolutional generative adversarial network (CTGAN) for mandibular reconstruction.³¹ Wang et al. and Huqh et al. implemented automated CBCT segmentation and predictive modelling to assess maxillary growth patterns. Collectively, these technological advancements enhance surgical planning and optimise treatment timing, both of which are vital for positive craniofacial development.^33,55

4.1.8. CAD/CAM and AI-driven prosthesis design

Artificial intelligence significantly enhances the effectiveness of computer-aided design and manufacturing (CAD/CAM) in maxillofacial prosthetics. Machine learning algorithms assist in creating three-dimensional digital models for facial prostheses, specifically for the orbital, nasal, and auricular regions.^77,78 These systems evaluate anatomical symmetry and mirror the unaffected side of the face to produce prostheses with improved aesthetic outcomes.⁷⁹ CAD/CAM provides healthcare professionals with a streamlined method for designing and manufacturing highly precise facial implants. Moreover, these technologies reduce surgical duration and help future specialists establish clear treatment objectives for patients. This study explores manufacturing technologies for customized implants, emphasizing the benefits of three-dimensional printing, including increased time efficiency, accuracy, and production speed. Despite these advances, the design and fabrication of facial implants remain complex due to the need for multiple measurements and extended waiting periods before final surgery.⁸⁰

Preoperative planning, image processing, and implant design and production are complex, costly, and time-consuming processes. To address these challenges, solutions such as image analysis algorithms have been developed to support each stage of the process. Collaboration between CAD-CAM designers and medical professionals ensures precise and patient-specific facial implant designs. Advanced design software with user-friendly interfaces can further reduce planning and design durations.⁸¹ After finalizing the design, additive manufacturing methods such as 3D printing are used for production. Artificial intelligence optimizes fit, material usage, and post-processing, thereby reducing human error and shortening turnaround times.⁸²

Studies by Pathak et al.⁴⁵ and Kurt et al.⁴⁶ demonstrate that AI-driven imaging and deep learning techniques enhance the precision of maxillofacial prosthesis production and silicone tinting. Similarly, Patcas et al.⁴¹ and Rosero et al.³⁸ highlight AI's potential to objectively assess facial attractiveness and lip symmetry, supporting surgeons in pre- and post-operative evaluations. Pathak et al.⁴⁵ report that integrating AI with implant-supported prosthesis manufacturing has transformed facial reconstruction, especially in replicating the complex anatomy of the ear. AI algorithms enable accurate reproduction of facial features through comprehensive scanning and analysis, thereby restoring symmetry and individuality to patients' faces.⁸³ Kurt et al.⁴⁶ identify the Gated Recurrent Units (GRU) model as an effective deep learning method for improving the coloration of maxillofacial prostheses.⁴⁶

4.1.9. Aesthetic and functional outcomes

Beyond structural restoration, facial prosthetics must also achieve both aesthetic harmony and functional integration. Artificial intelligence tools evaluate facial proportions, skin tone, and symmetry using facial recognition algorithms. These assessments facilitate the creation of prostheses that closely resemble natural features, enhancing patient acceptance and overall quality of life.⁸⁴ Additionally, AI models simulate stress distribution and soft tissue interactions, which improve prosthesis retention and reduce complications such as irritation or displacement.⁸⁵ Rosero et al. conducted a study employing a deep learning approach to assess lip symmetry in individuals with repaired cleft lips. The CLP transformation, designed to mimic unilateral lip asymmetry, showed a stronger correlation with human assessments and achieved a weighted categorization accuracy of 75.34 % for evaluating lip symmetry.³⁸

4.1.10. Chatbots to patient inquiries about dental and maxillofacial prostheses

All chatbots provided accurate information; however, the domain-specific GPT generated more detailed responses than general-purpose chatbots. In contrast, Dental GPT showed lower readability, indicating a need to improve response clarity. The chatbots evaluated were highly beneficial for patients and demonstrated strong quality and coherence.⁵⁹ Natural language processing (NLP) and large language models (LLMs), such as ChatGPT, have also been studied for their roles in educational support and caregiver assistance. Chaker et al.⁵⁸ and Mahendia et al.⁶⁰ found that these models produce explicit educational content for parents and patients, which helps reduce caregiver burden. Shirk et al. advanced the application of LLMs by developing a modified Whisper-based model to identify velopharyngeal dysfunction, achieving higher accuracy than traditional machine learning methods.³⁹

Boyaci et al. employed deep learning to develop a computerized model that generates realistic facial images customized to each patient's unique features. This method helps surgeons with better planning tools and provides educational resources for patients.⁴³ Additionally, AI language models have also been created to tackle the challenges faced by caregivers of children with cleft lip and palate.⁵⁸

4.1.11. AI in cleft lip and palate

AI-powered tools support speech analysis, facial symmetry evaluation, and treatment monitoring. These applications enable more personalized and efficient care for patients with cleft lip and palate (CLP). Continued advancements in AI are expected to foster greater collaboration among specialties involved in cleft care, thereby enhancing both functional and aesthetic results.^38,86,87

The investigation of innovative AI frameworks, such as Vision Transformers (ViTs) used alongside Siamese Neural Networks by Nantha et al., showcases a new strategy for complex image classification in cleft diagnosis. This framework has performed exceptionally well in situations with limited data, indicating its future applicability in low-resource environments.³⁴ To forecast the maxillary expansion technique pertinent to cleft lip and palate, Huqh et al. employed binary logistic regression (BLR) using R syntax.³⁴ Patcas et al. conducted a study assessing facial attractiveness using neural network algorithms for patients with cleft lip and palate.⁴¹ Mathad et al. applied a Deep Learning Algorithm for the objective evaluation of hypernasality in children affected by cleft palate.³⁷ Shrik et al. explored the use of large language models for the automated identification of velopharyngeal dysfunction in cleft palate patients.³⁹ Additionally, Arslan et al. demonstrated that deep learning models could be utilized to detect teeth in individuals with cleft lip and palate.⁷² Despite the encouraging results, numerous challenges persist. For instance, some research involving smartphone-based 3D scanning indicates promise but still needs further validation within clinical settings.⁴² Additionally, aspects such as model interpretability, potential biases within datasets, and the ability of AI models to generalize across diverse ethnic and geographic populations must be carefully evaluated before widespread clinical implementation.

5. Limitations and future perspectives

Several limitations should be considered when interpreting this review. The scoping review methodology limits the ability to assess the quality of included studies, which may affect the credibility of the identified AI methods. Restricting the review to published research introduces publication bias, as studies with negative or inconclusive results might be underrepresented. Variability in study designs and AI methodologies can lead to an incomplete understanding of the research landscape. Moreover, the clinical significance of identified AI methods remains unclear due to a lack of longitudinal studies and standardised guidelines. The heterogeneity of studies and the absence of large-scale clinical trials further restrict their generalizability.

Integrating artificial intelligence with virtual and augmented reality can significantly improve the planning and visualization of prosthetic outcomes during surgical procedures. Furthermore, AI-driven robotics can help automate impression-taking and streamline the delivery of prosthetics. The clinical integration of AI will depend on developing comprehensive databases, encouraging interdisciplinary collaboration, and maintaining ethical standards. These findings lay the groundwork for future AI-based tools designed to improve access to diagnostic and therapeutic services for individuals with cleft-related velopharyngeal dysfunction. As technological capabilities grow, the dental profession must actively address ethical challenges while embracing innovative solutions. This involves ongoing research to develop robust security protocols, minimize algorithmic bias, clarify professional responsibilities, and protect patient privacy.

6. Conclusion

The integration of artificial intelligence into maxillofacial prosthodontics marks a significant advancement in clinical practice, research, and patient care. This review demonstrates that AI significantly enhances diagnostic imaging, the recognition of craniofacial anomalies, genetic risk evaluation, surgical simulation, treatment planning, and the design and manufacture of prostheses using CAD/CAM technologies. Furthermore, AI-driven tools have improved patient communication and support for caregivers. AI has also enabled the systematic discovery of biomaterials, accelerating progress in the field.

Although the benefits of AI are apparent, such as improved diagnostic accuracy, shorter treatment times, greater customisation of prosthetics, and more consistent treatment results, these systems should support rather than replace the expertise of prosthodontists. Clinical insight, compassion, and complex decision-making by human practitioners remain vital for comprehensive patient care.

However, ongoing challenges remain, such as the need to standardize data, validate findings across different populations, address ethical concerns regarding patient confidentiality, and train clinicians in AI-assisted procedures. Addressing these issues is essential to ensure the responsible, ethical, and equitable use of AI in maxillofacial prosthodontics.

Ongoing collaboration among clinicians, engineers, data scientists, and researchers will be essential for advancing AI systems and turning laboratory innovations into routine clinical practice. Through continuous research and careful integration, AI has the potential to improve clinical efficiency and enhance the quality of care, making maxillofacial prosthodontic rehabilitation more precise, patient-centered, and accessible.

Patient's/guardian's consent

Patient/Guardian's consent is not applicable, as it is a review article.

Authors’ contributions

Study conception and design: AA, RMB, SK, SHK, PCK.

Data collection: AA, RMB, SK, PCK.

Analysis and interpretation of results: AA, RMB, SK, SHK, PCK.

Draft manuscript preparation: AA, RMB, SK, SHK, PCK, RKN, GS, NVK.

All authors reviewed the results and approved the final version of the manuscript.

Ethical clearence

Not required Patient Consent- Not required.

Funding

The authors declare no sources of funding.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.