Exploring the Potential of Artificial Intelligence in Oral Diseases

Komal Vyas

, Devesh Nawgaje

and Ashwini Deshmukh

Institute Shri Sant Gajanan Maharaj College of Engineering Shegaon, India

Department of Electronics and Telecommunication Engineering, India

Keywords: Artificial Intelligence, CNN: Convolutional Neural Network, Oral Diseases, ML: Machine Learning, TMD:

Temporomandibular Joint Disorders, DL: Deep Learning

Abstract: Integration of AI (Artificial Intelligence) into healthcare has revolutionized numerous medical fields,

including dentistry. This paper explores the potential of AI in the diagnosis, treatment, and management of

oral diseases. We review current AI applications in oral pathology, including the application of ML

(machine learning) algorithms for early detection of oral cancers, dental caries, periodontal diseases, and

other oral conditions. Additionally, we examine how AI-driven tools enhance diagnostic accuracy, improve

patient outcomes, and streamline clinical workflows. The paper also discusses the challenges with the

deployment of AI technologies in dentistry, such as data privacy, bias in AI models, and the need for

standardization. Through a comprehensive analysis of recent advancements and case studies, this study

highlights the transformative impact of AI on oral health and underscores the need for continued research

and collaboration between technologists and dental professionals to fully realize its potential.

1 INTRODUCTION

Dentistry is a field which is undergoing significant

transformation with the advent of Artificial

Intelligence (AI), a technology that is reshaping

various aspects of healthcare. AI, has ability to

analyse large datasets, recognize patterns, and make

predictions which is increasingly being leveraged to

enhance the diagnosis, treatment, and management

of oral diseases. Advancements in the area of oral

health have far-reaching implications, and primary

reason behind this is the vital role that is being

played by the oral health when considering the well-

being of an individual.

Oral diseases under which we have considered

dental caries, periodontal disease, oral cancers and

TMD, remain prevalent globally and pose

substantial challenges to healthcare systems.

Traditional diagnostic methods, though effective,

often rely heavily on the expertise of dental

professionals and can be time-consuming and

subject to human error. AI offers a promising

alternative by providing tools that can support

clinicians in making more accurate and timely

diagnoses, personalizing treatment plans, and

predicting disease outcomes.

This paper seeks to delve into the burgeoning

potential of AI in the domain of oral health. We will

examine the current AI applications in dentistry,

which is image recognition for the detection of oral

diseases at early stage, predictive analytics for

patient risk assessment, and AI-assisted treatment

planning. Additionally, we will discuss the

limitations and ethical considerations associated

with AI in oral healthcare, including concerns

related to data privacy, the potential for algorithmic

bias, and the need for rigorous validation of AI

systems.

By exploring the current state of AI in oral

diseases, this study aims to present a comprehensive

overview of how AI is transforming the landscape of

dental care. We will highlight both the opportunities

and challenges that come with integrating AI into

dental practices. We will provide insights into the

future directions of research in this exciting and

rapidly evolving field.

2 ORAL IMAGING AND AI

INTEGRATION

The incorporation of AI (Artificial Intelligence) into

oral imaging represents one of the most

Vyas, K., Nawgaje, D. and Deshmukh, A.

Exploring the Potential of Artiﬁcial Intelligence in Oral Diseases.

DOI: 10.5220/0013594600004664

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 3rd International Conference on Futuristic Technology (INCOFT 2025) - Volume 2, pages 457-464

ISBN: 978-989-758-763-4

457

transformative advancements in modern dentistry.

Oral imaging, a cornerstone of diagnostic procedures

in dental practice, includes techniques such as

radiography, CBCT (cone-beam computed

tomography), and intraoral scanning. These imaging

methods are crucial for diagnosing a broad range of

oral conditions, from dental caries and periodontal

disease to complex maxillofacial abnormalities.

However, the interpretation of these images can be

highly subjective, requiring significant expertise and

experience. AI is now being utilized to enhance the

accuracy, efficiency, and consistency of image

analysis, thereby improving diagnostic outcomes.

2.1 AI in Radiographic Analysis

AI algorithms, particularly those based on deep

learning; demonstrate significant potential in the

analysis of radiographic images. These algorithms

are trained on labelled datasets of images, enabling

them to recognize patterns and anomalies that may

be unnoticeable to the human eye. For example, AI

systems have been developed to detect early signs of

dental caries, bone loss, and periapical lesions with

high accuracy. Studies have demonstrated that AI

can match or even surpass the diagnostic

performance of experienced radiologists in certain

tasks, such as identifying early-stage dental caries or

subtle fractures.

2.2 CBCT: Cone-Beam Computed

Tomography and AI

CBCT provides three-dimensional imaging, offering

detailed views of the maxillofacial region. This

modality is particularly useful in implant planning,

orthodontics, and the assessment of complex

anatomical structures. AI integration with CBCT has

enhanced the ability to automatically segment and

analyse anatomical features, such as the mandibular

canal or sinus cavities, reducing the time required

for manual analysis. AI algorithms can also assist in

identifying pathologies such as cysts, tumours, or

impacted teeth, improving the speed and accuracy of

diagnosis.

2.3 Intraoral Scanning and AI

Intraoral scanning has revolutionized restorative

dentistry by providing precise digital impressions of

the oral cavity. AI plays a crucial role in refining

these digital impressions, ensuring accurate margin

detection, bite alignment, and occlusal analysis. The

integration of AI with intraoral scanners allows for

real-time analysis, enabling dentists to make

immediate adjustments and reduce the need for

remakes. Additionally, AI-powered software can

predict the longevity of restorations by analysing

wear patterns and material properties, helping to

optimize treatment plans.

2.4 Benefits and Challenges

The integration of AI in oral imaging offers

numerous benefits, including enhanced diagnostic

accuracy, reduced human error, and increased

efficiency in clinical workflows. AI systems can

quickly process and analyse large volumes of

imaging data, providing dentists with detailed

insights that support better clinical decision-making.

Moreover, AI's capability to learn from new data

continuously enhances its diagnostic capabilities

over time.

3 ROLE OF AI IN PREDICTION

OR DIAGNOSIS OF ORAL

DISEASES

AI has demonstrated remarkable potential in the

early prediction and detection of various oral

diseases, offering a transformative impact on

preventive dentistry and early intervention. By

leveraging machine learning algorithms, neural

networks, and large datasets, AI systems can analyse

clinical and imaging data with remarkable accuracy,

aiding in the identification of oral conditions that

might otherwise go undetected until they have

progressed to more advanced stages.

3.1 Dental Caries

While considering the world wide predictions of, the

oral disease which is mostly seen is dental caries.

Dental caries are sometimes also termed as tooth

decay. Early detection is critical to preventing the

progression of caries and avoiding invasive

treatments. AI algorithms, particularly those based

on deep learning, have been successfully trained to

analyse radiographic images and identify carious

lesions with high precision. These systems can

detect early-stage caries that may be missed by the

naked eye, even in challenging areas such as the

interproximal spaces between teeth. By integrating

AI into routine dental check-ups, dentists can

improve their diagnostic accuracy and intervene

INCOFT 2025 - International Conference on Futuristic Technology

458

earlier, thus preserving more of the natural tooth

structure.

3.2 Periodontal Disease

Periodontal disease, encompassing conditions such

as gingivitis and periodontitis, can be considered as

one of the primary reason of tooth loss especially in

adults. It is also linked to systemic health issues such

as cardiovascular disease and diabetes. AI models

have been developed to predict the risk of

periodontal disease by analysing a combination of

patient data, including clinical history, lifestyle

factors, and genetic predispositions. Additionally, AI

can assist in the early detection of periodontal

disease by evaluating radiographs for signs of bone

loss, changes in the periodontal ligament space, and

other indicators of disease progression. This enables

personalized treatment plans and more effective

management of periodontal health.

3.3 Oral Cancer

Oral cancer is a critical and life-threatening illness

that is frequently identified in its advanced stages,

when treatment choices are limited and the outlook

is generally unfavourable. AI has demonstrated

potential in enhancing early detection of oral cancer

by analysing imaging data and histopathological

slides. Machine learning models can be trained to

identify subtle patterns and abnormalities in oral

tissues, which may signal the presence of

precancerous lesions or early-stage tumours. For

instance, AI-powered image recognition systems can

analyse photographs of oral mucosa or microscopic

images of tissue biopsies to identify abnormal cell

structures. Early detection facilitated by AI can

significantly improve survival rates by enabling

timely and targeted interventions.

3.4 Temporomandibular Joint Disorders

(TMD)

TMD involve the dysfunction of the jaw joint and

surrounding muscles, leading to pain, difficulty

chewing, and other complications. Diagnosing TMD

can be complex due to the wide range of symptoms

and potential underlying causes. AI systems have

been developed that help in the diagnosis and

management of TMD by analyzing patient

symptoms, medical history, and imaging data such

as MRI or CBCT scans. These systems can help

identify patterns indicative of specific TMD

subtypes, aiding in more accurate diagnoses and

personalized treatment strategies.

In this work we have reviewed papers from

2018 to 2024 on oral diseases including Dental

caries, Periodontal disease, oral cancer and TMD

and have tried to summarize the key points including

AI Technique used, data type used for training and

testing , AI architecture or model used, results

obtained, future scope, key findings, advantages,

limitations , parameters measured. This findings

made by us are summarized in the section of

literature review for oral diseases including Dental

caries, Periodontal disease, oral cancer and TMD.

Paper using Images for detection or prediction are

only included in this study.

4 LITERATURE REVIEW

4.1 Dental Caries

Lee et al. (2018) exhibited the effectiveness of

CNNs in detecting dental caries from intraoral

radiographs, and achieved high diagnostic accuracy

comparable to expert clinicians. Using a dataset of

3,000 images split into training/validation (2,400)

and test (600), the model highlighted the potential

for automating caries detection to improve

diagnostic workflows. While promising, the

approach relies on high-quality images and requires

further validation across diverse populations and

clinical settings (Lee, Kim, et al. , 2018). Casalegno

et al. (2019) used a custom CNN (convolutional

neural network) to analyse 217 grayscale NIR-TI

(near-infrared transillumination) images of molars

and premolars for detecting and localizing dental

caries. The model, validated with Monte Carlo

cross-validation, demonstrated the feasibility of non-

invasive, automated caries detection but faced

challenges with small dataset size, imaging

inconsistencies, and dependence on high-quality

hardware (Casalegno, Newton, et al. , 2019). M. T.

G. Thanh et al. (2022) evaluated deep learning

models like YOLOv3 and Faster R-CNN for

detecting cavitated and non-cavitated caries from

1,902 intraoral photos captured with a smartphone.

Use of YOLOv3 demonstrated the highest

sensitivity (87.4%) for cavitated caries, while

performance for non-cavitated caries was lower

across all models. The study highlights the potential

of smartphone-based diagnostics for improving

accessibility in low-resource settings but notes

limitations such as reduced sensitivity for non-

cavitated lesions and dependence on high-quality

Exploring the Potential of Artiﬁcial Intelligence in Oral Diseases

459

imaging conditions (Thanh, Toan, et al. , 2022).

Huang et al. (2021) utilized CNNs on OCT (Optical

Coherence Tomography) images to detect dental

caries, achieving high sensitivity and specificity,

surpassing conventional diagnostic methods.

Combining AI with OCT allowed for detailed

visualization of demineralized zones and early

detection of caries invisible to traditional techniques.

(Huang and Lee, 2021). Young et al. (2022)

combined U-Net for segmentation and Faster R-

CNN for object detection to analyse 2,348 intraoral

images, achieving high sensitivity and specificity for

detection of dental caries. The approach significantly

reduced false positives by isolating tooth surfaces

and minimizing background noise, outperforming

conventional diagnostic methods. (Park, Cho, et al. ,

2022). A Holtkamp et al.(2021)explored the use of

DL models, specifically CNNs, for detecting dental

caries in NILT images. Despite achieving strong

detection performance, the models faced challenges

in generalizing across diverse datasets and varying

imaging conditions, highlighting the need for more

diverse data. The key advantages include automated,

accurate caries detection, while limitations include

variability in data quality and imaging protocols,

affecting generalizability and performance. A. Tareq

et al. (2023) developed a hybrid YOLO ensemble

model with transfer learning, achieving high

sensitivity, specificity, and F1-scores (0.82–0.93) for

caries detection from non-standardized dental

photographs. The model demonstrated strong

performance across varying conditions, with real-

time detection capabilities ideal for clinical

workflows. Challenges include dependence on high-

quality annotated datasets and variability from non-

standardized imaging sources. (Tareq, Faisal, et al. ,

2023). Cascade R-CNN (Region-based

Convolutional Neural Network) analyzed 24,578

intraoral photographs for automatically recognizing

number of tooth and detecting dental caries, an

average mAP of 0.880 for tooth recognition and

0.769 for caries detection is achieved. The model

demonstrated high accuracy in caries localization

and staging, streamlining clinical workflows by

automating multi-tooth and multi-stage diagnosis.

Challenges include lower performance for certain

teeth, such as tooth 48, and reliance on high-quality

annotated datasets, with future plans to extend the

model for diagnosing other oral diseases and

improving generalizability (Yoon, Jeong, et al. ,

2024).

4.2 Periodontal Disease

In their 2023 study, J. Ryu et al. employed a Faster

R-CNN algorithm to analyse a dataset of 4,083

anonymized digital panoramic radiographs. These

radiographs were obtained from the Proline XC

machine, to identify periodontally compromised

teeth. The model demonstrated impressive

performance, achieving an Area Under the Curve

(AUC) of 0.88 for detecting periodontally

compromised teeth and 0.91 for overall detection,

which included edentulous regions. The study also

exhibited excellent consistency and reproducibility,

as evidenced by intraclass correlation coefficients

(ICC) of 0.94 for both inter- and intra-examiner

assessments. The results suggest potential for

automating periodontal disease diagnosis and

reducing human error, though high-quality

radiographs and a diverse dataset are essential for

optimal performance (Ryu, Lee, et al. , 2023). I.D.S.

Chen et al. (2023) applied the YOLOv7 algorithm

for object detection and a pre-trained EfficientNet-

B0 model for the classification of periodontal

diseases and dental caries in 1,525 periapical dental

X-ray images. The YOLOv7 model reached an

average precision of 97.1% for detecting teeth,

whereas the EfficientNet-B0 model achieved an

Area Under the Curve (AUC) of 98.67% for

identifying periodontitis and 98.31% for detecting

dental caries. The approach provides simultaneous

recognition of both conditions, offering improved

diagnostic support, though its performance is

dependent on high-quality X-ray images (Chen,

Yang, et al. , 2024). H. Amasya et al. (2023)

developed an AI system with two separate models

for detecting teeth and predicting periodontal bone

loss, using Mask R-CNN for tooth detection and

Cascade R-CNN for prediction of bone loss. Trained

on approximately 100 panoramic radiographs, the

model achieved high performance, with an F-score

of 0.948 for tooth detection and an F-score of 0.985

for detection of bone loss. The system showed high

accuracy and reliability, with Cohen's kappa

coefficients of 0.933 for tooth detection and 0.974

for bone loss detection, making it a promising tool

for dental diagnostics. (Amasya, et al. , 2024).

Kubilay Muhammed Sunnetc developed a hybrid AI

system that combines deep learning (CNNs) and ML

(machine learning) techniques to enhance the

detection of periodontal bone loss and the

classification of periodontitis stages. Trained on

1,432 panoramic radiographs with varying levels of

bone loss, the model utilized AlexNet and

SqueezeNet for feature extraction, achieving high

INCOFT 2025 - International Conference on Futuristic Technology

460

accuracy and strong F-scores for classification. The

system is user-friendly, enabling dental

professionals to efficiently assess periodontal health

from radiographs (Sunnetci, Ulukaya, et al. , 2022).

4.3 Oral Cancer

G. Tanriver et al. (2021) developed a DL (deep

learning) -based system with a two-stage pipeline for

detecting oral lesions and classifying them into

benign, oral potentially malignant disorders

(OPMDs), or carcinoma categories. The model,

using convolutional neural networks (CNNs) and a

pre-trained network like ResNet for feature

extraction, achieved over 95% accuracy and an AUC

of more than 0.97 for OPMD and carcinoma

classification. The system effectively detects early-

stage oral cancer and differentiates it from benign

conditions, demonstrating high sensitivity and

specificity (Tanriver, Tekkesin, et al. , 2021). S.

Krishna P et al. (2022) applied DL, particularly

CNNs, for classifying and segmenting oral lesions,

using histopathological images and digital

photographs. The model, utilizing ResNet-50 for

feature extraction and YOLO for real-time object

detection, achieved 92% classification accuracy for

oral squamous cell carcinoma (OSCC), with 93%

sensitivity and 89% specificity. The system's strong

F1-score demonstrated balanced precision and

recall, making it suitable for clinical use. The

model's potential for integration into real-time

diagnostic workflows and mobile applications is

promising for improving accessibility in resource-

limited settings (Krishna, Lavanaya, et al. , 2022).

K. Warin et al. (2022) utilized deep learning,

specifically deep CNNs, to detect oral cancers at

earlies stages, through feature extraction and

classification of oral lesion images. The model,

trained on a huge dataset of labelled

histopathological slides, oral cavity photographs,

and autofluorescent imaging, achieved 90-95%

accuracy, with sensitivity ranging from 88-93% and

specificity between 92-96%. The study highlighted

the model's potential for integration into clinical

workflows and real-time applications, with a focus

on minimizing false negatives for early detection.

(Warin, Limprasert, et al. , 2024). E.S. Mira (2024)

applied deep convolutional neural networks (CNNs)

such as DenseNet-169, ResNet-101, SqueezeNet,

and Swin-S for the classification of oral lesions,

including oral squamous cell carcinoma (OSCC),

oral potentially malignant disorders (OPMDs), and

non-pathological oral regions. The study used a

dataset of 980 annotated oral photographic images,

achieving promising classification results that

outperformed traditional diagnostic methods, though

specific metrics like sensitivity and specificity were

not detailed. The model's future scope includes

integration with clinical workflows, development of

public datasets for benchmarking, and expansion

with diverse imaging modalities. (Mira, Sapri, et al. ,

2024). K. Vinay Kumar (2024) utilized DL with

CNNs and advanced hybrid models like

InceptionResNetV2 for detection of oral cancer. For

the detection purpose they utilized histopathological

images sourced from public datasets, including

5,685 images (3,099 cancerous, 2,586 non-

cancerous). The model demonstrated high

classification accuracy, outperforming other models,

and showed strong potential for early diagnosis by

distinguishing between cancerous and non-

cancerous lesions. However, limitations include

dataset diversity and the need for extensive

validation before clinical use (Kumar, Palakurthy, et

al. , 2024). I. U. Haq et al. (2023) employed a hybrid

AI approach combining deep learning models

(CNNs like ResNet and Inception) with feature-

based machine learning techniques (e.g., SVMs and

random forests) for OSCC detection in

histopathological images. The model achieved high

accuracy, sensitivity, and specificity, significantly

reducing diagnostic time compared to manual

evaluations. The future scope includes expanding

datasets for diverse populations, integrating real-

time diagnostic tools, and exploring multimodal

data. The hybrid model demonstrated improved

diagnostic reliability and accuracy, offering a

scalable solution for OSCC diagnosis. (Haq, Ahmed,

et al. , 2023). H. Lin et al. (2021) used CNNs to

classify oral lesion images captured with

smartphones into five categories: normal, aphthous

ulcers, low-risk OPMD, high-risk OPMD, and oral

cancer. The system demonstrated high accuracy and

efficiency for early detection, making it a promising

solution for field use with no need for specialized

equipment. The smartphone-based approach offers a

cost-effective, portable, and accessible method for

oral cancer detection, though its performance is

dependent on image quality and lighting conditions

(Lin, Chen, et al. , 2021).

4.4 Temporomandibular Joint Disorders

(TMD)

E. Choi et al. (2021) used a ResNet-based CNN to

classify orthopantomograms (OPGs) into categories

of normal, indeterminate osteoarthritis (OA), and

Exploring the Potential of Artiﬁcial Intelligence in Oral Diseases

461

OA. The model initially struggled with multi-label

classification but improved when "indeterminate

OA" was reclassified as either normal or OA,

achieving 78% accuracy, 73% sensitivity, and 82%

specificity. This approach demonstrated diagnostic

performance comparable to expert radiologists and

could be integrated into clinical workflows as a cost-

effective screening tool for temporomandibular joint

osteoarthritis. (Choi, Kim, et al. , 2021). W.M.

Talaat et al. (2023) utilized a CNN with regression-

based object detection for analyzing CBCT images

from 943 patients. The model demonstrated higher

agreement with gold-standard references compared

to expert radiologists, improving diagnostic

accuracy for subcortical cysts and osteoarthritic

signs. The AI system showed potential for

standardizing temporomandibular joint (TMJ)

osteoarthritis diagnosis, reducing subjectivity, and

expediting CBCT scan analysis. (Talaat, Shetty, et

al. , 2023). Y.H. Lee et al. (2024) developed a DL

model using CNNs to detect temporomandibular

joint (TMJ) effusion from MRI images, with data

from 1,474 patients and 2,948 images. The model

employed the VGG16 architecture and was fine-

tuned for effective interpretation of PD: proton

density and T2W: T2-weighted MRI scans. While

the model demonstrated excellent specificity, it

showed lower sensitivity compared to human

experts, suggesting room for improvement. The

study highlighted the potential for further

optimization, including the use of different MRI

sequences or combining imaging modalities, to

enhance diagnostic performance and reduce reliance

on expert interpretation(Lee, Jeon, et al. , 2024).

T.Y. Su et al. (2024) applied CNNs for automatic

classification of temporomandibular joint disorders

(TMD) using MRI or CT scans. The study reports

high accuracy in distinguishing TMD cases from

healthy ones, often outperforming traditional

diagnostic methods. Sensitivity and specificity

metrics indicate the model's effectiveness in

correctly identifying true positives and true

negatives, respectively. The research suggests

incorporation of AI into clinical workflows and

emphasizes the potential of data augmentation and

multi-modal analysis (combining imaging with

clinical data) for better accuracy. While CNNs

provide efficient, non-invasive, and scalable

diagnostics, the model's generalizability is limited

by dataset quality, and concerns about the lack of

interpretability and hardware demands were noted.

Performance metrics included accuracy, sensitivity,

specificity, and recall (Su, Wu, et al. , 2024). K.S.

Lee et al. (2020) employed a deep learning-based

Single-Shot Detection (SSD) framework to detect

and classify osseous changes in temporomandibular

joints (TMJ) from CBCT images. The study used

3,514 sagittal CBCT images for training and two

independent test sets with 300 images for evaluation.

The model attained an accuracy of 86%, precision of

85%, and recall of 84%, demonstrating reliable

performance. It can efficiently classify TMJ osseous

changes, supporting early TMJ osteoarthritis

(TMJOA) diagnosis. Challenges include its reliance

on sagittal images, which may miss changes in other

planes, and observer bias in dataset labelling.

(Farook, Dudley, et al. , 2020)

5 OBSERVATIONS

The review of studies from 2018 to 2024

demonstrates a significant shift towards the

integration of AI, ML and DL techniques in

predicting and detecting various oral conditions,

including dental caries, periodontal disease, oral

cancer, TMD. Researchers have utilized a diverse

range of AI models, including CNNs, Random

Forests, Support Vector Machines (SVMs), and

Decision Trees, applied to various data types such as

clinical data, medical images, genetic profiles, and

patient history.

The studies reviewed demonstrate significant

advancements in the application of ML and DL

techniques in the field of dental diagnostics,

specifically for dental caries, periodontal disease,

oral cancer, and TMD. These approaches,

particularly CNNs and hybrid models, have shown

promising results in automating and enhancing the

accuracy of diagnostics using various imaging

modalities such as X-rays, CBCT, MRI, and

histopathological images.

For dental caries, CNNs and advanced object

detection models like YOLOv3 and Faster R-CNN

have proven to be highly effective, particularly in

detecting cavitated lesions and improving diagnostic

workflows. The integration of more diverse datasets

and advanced preprocessing techniques could

address these limitations.

In the realm of periodontal disease, models

utilizing Faster R-CNN, YOLOv7, and Mask R-

CNN have demonstrated high performance in

detecting periodontal bone loss and classifying

disease stages. These models have shown promise in

standardizing diagnoses and reducing human error.

However, as with caries detection, performance is

highly dependent on high-quality imaging and

INCOFT 2025 - International Conference on Futuristic Technology

462

diverse datasets, which are crucial for optimizing

generalization and real-world applicability.

For oral cancer, CNN-based models have

exhibited exceptional accuracy and sensitivity,

particularly in differentiating benign from malignant

lesions, with some models achieving over 95%

accuracy. However, challenges related to image

quality, dataset diversity, and model generalization

remains, necessitating the expansion of training

datasets and further validation in clinical settings.

The potential for real-time applications in resource-

limited environments is promising, particularly with

mobile solutions for early detection.

Lastly, in temporomandibular joint disorders

(TMD), CNNs have been successfully employed to

classify and diagnose TMJ disorders using CT and

MRI scans. These models have outperformed

traditional methods in terms of diagnostic accuracy

and efficiency, though integration of multimodal

data and addressing challenges related to data

quality, interpretability, and computational demands

are essential for improving model robustness

Table 1: Statistical Analysis.

Categor

Statistical Metrics

Overall Trends

2018-2024

80-90% of reviewed studies

reported improved diagnostic

accuracy using AI

techni

ues.

Dental Caries CNN models achieve 85-

95% accuracy in detecting

cavitated lesions.

Object detection models like

YOLOv3 and Faster R-CNN

demonstrate 10-15%

improvement in diagnostic

workflows efficienc

Periodontal Disease Faster R-CNN and YOLOv7

models show 90-95%

accuracy in detecting bone

loss and staging diseases.

Mask R-CNN achieves up to

20% reduction in human

dia

nostic errors.

Oral Cancer CNN models achieve

sensitivity and accuracy

exceeding 95% in

differentiating benign and

malignant lesions

Mobile-based solutions have

potential for early detection

in 60-80% of resource-

limited environments.

Temporomandibular

Joint Disorders

CNN models improve

diagnostic accuracy by 15-

25% compared to traditional

methods.

Efficiency gains of 10-20%

due to automated analysis

using CT and MRI scans.

6 CONCLUSION

In conclusion, while ML and DL models have

shown significant promise in automating dental

diagnostics, the future of these technologies depends

on overcoming challenges related to data quality,

model generalization, and integration into real-world

clinical workflows. The continued expansion of

diverse datasets, along with the incorporation of

multimodal data, will be key to further enhancing

the accuracy, accessibility, and real-time

applicability of these AI-driven diagnostic tools.

7 FUTURE SCOPE

The future of AI in dental diagnostics lies in

improving model generalization and real-world

applicability. Expanding datasets with diverse

patient demographics, imaging modalities, and

clinical data is essential for developing more robust

models. Emphasis should be placed on multimodal

integration, where combining different data sources,

such as radiographs, clinical records, and advanced

imaging techniques, can enhance diagnostic

accuracy. Additionally, future research should focus

on creating real-time, point-of-care diagnostic tools,

especially for underserved regions, and ensuring

seamless integration of AI systems into clinical

workflows. Addressing challenges like model

interpretability, bias mitigation, and clinical

validation through large-scale studies will be crucial

to building trust and ensuring the effectiveness of

AI-driven solutions in routine practice. Ultimately,

the focus will be on developing early detection

models, improving diagnostic precision, and

ensuring equitable access to advanced dental care

technologies.

REFERENCES

Lee, J.-H., Kim, D.-H., Jeong, S.-N., & Choi, S.-H.

(2018). Detection and diagnosis of dental caries using

a deep learning-based convolutional neural network

algorithm. ScienceDirect, 77, 106–111

Casalegno, F., Newton, T., Daher, R., Abdelaziz, M.,

Lodi-Rizzini, A., Schürmann, F., Krejci, I., &

Markram, H. (2019). Caries detection with near-

Exploring the Potential of Artiﬁcial Intelligence in Oral Diseases

463

infrared transillumination using deep learning. Journal

of Dental Research, 98(11), 1227–1233.

Thanh, M. T. G., Toan, N. V., Ngoc, V. T. N., Tra, N. T.,

Giap, C. N., & Nguyen, D. M. (2022). Deep learning

application in dental caries detection using intraoral

photos taken by smartphones. Applied Sciences,

12(11), 5504. https://doi.org/10.3390/app12115504

Huang, Y.-P., & Lee, S.-Y. (2021). Deep learning for

caries detection using optical coherence tomography.

medRxiv. Preprint, May 23.

Park, E. Y., Cho, H., Kang, S., Jeong, S., & Kim, E.-K.

(2022). Caries detection with tooth surface

segmentation on intraoral photographic images using

deep learning. BMC Oral Health, 22, 573.

Holtkamp, A., Elhennawy, K., Cejudo Grano de Oro, J. E.,

Krois, J., Paris, S., & Schwendicke, F. (2021).

Generalizability of deep learning models for caries

detection in near-infrared light transillumination

images. Journal of Clinical Medicine, 10(5), 961.

Tareq, A., Faisal, M. I., Islam, M. S., Rafa, N. S.,

Chowdhury, T., Ahmed, S., Farook, T. H.,

Mohammed, N., & Dudley, J. (2023). Visual

diagnostics of dental caries through deep learning of

non-standardized photographs using a hybrid YOLO

ensemble and transfer learning model. International

Journal of Environmental Research and Public Health,

20(7), 5351.

Yoon, K., Jeong, H.-M., Kim, J.-W., Park, J.-H., & Choi,

J. (2024). AI-based dental caries and tooth number

detection in intraoral photos: Model development and

performance evaluation. Journal of Dentistry,

141,Article 014821.

Ryu, J., Lee, D.-M., Jung, Y.-H., Kwon, O., Park, S.,

Hwang, J., & Lee, J.-Y. (2023). Automated detection

of periodontal bone loss using deep learning and

panoramic radiographs: A convolutional neural

network approach. Applied Sciences, 13(9), 5261.

Chen, I. D. S., Yang, C.-M., Chen, M.-J., Chen, M.-C.,

Weng, R.-M., & Yeh, C.-H. (2023). Deep learning-

based recognition of periodontitis and dental caries in

dental X-ray images. Bioengineering, 10(8), 911.

https://doi.org/10.3390/bioengineering10080911

Amasya, H., et al. (2024). Development and validation of

an artificial intelligence software for periodontal bone

loss in panoramic imaging. International Journal of

Imaging Systems and Technology, 34, e22973.

Sunnetci, K. M., Ulukaya, S., & Alkan, A. (2022).

Periodontal bone loss detection based on hybrid deep

learning and machine learning models with a user-

friendly application. Biomedical Signal Processing

and Control, 77, 103844.

Tanriver, G., Tekkesin, M. S., & Ergen, O. (2021).

Automated detection and classification of oral lesions

using deep learning to detect oral potentially

malignant disorders. Cancers, 13(11), 2766.

Krishna, S. P., & Lavanaya, J. (2022). Oral cancers

diagnosis using deep learning for early detection. In

Proceedings of the IEEE Conference ICEARS.

Warin, K., Limprasert, W., Suebnukarn, S., Jinaporntham,

S., Jantana, P., & Vicharueang, S. (2024). AI-based

analysis of oral lesions using novel deep convolutional

neural networks for early detection of oral cancer.

PLOS ONE.

Mira, E. S., Sapri, A. M. S., Aljehani, R. F., Jambi, B. S.,

Bashir, T., El-Kenawy, E.-S. M., & Saber, M. (2024).

Early diagnosis of oral cancer using image processing

and artificial intelligence. Vol. 14(01), 293–308.

Kumar, K. V., Palakurthy, S., Balijadaddanala, S. H.,

Pappula, S. R., & Lavudya, A. K. (2024). Early

detection and diagnosis of oral cancer using deep

neural network. Journal of Computer Allied

Intelligence, 2(2), 22–34.

]Haq, I. U., Ahmed, M., Assam, M., Ghadi, Y. Y., &

Algarni, A. (2023). Unveiling the future of oral

squamous cell carcinoma diagnosis: An innovative

hybrid AI approach for accurate histopathological

image analysis. IEEE Access, 11, 118281–118290.

Lin, H., Chen, H., Weng, L., Shao, J., & Lin, J. (2021).

Automatic detection of oral cancer in smartphone-

based images using deep learning for early diagnosis.

Journal of Biomedical Optics, 26(8), 086007-1.

Choi, E., Kim, D., Lee, J.-Y., & Park, H.-K. (2021).

Artificial intelligence in detecting temporomandibular

joint osteoarthritis on orthopantomogram. Scientific

Reports, 11, 10246.

Talaat, W. M., Shetty, S., Al Bayatti, S., Talaat, S.,

Mourad, L., Shetty, S., & Kaboudan, A. (2023). An

artificial intelligence model for the radiographic

diagnosis of osteoarthritis of the temporomandibular

joint. Scientific Reports, 13, 15972.

Lee, Y.-H., Jeon, S., Won, J.-H., Auh, Q.-S., & Noh, Y.-

K. (2024). Automated detection and visualization of

temporomandibular joint effusion with deep neural

network. Scientific Reports, 14, 18865.

https://doi.org/10.1038/s41598-024-69848-9

Su, T.-Y., Wu, J. C.-H., Chiu, W. C., Chen, T.-J., Lo, W.-

L., & Lu, H. H.-S. (2024). Automatic classification of

temporomandibular joint disorders by magnetic

resonance imaging and convolutional neural networks.

Journal of Dental Sciences.

Farook, T. H., & Dudley, J. (2020). Understanding

occlusion and temporomandibular joint function using

deep learning and predictive modeling. Journal of

Dental Research, 99, 1–5.

INCOFT 2025 - International Conference on Futuristic Technology

464