Review Article

ARTIFICIAL INTELLIGENCE IN ORAL PATHOLOGY

PRACTICE– AN OVERVIEW
Nagjyothi Kariamal1, Punnya Vaijanath Angadi1*
1
Department of Oral Pathology and Microbiology, V.K. Institute of Dental Sciences, KLE Academy of Higher Education and Research (KAHER), Belgaum-590010, Karnataka,
India. [email protected]

https://doi.org/10.51847/AQAJtO9N1U

ABSTRACT
Despite more recent medical achievements and a significant amount of information on various diseases, it is impossible to
anticipate disease diagnosis and progress accurately. The advent of artificial intelligence (AI) has opened up many new
possibilities for healthcare improvement and has ushered in a new era of increased exactitude in pathology. Artificial
intelligence (AI) can be helpful in the diagnosis of diseases, in making prognostications, or in creating patient-specific
treatment plans. AI can help pathologists in particular when they need to make important judgements quickly. It can
eliminate human mistake from the judgement process, resulting in improved and standardised health care while
diminishing the strain on the doctor. Pathologists are going to employ AI to find specific imaging indicators connected to
disease processes in order to improve early diagnosis, ascertain prognosis, and select the treatments that are most likely to
be successful. In the upcoming years, this trend is frequently anticipated to continue and change the pathology landscape.
Given the ageing population and rising patient volume, as well as the fact that there is a dearth of pathologists globally,
this is especially crucial. Even though AI models have been effective, it has taken a while for them to be translated from
research to clinical use. AI and pathologists working together can produce outcomes that are superior to what humans are
capable of in terms of accuracy, consistency, timeliness, and utility. This review presents the components, emerging
techniques and applications of AI in oral pathology.

Key words: Artificial intelligence, Pathology, Digital pathology, Oral pathology.

Introduction spatial relationships, which allowed categorization of

different cells present in the image [3-5].
The amalgamation of artificial intelligence (AI) in
healthcare will be the biggest revolution in medicine and McCarthy et al. first adopted the phrase "artificial
pathology, right at the center of this development. The intelligence" (A.I.) in the 1950s to refer to computer science,
earliest change evidenced over the years includes the large where machine-based techniques are used to simulate what
reliance on digital pathology, which is becoming a a rational person would do in a certain circumstance, such as
progressively more imperative technological requisite in making a prediction [6-8].
laboratory settings in most countries. Pathologists can now
manage digital data owing to advances in computer Artificial intelligence (AI) has been cited as being helpful in
resources, faster networks, and more affordable storage. the diagnosis of diseases, in making prognostications, or in
Slide photos can now be shared for telepathology and creating patient-specific treatment plans. AI can help
clinical use with greater simplicity and versatility than in pathologists, particularly when they need to make important
previous years [1, 2]. However, the growth evidenced in judgments quickly. It can eliminate human mistakes from
pathology has not extended to oral pathology, with digital the judgment process, resulting in improved and
pathology still in its infancy here. standardized health care while diminishing the strain on the
doctor [9]. This review outlines the components, recent
Digital pathology involves the creation of virtual slides using applications, and advancements of artificial intelligence in
Whole-slide scanners, and these digitized whole-slide digital oral pathology.
images are analyzed using computer methods. The whole
slide scanners came into existence more than two decades Units of artificial intelligence
earlier. Still, the origins of the digitalization of pathology
date back to 1960, when Prewitt and Mendelsohn established Machine learning
a system that scanned the microscopic images of a Any Pathology AI (Artificial Intelligence) system must have
conventional blood smear preparation and converted these machine learning, which enables a task to be learned from
images based on their optical density values into various data (without being directly designed). Machine learning
techniques ' primary goal is creating a model that may be

© 2023 Annals of Dental Specialty. Open Access – This Article is licensed under CC BY NC SA 4.0. To view a copy of this license, visit
https://creativecommons.org/licenses/by-nc-sa/4.0/ 82
Kariamal and Angadi

used for categorization, prediction, approximation, or any times and reduced errors during specimen handling, thus
other similar activity. enhancing the efficiency of the services.

ML techniques fall into two categories: supervised learning Artificial intelligence in clinical oral pathology
and unsupervised learning. In supervised learning, the input AI can be integrated with clinical data, patient demography,
data are approximated or matched to the intended output and histopathological and genetic data that can efficiently
using a labeled training data set. Unsupervised learning and accurately diagnose. Therefore, it is imperative to adopt
techniques don't give out labeled samples or know the results a more encompassing approach with input from A.I.
after completing the learning. It is up to the training instances researchers, pathologists, and other participants like
or model to identify the groups or characteristics in the oncologists and surgeons while training the algorithms [14,
incoming data. Semi-supervised learning combines 15].
supervised and unsupervised learning, is another class of ML
techniques that has seen extensive use. To build a precise An effective screening test that can be applied in primary
learning model, it integrates labeled and unlabelled data. care settings is needed to improve the effectiveness of
This type of learning is typically employed if there are more secondary or tertiary care referrals. This should distinguish
unidentified datasets than identified [10, 11]. the commonly encountered benign lesions without any
malignant potential and potentially malignant oral lesions
Deep learning (PMOLs) at "risk" for becoming cancerous or those with a
Deep learning, which is accurately referred to as a subfield heightened risk for cancer progression [16].
of machine learning, is thought to be the most recent
evolution of machine learning. All of the preceding machine A minimally invasive brush cytology sample has been
learning applications can be achieved with deep developed to analyze cell specimens using the Point-of-Care
understanding. Still, it also has more sophisticated Oral Cytology Tool. The cell suspension obtained is
capabilities, decision-making abilities, and the ability to subjected to a sophisticated image recognition program and
handle big datasets. Convolutional, pooling and fully linked pattern recognition techniques aided by cutting-edge
layers make up CNN compositions. The convolutional statistical techniques, enabling the concurrent quantitative
layer's main function is recognizing patterns, lines, and other measurement of cell morphologic information and the
features like edges. Convolutional layers in each of CNN's appearance of molecular biomarkers of malignant potential
hidden layers convolve the input array with weight- in an automated manner. This advanced procedure produces
parameterized convolution cores. The cores provide cytology findings quickly instead of the typical
numerous feature images and enable success in various histopathology preparation of slides that may take days,
vision tasks, including segmentation and classification [12, making it appropriate for screening. This test will allow
13]. dental professionals and primary care dentists to avoid the
need for numerous referrals and appointments before
Artificial intelligence in oral pathology workflow acquiring an assessment of the molecular risk of PMOL,
In routine clinical practice, pathologists make histological which is likely to affect the management of the disease
diagnoses by analyzing the slide for its morphological significantly [16].
features, semi-quantifying the relevant components, and
visually recognizing the pathology. However, this The Mobile Mouth Screening Anywhere (MeMoSA) app,
histopathological analysis has inherent subjectivity, and developed by Kingston University of UK and University of
considerable differences may be observed in visual Malay scientists, takes pictures of the mouth for expert
perception and judgment among individual oral pathologists, remote interpretation. Following the training of a Deep
leading to variation and lack of standardization in diagnosing learning algorithm that can distinguish between photos with
wide-ranging lesions. Further, this problem is intensified by and without oral cancer symptoms in dozens of images, the
the emergence of non-invasive and minimally invasive app has included the algorithm. The implementation of
biopsies, which are presently in trend that reduce the size and Artificial intelligence into MeMoSA, according to Professor
quality of samples received. The introduction of numerous Dr. Sok Ching Cheong of Cancer Research Malaysia, offers
immunohistochemical markers and genomics components huge potential in assuring that the measures allow for timely
for diagnosing various lesions adds to the dilemma due to identification in areas that show high prevalence of this
the lack of uniform reporting guidelines. Thus, A.I.-based disease [17-19].
approaches, thought to be more robust and reproducible,
may serve as a foundation for assuaging these challenges Artificial intelligence in oral histopathology
faced by pathologists [1, 2]. Some applications for A.I. has been demonstrated to increase agreement among
improving the workflow include quality control and quality pathologists in a variety of contexts, including the evaluation
assurance, improving the pathology workstream, triaging of certain subjective features like nuclear and cellular
cases into high/low priority, and automated requests for pleomorphism in the cells and degree of dysplasia, amount
additional investigations. This may lead to faster processing of cellularity, and assessment of mitotic figures, grading of

Annals of Dental Specialty Vol. 11; Issue 3. Jul – Sep 2023 | 83

Kariamal and Angadi

tumors infiltrating lymphocytes, and assessment of Shabana et al. (2019) created a special Deep-learning
proliferation using Ki-67 levels [14]. method to measure cancer infiltrating lymphocytes utilizing
whole slide images of oral Squamous cell carcinoma. They
Das et al. presented a two-stage method for computing oral reached an efficiency of ninety-six percent.
histology images using a 12-layered deep convolution neural
network (CNN) that separates component levels in the Eventually, all morphological traits detected individually by
primary and secondary stages. From these isolated keratin pathologists should be combined into a single classification
areas, developed random forests of the Gabor filter are system to describe the type and behavior of tumors. These
employed to locate the keratin pearls. When utilizing these may offer crucial directions for patient care and targeted
random forest classifiers to recognize keratin pearls, therapy heading in the future [21].
detection accuracy was reported to be 96.88 percent overall
[17, 20]. Limitations and challenges in adoption of AI in oral
pathology
A Convolutional neural network-operated automatic Adopting AI image models in everyday oral pathology
computer-aided hyperspectral image detection method was practice has taken longer than anticipated despite their
used to classify OSCC by Jeyaraj and Nadar. OSCC vs. preliminary success in pathology and other medicinal areas.
benign lesions image classification accuracy for this system The 'black box' nature of these models, which makes it
was 91 percent overall, and OSCC vs. normal tissue image difficult to comprehend why the algorithm generates certain
classification accuracy was 95 percent overall [17, 21]. estimates, is probably one of the main causes of this issue.
Algorithm developers are attempting to incorporate
Only a few research have examined oral squamous cell feedback from the algorithms' customers to boost the
carcinoma digital histopathologic images for machine acceptance of A.I. applications for regular use. Although the
learning techniques. A computer-assisted histomorphometry expansion of the pathologic sector in A.I. to include cancer
classifier based on nuclear morphology was developed by Lu severity evaluation and prognosis prediction is intriguing, a
et al. (2017). By using a digitized tissue microarray to significant amount of data is still required to create A.I. that
analyze 2mm sections of OSCC (115), an ML classifier was may be used in various healthcare settings. It is still
able to identify cases at high and moderate risk for disease- necessary to assess the medical relevance of integrating A.I.
specific survival with an AUC (0.72). This study yielded a technologies into the histopathological practice and its
low specificity (71%) and sensitivity (62%), but it impact on the outcomes, as well as their pricing system [14,
demonstrated the potential of analyzing even small amounts 29-32].
of tissue [21, 22].
Further, the perception that A.I. may replace the pathologists
Using AI-based techniques on potentially malignant oral lingers in many minds. However, this notion is not
disorders offers tremendous potential to improve OSCC particularly true since the final say will always be with the
early detection. Shamim et al. 2019 trained deep-learning pathologist. However, AI can assist in enhancing the
convolutional networks for classification tasks using an accuracy and speed of diagnosis in heavy workflow
annotated information base of clinical and histopathological conditions where the pathologist may get overwhelmed by
photos of potentially malignant and benign tongue lesions. the sheer amount of work, leading to inadvertent errors.
In this investigation, their model was able to differentiate
between potentially malignant and benign tongue lesions Conclusion
with a mean classification accuracy of 0.98 [21, 23].
The creation and application of Automated tools, such as
Artificial intelligence in prognosis in oral pathology image-based algorithms that are utilized in pathology, has
Many studies have been conducted about prognostic systems increased significantly, and it is anticipated that they will
for anticipating the survival of patients and local recurrence soon rule the area of oral pathology, too. AI and pathologists
in OSCC patients in order to improve the prognostication of working together can produce outcomes that are superior to
oral cancer [21, 24-27]. what humans are capable of in terms of accuracy,
consistency, timeliness, and utility. Artificial intelligence
These models examined huge information repositories of applications will also enable more sophisticated diagnostics,
Oral cancer patients using AI-based methods. Still, the scope allowing developers and healthcare practitioners to
of the data analysis was restricted to demography, clinical exchange expertise and employ automated systems to
correlation, or genetic data. In 2013, Chang et al. created an evaluate and offer insightful contributions that could
OSCC prognostication model that included genomic data of eventually result in a more thorough pathology diagnosis.
p53 and p63 outcomes from IHC slides and The future of precision oncology will benefit from this
clinicopathologic material. Their model, which used combination, which has the potential to produce
machine learning techniques, produced an AUC (0.90) in the personalized care plans.
prognosis of OSCC depending on the presentation,
marker(P63), and alcoholic intake [21, 28]. Acknowledgments: None

Annals of Dental Specialty Vol. 11; Issue 3. Jul – Sep 2023 | 84

Kariamal and Angadi

Conflict of interest: None artificial intelligence in pathology: a clinical

perspective. J Clin Pathol. 2021;74(7):409-14.
Financial support: None 15. Ali HR, Dariush A, Provenzano E, Bardwell H,
Abraham JE, Iddawela M, et al. Computational
Ethics statement: None pathology of pre-treatment biopsies identifies
lymphocyte density as a predictor of response to
References neoadjuvant chemotherapy in breast cancer. Breast
Cancer Res. 2016;18(1):21.
1. Niazi MKK, Parwani AV, Gurcan MN. Digital 16. McRae MP, Modak SS, Simmons GW, Trochesset
pathology and artificial intelligence. Lancet Oncol. DA, Kerr AR, Thornhill MH, et al Point-of-care oral
2019;20(5):e253-e61. cytology tool for the screening and assessment of
2. Farahani N, Parwani AV, Pantanowitz L. Whole slide potentially malignant oral lesions. Cancer Cytopathol.
imaging in pathology: advantages, limitations, and 2020;128(3):207-20.
emerging perspectives. Pathol Lab Med Int. 17. Krishna AB, Tanveer A, Bhagirath PV, Gannepalli A.
2015;7:23-33. Role of artificial intelligence in diagnostic oral
3. Prewitt JM. Intelligent microscopes: recent and near- pathology-A modern approach. J Oral Maxillofac
future advances. InRecent and Future Developments in Pathol. 2020;24(1):152-6.
Medical Imaging II 1979 Dec 26 (Vol. 206, pp. 206- 18. Park A. Researchers Training AI Mobile App to Detect
219). SPIE. Early Signs of Oral Cancer. Available from:
4. Prewitt JMS. Parametric and nonparametric https://oralcancernews.org/wp/researchers training ai
recognition by computer: an application to leukocyte mobile app to detect early signs of oral cancer/. [Last
image processing. Adv Comput. 1972;12:285-414. accessed on 2022 Sep 30].
5. Prewitt JM, Mendelsohn ML. The analysis of cell 19. Can Artificial Intelligence Spot Warning Signs of Oral
images. Ann N Y Acad Sci. 1966;128(3):1035-53. Cancer? Kingston University Experts Develop System
6. McCarthy J, Minsky ML, Rochester N, Shannon CE. for Mobile Phone App in New Study. Available from:
A proposal for the Dartmouth summer research project https://www.kingston.ac.uk/news/ article/2201/04 jun
on artificial intelligence. AI Mag. 2006;27(4):12-6. 2019 can artificial intelligence spotwarning signs of
7. McCarthy JJ, Minsky ML, Rochester N. Artificial oral cancer kingston universityexperts develop system
intelligence. Research Laboratory of Electronics for/. [Last accessed on 2022 Sep 30].
(RLE) at the Massachusetts Institute of Technology 20. Das DK, Bose S, Maiti AK, Mitra B, Mukherjee G,
(MIT); 1960. Available from: Dutta PK. Automatic identification of clinically
https://dspace.mit.edu/handle/ 1721.1/52263 (1959). relevant regions from oral tissue histological images
8. Bera K, Schalper KA, Rimm DL, Velcheti V, for oral squamous cell carcinoma diagnosis. Tissue
Madabhushi A. Artificial intelligence in digital Cell. 2018;53:111-9.
pathology - new tools for diagnosis and precision 21. Sultan AS, Elgharib MA, Tavares T, Jessri M, Basile
oncology. Nat Rev Clin Oncol. 2019;16(11):703-15. JR. The use of artificial intelligence, machine learning
9. Patil S, Albogami S, Hosmani J, Mujoo S, Kamil MA, and deep learning in oncologic histopathology. J Oral
Mansour MA, et al. Artificial Intelligence in the Pathol Med. 2020;49(9):849-56.
Diagnosis of Oral Diseases: Applications and Pitfalls. 22. Lu C, Lewis JS, Dupont WD, Plummer WD,
Diagnostics (Basel). 2022;12(5):1029-32. Janowczyk A, Madabhushi A. An oral cavity
10. Kourou K, Exarchos TP, Exarchos KP, Karamouzis squamous cell carcinoma quantitative
MV, Fotiadis DI. Machine learning applications in histomorphometric-based image classifier of nuclear
cancer prognosis and prediction. Comput Struct morphology can risk stratify patients for disease-
Biotechnol J. 2014;13:8-17. specific survival. Mod Pathol. 2017;30(12):1655-65.
11. Pang-Ning T, Steinbach M, Kumar V. Introduction to 23. Shamim M, Syed S, Shiblee M, Usman M, Ali S.
data mining; 2006. 2nd Edition, Web book: Automated detection of oral precancerous tongue
https://www- lesions using deep learning for early diagnosis of oral
users.cse.umn.edu/~kumar001/dmbook/index.php cavity cancer. Comput J. 2022;65(1):91-104.
12. Jiang Y, Yang M, Wang S, Li X, Sun Y. Emerging role 24. Karadaghy OA, Shew M, New J, Bur AM.
of deep learning-based artificial intelligence in tumor Development and Assessment of a Machine Learning
pathology. Cancer Commun (Lond). 2020;40(4):154- Model to Help Predict Survival Among Patients With
66. Oral Squamous Cell Carcinoma. JAMA Otolaryngol
13. Choi RY, Coyner AS, Kalpathy-Cramer J, Chiang MF, Head Neck Surg. 2019;145(12):1115-20.
Campbell JP. Introduction to Machine Learning, doi:10.1001/jamaoto.2019.0981
Neural Networks, and Deep Learning. Transl Vis Sci 25. Bur AM, Holcomb A, Goodwin S, Woodroof J,
Technol. 2020;9(2):14. doi:10.1167/tvst.9.2.14 Karadaghy O, Shnayder Y, et al. Machine learning to
14. Rakha EA, Toss M, Shiino S, Gamble P, Jaroensri R, predict occult nodal metastasis in early oral squamous
Mermel CH, et al. Current and future applications of cell carcinoma. Oral Oncol. 2019;92(2):20-5.

Annals of Dental Specialty Vol. 11; Issue 3. Jul – Sep 2023 | 85

Kariamal and Angadi

26. Alabi RO, Elmusrati M, Sawazaki-Calone I, Kowalski 29. Benitez JM, Castro JL, Requena I. Are artificial neural
LP, Haglund C, Coletta RD, et al. Machine learning networks black boxes? IEEE Trans Neural Netw.
application for prediction of locoregional recurrences 1997;8(5):1156-64.
in early oral tongue cancer: a Web-based prognostic 30. Hartman DJ, Pantanowitz L, McHugh JS, Piccoli AL,
tool. Virchows Arch. 2019;475(4):489-97. OLeary MJ, Lauro GR. Enterprise implementation of
27. Arora A, Husain N, Bansal A, Neyaz A, Jaiswal R, Jain digital pathology: feasibility, challenges, and
K, et al. Development of a New Outcome Prediction opportunities. J Digit Imaging. 2017;30:555-60.
Model in Early-stage Squamous Cell Carcinoma of the 31. Madabhushi A, Lee G. Image analysis and machine
Oral Cavity Based on Histopathologic Parameters learning in digital pathology: challenges and
With Multivariate Analysis: The Aditi-Nuzhat Lymph- opportunities. Med Image Anal. 2016;33:170-5.
node Prediction Score (ANLPS) System. Am J Surg 32. Tizhoosh HR, Pantanowitz L, Intelligence A. Artificial
Pathol. 2017;41(7):950-60. intelligence and digital pathology: challenges and
28. Chang SW, Abdul-Kareem S, Merican AF, Zain RB. opportunities. J Pathol Inform. 2018;9(1):38-9.
Oral cancer prognosis based on clinicopathologic and
genomic markers using a hybrid of feature selection
and machine learning methods. BMC Bioinform.
2013;14(1):170-4.

Annals of Dental Specialty Vol. 11; Issue 3. Jul – Sep 2023 | 86