Artificial Intelligence and Otolaryngology-Head and Neck Surgery

Aru Panwar, MD, member, and Michael J. Brenner, MD, Chair, Outcomes Research and Evidence-based Medicine Committee

Artificial intelligence (AI), encompassing large language models, machine learning, natural language processing, and more, has captured the attention and imagination of not only otolaryngologists but also the lay public. The widespread accessibility of platforms such as Open AI's ChatGPT and Dall-E, Google’s Bard, and others has spurred innovative use of this technology in diverse fields of application.

AI use may interface with care delivery in otolaryngology-head and neck surgery in a variety of ways, offering new opportunities and presenting potential pitfalls. Otolaryngologists need to have an evolving understanding of AI, and broad stakeholder engagement is needed to develop agreed recommendations for ethical and effective integration of AI into otolaryngology education, research, and practice.

AI models can learn to recognize patterns from large datasets, allowing for problem solving. Specific AI models may help to classify data by identifying patterns within the source data – such as images, molecular structures, or software code – or make predictions through natural language processing. Many AI-based applications surround us already, such as voice assistants on our phones, bots providing customer service, or recommendations from online shopping portals. However, the use of AI in healthcare is still in its infancy. The promise of AI includes enhanced diagnostic accuracy, personalized treatment planning, predictive analytics for patient outcomes, and streamlined administrative processes. Unfortunately, AI is susceptible to bias and other pitfalls. In this article, we offer an overview of AI in otolaryngology-head and neck surgery.

How AI Is Transforming Clinical Practice

AI is already a part of clinical workflow for many otolaryngologists. Many clinicians have already embraced voice recognition software for transcribing dictated notes as well as AI-based scheduling tools. In a survey study published in 2023, 82% of the respondents intended to use AI as a decision aid and 74% felt comfortable with AI-recommended treatment proposals.¹ Clinicians were, however, more cautious in relying on AI technology to recognize malignancies or interpret radiological imaging. However, this reluctance may diminish as technology advances and AI is validated across diverse clinical applications. Publications on AI have accelerated in the last decade, with more than 280 PubMed indexed articles relating to AI in otolaryngology published in 2023 alone.²

AI can influence otolaryngology care by utilizing imaging, histology, and other clinical variables to predict clinical outcomes. For example, machine learning algorithms can predict risk for nodal metastases or overall prognosis for patients with oral cavity squamous cell carcinoma, papillary thyroid carcinoma, and laryngeal cancer.^3-6 AI also has an emerging role in oropharyngeal squamous cell carcinoma detection, margin assessment, and noninvasive diagnosis of mucosal lesions including malignancy.^7-9 Machine learning algorithms have been trained on images from endoscopic narrow band imaging, using blue and green light to examine mucosal surface architecture and vascular patterns, to automate and improve identification of malignant lesions of the oropharynx.⁷ In some studies, radiomics-based AI analyses may outperform human experts in identifying clinically important information such as extra-nodal extension on diagnostic imaging.¹⁰ Evolving uses of AI may also include improved analysis of hearing, speech, or sleep-disordered breathing. These and other applications can provide clinical decision support, streamline care, and transform how we practice otolaryngology-head and neck surgery.

AI applications may ease administrative tasks through automation of billing and task reminders and enhance communication between physicians and other stakeholders including patients, insurers, healthcare systems and others. However, clinicians are advised to remain cautious about risks relating to breaches of confidentiality that could result from unauthorized input of patient data into generative AI platforms. Clinicians should follow institutional guidance and comply with processes that ensure the validation and safe use of AI in clinical practice.

How AI Is Improving Clinician, Patient, and Public Education

The ability of AI to synthesize large volumes of data will likely prove influential in educating clinicians, patients, and the public. Currently available generative AI models are imperfect but can already achieve a passing grade on at least five of the 10 categories on the self-assessment examination for sleep medicine certification, and can correctly answer more than half of the questions on the otolaryngology residency in-service examination.^11,12 Assimilation of AI in simulation-based surgical training is already underway, facilitating objective skills assessment and structured feedback.¹³ Although the results of using AI in otolaryngology professional education have been mixed thus far, continued technological refinements will likely allow the meaningful incorporation of AI in training programs and continuing education.

The appeal of generative AI to the patient and lay public for healthcare is readily apparent, but few reliable data are available on its implications for care. Freely available generative AI platforms might afford patients a better understanding of their health conditions, treatment options, and prognosis. This may empower patients through greater engagement, more nuanced understanding of complex medical issues, or a more streamlined experience when accessing otolaryngology-head and neck surgery-related care. However, having access to such information may exacerbate challenges, similar to when patients develop a flawed understanding after internet-based searches. Patients with these erroneous impressions may be more resistant to advice offered by their clinicians as the patients may place greater weight on an AI algorithm that advertises data-driven objectivity.

Recent research suggests that preoperative counseling from ChatGPT may be similar in quality to traditional internet search-based research. Clinicians already report using generative AI for counseling for otorhinolaryngology procedures such as a rhinoplasty. In a study comparing education offered by an expert clinician to that offered by a generative AI platform pertaining to septorhinoplasty, observers preferred the education provided by the AI platform.^14-16 In many instances, AI-generated responses may be above the traditional sixth grade reading level that is recommended for most medical literature aimed at patient education.¹⁷ Additionally, generative AI platforms may be sensitive to suggestion and offer inaccurate information (also known as ”hallucination”) that may create potential for harm.¹⁷ As AI use increases, both individual otolaryngologists and professional otolaryngology organizations will need to take a proactive role in curating patient education and resources from AI platforms

Research Implications of AI

Large language models, machine learning, and natural language processing have found broad diverse application in research. AI has been leveraged to create, manage, and curate head and neck cancer survivor registries, to draft and proofread manuscripts, and to assist in peer review and dissemination of innovations.^18-20 Advancements in AI may accelerate innovation, bring down barriers in research, and allow recognition of actionable insights from data that may otherwise be difficult to interpret or manage through conventional analytical methodology.

However, AI algorithms can also be remarkably opaque and are dependent on the quality, validity, and integrity of data used to train the algorithm. Traditional training in research methods may leave clinicians and researchers ill-equipped to assess the methodologic integrity and validity of research conducted using AI models. This limitation constrains the epistemological value of peer reviewers, editors, and thought leaders who typically shape the innovation landscape. One could argue that this evolution is a step toward greater democratization in research priorities; however, human input is critical to ensuring that research has societal and clinical value, accurately represents the population being served, and safeguards the ethical principles of respect, beneficence, justice, and confidentiality.

Caution and Regulation Ahead

Since AI algorithms use pre-existing data for training and validating their future outputs, the interpretation is susceptible to several biases. AI-assisted research models that rely on unrepresentative datasets may perpetuate pre-existing stereotypes. AI may also discount novel ideas or observations that diverge from previous experience, and misinformation from hallucinations can impair patient care. These concerns have prompted the development of guideline statements that afford guardrails for use of AI, authorship, and proposals for minimum reporting standards for machine learning based science.^{22, 23} This guidance must evolve as technology evolves to ensure continued relevance and fidelity to the core ethical principles. 

Applications based on AI are already part of everyday life in clinical and nonclinical settings and will continue to find utility in ways both mundane and surprising. It will serve our patients and the field of otolaryngology-head and neck surgery to engage in real-world use cases for AI with equal measure of enthusiasm and skepticism. Doing so will ensure that technology promotes equitable care and alleviation of human suffering through enhancements to clinical care, research, and education.  

Disclosures
Aru Panwar, MD, serves as an associate editor for the head and neck/skull base surgery section for the journal Otolaryngology–Head and Neck Surgery.

Michael J. Brenner, MD, serves as an associate editor for the Health Policy, Patient Safety, and Quality Improvement section for the journal Otolaryngology–Head and Neck Surgery.

References

1. Asokan A, Massey CJ, Tietbohl C, Kroenke K, Morris M, Ramakrishnan VR. Physician views of artificial intelligence in otolaryngology and rhinology: A mixed methods study. Laryngoscope Investig Otolaryngol. 2023 Oct 31;8(6):1468-1475.
2. Pubmed Search. Accessed January 03, 2024 and available at: https://pubmed.ncbi.nlm.nih.gov/?term=artificial%20intelligence%20otolaryngology&sort=date&page=1
3. Esce AR, Baca AL, Redemann JP, Rebbe RW, Schultz F, Agarwal S, Hanson JA, Olson GT, Martin DR, Boyd NH. Predicting nodal metastases in squamous cell carcinoma of the oral tongue using artificial intelligence. Am J Otolaryngol. 2024 Jan-Feb;45(1):104102.
4. Nakajo M, Nagano H, Jinguji M, Kamimura Y, Masuda K, Takumi K, Tani A, Hirahara D, Kariya K, Yamashita M, Yoshiura T. The usefulness of machine-learning-based evaluation of clinical and pretreatment ¹⁸F-FDG-PET/CT radiomic features for predicting prognosis in patients with laryngeal cancer. Br J Radiol. 2023 Sep;96(1149):20220772.
5. Petruzzi G, Coden E, Iocca O, di Maio P, Pichi B, Campo F, De Virgilio A, Francesco M, Vidiri A, Pellini R. Machine learning in laryngeal cancer: A pilot study to predict oncological outcomes and the role of adverse features. Head Neck. 2023 Aug;45(8):2068-2078.
6. Esce A, Redemann JP, Olson GT, Hanson JA, Agarwal S, Yenwongfai L, Ferreira J, Boyd NH, Bocklage T, Martin DR. Lymph Node Metastases in Papillary Thyroid Carcinoma can be Predicted by a Convolutional Neural Network: a Multi-Institution Study. Ann Otol Rhinol Laryngol. 2023 Nov;132(11):1373-1379.
7. Mascharak S, Baird BJ, Holsinger FC. Detecting oropharyngeal carcinoma using multispectral, narrow-band imaging and machine learning. Laryngoscope. 2018 Nov;128(11):2514-2520.
8. Loperfido A, Celebrini A, Marzetti A, Bellocchi G. Current role of artificial intelligence in head and neck cancer surgery: a systematic review of literature. Explor Target Antitumor Ther. 2023;4(5):933-940.
9. You Z, Han B, Shi Z, Zhao M, Du S, Yan J, Liu H, Hei X, Ren X, Yan Y. Vocal cord leukoplakia classification using deep learning models in white light and narrow band imaging endoscopy images. Head Neck. 2023 Dec;45(12):3129-3145.
10. Kann BH, Likitlersuang J, Bontempi D, Ye Z, Aneja S, Bakst R, Kelly HR, Juliano AF, Payabvash S, Guenette JP, Uppaluri R, Margalit DN, Schoenfeld JD, Tishler RB, Haddad R, Aerts HJWL, Garcia JJ, Flamand Y, Subramaniam RM, Burtness BA, Ferris RL. Screening for extranodal extension in HPV-associated oropharyngeal carcinoma: evaluation of a CT-based deep learning algorithm in patient data from a multicentre, randomised de-escalation trial. Lancet Digit Health. 2023 Jun;5(6):e360-e369.
11. Cheong RCT, Pang KP, Unadkat S, Mcneillis V, Williamson A, Joseph J, Randhawa P, Andrews P, Paleri V. Performance of artificial intelligence chatbots in sleep medicine certification board exams: ChatGPT versus Google Bard. Eur Arch Otorhinolaryngol. 2023 Dec 20. doi: 10.1007/s00405-023-08381-3. Epub ahead of print.
12. Mahajan AP, Shabet CL, Smith J, Rudy SF, Kupfer RA, Bohm LA. Assessment of Artificial Intelligence Performance on the Otolaryngology Residency In-Service Exam. OTO Open. 2023 Nov 29;7(4):e98.
13. Varas J, Coronel BV, Villagrán I, Escalona G, Hernandez R, Schuit G, Durán V, Lagos-Villaseca A, Jarry C, Neyem A, Achurra P. Innovations in surgical training: exploring the role of artificial intelligence and large language models (LLM). Rev Col Bras Cir. 2023 Aug 25;50:e20233605. doi: 10.1590/0100-6991e-20233605-en.
14. Lee JC, Hamill CS, Shnayder Y, Buczek E, Kakarala K, Bur AM. Exploring the Role of Artificial Intelligence Chatbots in Preoperative Counseling for Head and Neck Cancer Surgery. Laryngoscope. 2023 Dec 21. doi: 10.1002/lary.31243.
15. Capelleras M, Soto-Galindo GA, Cruellas M, Apaydin F. ChatGPT and Rhinoplasty Recovery: An Exploration of AI's Role in Post-Operative Guidance. Facial Plast Surg. 2023 Nov 29. doi: 10.1055/a-2219-4901.
16. Durairaj KK, Baker O, Bertossi D, Dayan S, Karimi K, Kim R, Most S, Robotti E, Rosengaus F. Artificial Intelligence Versus Expert Plastic Surgeon: Comparative Study Shows ChatGPT "Wins" Rhinoplasty Consultations: Should We Be Worried? Facial Plast Surg Aesthet Med. 2023 Nov 20.
17. Campbell DJ, Estephan LE, Sina EM, Mastrolonardo EV, Alapati R, Amin DR, Cottrill EE. Evaluating ChatGPT Responses on Thyroid Nodules for Patient Education. Thyroid. 2023 Dec 26.
18. Gautamjit RK, Gupta R, Singh A, Panda NK, Ghoshal S, Bakshi JB, Verma RK. Feasibility of Establishing an Artificial Intelligence Based Head and Neck Cancer Registry: Experience from a Tertiary Care Hospital. Indian J Otolaryngol Head Neck Surg. 2022 Dec;74(4):588-593.
19. Khoury CJ, Enver N, Paderno A, Ratti E, Rameau A. Using Generative Artificial Intelligence in the Production and Dissemination of Innovation in Otolaryngology-Ethical Considerations. Otolaryngol Head Neck Surg. 2023 Dec 3.
20. Lechien JR, Gorton A, Robertson J, Vaira LA. Is ChatGPT-4 Accurate in Proofread a Manuscript in Otolaryngology-Head and Neck Surgery? Otolaryngol Head Neck Surg. 2023 Sep 17.
21. Nwosu OI, Crowson MG, Rameau A. Artificial Intelligence Governance and Otolaryngology-Head and Neck Surgery. Laryngoscope. 2023 Nov;133(11):2868-2870.
22. REFORMS: Reporting Standards for ML-Based Science. Accessed January 03, 2024. Available at: https://reforms.cs.princeton.edu/
23. Zielinski C, Winker MA, Aggarwal R, Ferris LE, Heinemann M, Lapeña JF, Pai SA, Ing E, Citrome L, Alam M, Voight M, Habibzadeh F; WAME Board. Chatbots, generative AI, and scholarly manuscripts: WAME recommendations on chatbots and generative artificial intelligence in relation to scholarly publications. Curr Med Res Opin. 2024 Jan;40(1):11-13.