Artificial Intelligence in Head and Neck Surgical Oncology: A State-of-the-Art Review.

Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice J Clin Med . 2026 Apr 6;15(7):2767. doi: 10.3390/jcm15072767 Search in PMC Search in PubMed View in NLM Catalog Add to search Artificial Intelligence in Head and Neck Surgical Oncology: A State-of-the-Art Review Steven X Chen Steven X Chen 1 Department of Otolaryngology—Head and Neck Surgery, Indiana University School of Medicine, Indiana University, Indianapolis, IN 46202, USA Find articles by Steven X Chen 1 , Maria Feucht Maria Feucht 1 Department of Otolaryngology—Head and Neck Surgery, Indiana University School of Medicine, Indiana University, Indianapolis, IN 46202, USA Find articles by Maria Feucht 1 , Aditya Bhatt Aditya Bhatt 1 Department of Otolaryngology—Head and Neck Surgery, Indiana University School of Medicine, Indiana University, Indianapolis, IN 46202, USA Find articles by Aditya Bhatt 1 , Janice L Farlow Janice L Farlow 1 Department of Otolaryngology—Head and Neck Surgery, Indiana University School of Medicine, Indiana University, Indianapolis, IN 46202, USA Find articles by Janice L Farlow 1, * Editors: Dustin A Silverman 1 , Alice L Tang 1 Author information Article notes Copyright and License information 1 Department of Otolaryngology—Head and Neck Surgery, Indiana University School of Medicine, Indiana University, Indianapolis, IN 46202, USA * Correspondence: [email protected] Roles Dustin A Silverman : Academic Editor Alice L Tang : Academic Editor Received 2026 Mar 1; Revised 2026 Mar 31; Accepted 2026 Apr 3; Collection date 2026 Apr. © 2026 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license . PMC Copyright notice PMCID: PMC13073410  PMID: 41977066 Abstract Artificial intelligence (AI) is rapidly reshaping head and neck surgical oncology by augmenting decision-making across the full perioperative continuum. This state-of-the-art review aims to provide head and neck surgical oncologists with a conceptual framework for understanding and critically appraising AI tools entering clinical practice, summarizing how machine learning, deep learning, and generative AI are being integrated into contemporary surgical workflows. Preoperative applications include detection of occult nodal metastasis and extranodal extension. Intraoperative innovations include augmented reality-assisted navigation, real-time margin assessment, and improving visual clarity and tissue handling for robotic platforms. Postoperatively, AI can predict complications like free flap failure and oncologic outcomes. Large language models are being operationalized for clinician-facing applications such as documentation and inbox support, as well as patient-facing education. Despite promising results, broad clinical deployment remains limited by concerns about privacy, validation, reliability, safety, and ethics. Widespread adoption will require prospective clinical trials, robust governance, and human-centered workflows that ensure AI remains a safe, assistive copilot. Keywords: artificial intelligence, machine learning, deep learning, large language model, computer vision, augmented reality, head and neck, surgery, otorhinolaryngology, oncology 1. Introduction In recent years, artificial intelligence (AI) has emerged as a promising tool to improve diagnostics, treatment planning, and outcomes in medicine. In head and neck surgical oncology, AI applications remain in relatively early stages but are rapidly expanding. AI techniques are being explored across the continuum of care from preoperative diagnostic imaging and surgical planning to intraoperative guidance and decision support, and postoperative outcome prediction and surveillance. While AI has previously been largely within the purview of computational scientists, a fluency with and critical appraisal of increasing applications of AI will soon be required of clinicians. This review highlights relevant key AI concepts, the current state of AI in head and neck cancer surgery, challenges and considerations, and future directions. The primary objective of this review is to equip head and neck surgical oncologists with a conceptual framework for understanding, evaluating, and responsibly integrating AI into clinical practice. It highlights both the current promise and the meaningful barriers that remain before widespread adoption is appropriate. 2. Methods This state-of-the-art narrative review was conducted to synthesize the current literature on AI applications across the perioperative continuum in head and neck surgical oncology. A structured literature search was performed in PubMed/MEDLINE and Google Scholar using the following Medical Subject Headings (MeSH) and keyword terms. AI Terms: “Artificial Intelligence” OR “Machine Learning” OR “Deep Learning” OR “Neural Networks, Computer” OR “Natural Language Processing” OR “Computer Vision” OR “Convolutional Neural Network” OR “Random Forest” OR “Support Vector Machine” OR “Large Language Model” OR “Robotics” OR “Radiomics” OR “Digital Pathology”. Head & Neck Surgery Terms: “Head and Neck Neoplasms” OR “Head and Neck Surgery” OR “Otolaryngology” OR “Laryngectomy” OR “Thyroidectomy” OR “Parotidectomy” OR “Neck Dissection” OR “Oral Cavity” OR “Oropharyngeal Neoplasms” OR “Laryngeal Neoplasms” OR “Salivary Gland Neoplasms” OR “Skull Base Surgery” OR “Free Flap”. Searches were limited to articles published in English between January 2015 and March 2026, with additional seminal references included irrespective of publication date when deemed foundational to a given topic. Articles were screened for relevance to the clinical application of AI in head and neck surgical oncology, including both primary research studies and systematic reviews. Studies were selected based on clinical relevance, methodological quality, novelty, and topical coverage across the perioperative workflow. Given the rapid evolution of this field, preprints and recently published peer-reviewed articles were included when they represented significant advances. A narrative review format was selected as most appropriate for this topic for several reasons. First, the field of AI in head and neck surgical oncology is heterogeneous and rapidly evolving, spanning diverse methodologies, anatomic subsites, and clinical contexts that resist the uniform eligibility criteria and quantitative synthesis required by systematic or scoping review frameworks. Second, the primary aim of this review is conceptual and educational: to provide surgeons with an accessible framework for understanding and critically appraising a broad technology landscape rather than to answer a narrow, pre-specified clinical question amenable to meta-analysis. Third, the existing literature is insufficiently standardized in outcome reporting to permit meaningful pooled estimates. A narrative approach therefore allows the authors to integrate evidence across heterogeneous study types, highlight methodological strengths and limitations, and draw clinically oriented conclusions that would be obscured by the constraints of more formal review methodologies. Because of the narrative review format, formal PRISMA methodology and meta-analysis were not employed. Instead, the authors synthesized findings thematically across preoperative, intraoperative, and postoperative domains, with attention to study design, validation approach, and clinical applicability. Key performance metrics (e.g., AUC, sensitivity, specificity) are summarized where reported. 3. Principles of Artificial Intelligence in Surgery AI is a broad category of computational methods and technology designed to simulate human learning, comprehension, problem solving, decision making, creativity and autonomy ( Figure 1 ). Within surgical oncology, it is best thought of as a set of methods that learn patterns from data to support decisions and workflows across the perioperative timeline, including preoperative (risk stratification, imaging interpretation, surgical planning), intraoperative (navigation, image guidance, margin assessment support), and postoperative (complication prediction, surveillance, patient-facing education) phases ( Figure 2 , Table 1 ) [ 1 ]. Input data is often multimodal, including cross-sectional imaging, endoscopy video, digital pathology, laboratory results, operative reports, tumor board notes, and longitudinal outcomes. Figure 1. Open in a new tab Hierarchical Framework of Artificial Intelligence. The hierarchy progresses from broad theoretical frameworks to specific architectural implementations. Artificial Intelligence (AI): The overarching domain encompassing computational systems designed to emulate human cognitive functions, such as reasoning, problem-solving, and decision-making in clinical environments. Machine Learning (ML): A critical subset of AI focusing on the development of algorithms that improve performance through experience. In medicine, ML is used to identify patterns in structured data (e.g., electronic health records) to facilitate predictive analytics without being explicitly programmed for every scenario. Deep Learning (DL): A specialized branch of ML utilizing multilayered neural networks to process complex, high-dimensional data. Convolutional Neural Networks (CNNs): Optimized for spatial data, primarily used for automated medical imaging analysis. Recurrent Neural Networks (RNNs): Designed for sequential data, such as real-time physiological monitoring or longitudinal patient history. Generative AI: A modern subset of deep learning capable of creating novel data outputs. Generative Adversarial Networks (GANs): Employed for creating synthetic medical data or enhancing image resolution. Large Language Models (LLMs): Trained on vast corpora of text to assist in natural language processing (NLP) tasks, such as automated clinical documentation or medical question-answering. Figure 2. Open in a new tab The Artificial Intelligence (AI) Continuum in Head and Neck Surgical Oncology. This schematic illustrates potential AI applications across the preoperative, intraoperative, and postoperative phases of care. Preoperative Phase: AI applications facilitate precision diagnostics through digital pathology for automated risk stratification and deep-learning-based surgical planning, optimizing patient selection and reconstructive mapping. Intraoperative Phase: Real-time assistance is provided via augmented reality (AR)-assisted navigation, which overlays genomic or radiologic data onto the surgical field, and AI-driven margin assessment (e.g., via hyperspectral imaging or optical coherence tomography) to ensure complete oncologic resection. Postoperative Phase: Predictive analytics utilize longitudinal data to model survival outcomes and complication risks, while computer-aided surveillance systems monitor for early recurrence, enabling personalized follow-up and timely intervention. Table 1. Summary of Select Studies Evaluating AI Applications in Head and Neck Surgical Oncology. Author (Year) Application Methods Key Metric Validation Limitations Preoperative Lan et al. (2024) [ 2 ] Occult cervical LN metastasis detection (MRI-based DL + radiomics) CNN on MRI for oral/oropharyngeal SCC; training + external cohort AUC 0.92 (training), AUC 0.72 (external validation) External Large performance drop across cohorts; retrospective design; lacks prospective validation Dayan et al. (2026) [ 3 ] Imaging-based ENE detection + outcome prediction (HPV+ OPC) DL model on pretreatment CT; HPV+ OPC Outperformed radiologist ENE detection; improved outcome stratification Internal cross validation Single cancer subtype (HPV+ OPC); CT-only input; institutional protocol variation Kann et al. (2023) [ 4 ] CT-based DL screening for ENE in HPV+ OPC (de-escalation trial context) DL algorithm evaluated on multicenter RCT patient data Validated sensitivity/specificity for ENE in multicenter trial setting Multicenter RCT data Prospective integration into clinical workflow not yet assessed; HPV+ only Klein et al. (2023) [ 5 ] DL prediction of HPV status in OPC from H&E histology slides CNN on whole-slide H&E images; multi-institutional cohort High AUC for HPV status classification directly from H&E Multi-institutional; internal cross validation OPC-specific; requires high-quality slide digitization; not yet adopted clinically Unlu et al. (2025) [ 6 ] LLM-assisted clinical trial prescreening vs. manual chart review RCT comparing AI vs. manual eligibility screening Comparable accuracy to manual review; significantly reduced time burden RCT General oncology (not head and neck-specific); may miss nuanced eligibility criteria Intraoperative Yang et al. (2026) [ 7 ] AR-guided tumor resection with deformable registration (tongue) Pre/post-resection surface registration framework; tongue specimens TRE improved by 33%; average error reduced from 9.8 mm to 4.8 mm Internal cross validation Investigational; not validated in live surgery; limited anatomic sites tested Halicek et al. (2019) [ 8 ] HSI + DL for intraoperative margin assessment HSI + CNN on 102 surgical specimens (HNSCC) 98% accuracy, 93% sensitivity vs. histopathology gold standard Internal cross validation Ex vivo specimens only; no real-time intraoperative trial; limited to HNSCC Postoperative Karadaghy et al. (2019) [ 9 ] ML model for overall survival prediction in oral SCC Random forest + clinical/pathologic features; SEER + institutional data AUC 0.77 for 5-year survival prediction Internal cross validation Retrospective; limited granularity on treatment details; no functional outcomes Formeister et al. (2020) [ 10 ] ML for predicting complications in HN microvascular free tissue transfer Logistic regression + random forest on institutional free flap cohort Improved complication prediction vs. traditional scoring tools Internal cross validation Single institution; small sample; models not externally validated You et al. (2025) [ 11 ] Ambient documentation AI: impact on documentation burden and burnout Large multi-site survey of clinicians using ambient AI-drafted notes Reduced perceived documentation burden; improved well-being scores Multi-site survey Self-report only; no objective outcome measures; no head and neck-specific subgroup Open in a new tab Abbreviations: AI, artificial intelligence; AUC, area under the curve; CNN, convolutional neural network; CT, computed tomography; DL, deep learning; ENE, extranodal extension; H&E, hematoxylin and eosin; HNSCC, head and neck squamous cell carcinoma; HPV, human papillomavirus; HIS, hyperspectral imaging; LLM, large language model; LN, lymph node; ML, machine learning; MRI, magnetic resonance imaging; OPC, oropharyngeal cancer; RCT, randomized controlled trial; SCC, squamous cell carcinoma; SEER, Surveillance Epidemiology and End Results Program; TRE, target registration error. Machine Learning (ML) refers broadly to algorithms that learn patterns from data without explicitly programmed rules [ 12 ]. Traditional ML often requires manual “feature engineering,” where domain experts select which variables the model will use. Deep Learning (DL), a subset of ML, eliminates the need for manual feature extraction, and is thus the engine behind the most advanced applications, particularly in image analysis and computer vision [ 13 ]. Convolutional Neural Networks (CNNs) are the workhorses of medical image analysis, designed to process grid-like data such as computed tomography slices [ 14 , 15 ]. The newest frontier is Generative AI, which include Large Language Models (LLMs) and Generative Adversarial Networks (GANs). Unlike discriminative models which classify data, generative models create new data. LLMs are being explored for automated note generation and summarizing patient histories [ 16 , 17 ]. Additionally, GANs can impute synthetic medical images, potentially filling in missing data in corrupted or low-resolution images to enhance diagnostic robustness. A variety of terms are used commonly for AI applications that may be less familiar to surgical oncologists ( Figure 3 ). A prompt is the instruction and constraints an LLM receives. Temperature controls how deterministic or variable the output is (lower is more precise, higher is more creative). Natural language processing (NLP) broadly refers to extracting/structuring information from text (e.g., margin status, nodal levels), while LLMs can also generate text (drafting summaries, letters, discharge instructions). The major safety principle here is acknowledging and engineering around hallucinations, or statements not supported by source data. Healthcare-focused evaluations show hallucinations and omissions can occur in medical summarization and extraction, motivating structured evaluation frameworks and clinician review [ 18 ]. Figure 3. Open in a new tab Glossary of Artificial Intelligence Terminology. The glossary is organized into four functional domains. (1) Model Development: Describes technical methodologies for processing surgical data. (2) Text Building Blocks: Defines the fundamental units of model interaction. (3) Model Behavior and Limits: Outlines critical safety and performance parameters. (4) Governance and Safety: Details the ethical and regulatory requirements for AI integration. To reduce hallucination and improve traceability, many teams use embeddings (numeric representations of text meaning) plus retrieval-augmented generation (RAG) to retrieve the most relevant source snippets (e.g., pathology result, radiology impression, operative report) and feed them into the model so the output is grounded in known documents. This enables “answer with citations to the chart,” which is attractive for tumor board summaries and staging extraction. Systematic reviews show RAG often improves factuality but requires standardized evaluation and careful choice of retrieval methods and reference sources [ 19 , 20 ]. Relatedly, fine-tuning (additional training on smaller, task-specific datasets) can match output to institutional-specific terminology, templates, and documentation patterns [ 21 ]. Safe deployment in surgical oncology hinges on governance. Protected health information (PHI) should be safeguarded with de-identification or controlled environments. Guardrails on what the system may and may not do should be defined. A human-in-the-loop where a human reviewer can verify and approve the AI output can also be deployed for clinical sign-off. Audit trails of sources, outputs, and edits should be carefully maintained. Finally, before scaling, validation of models against gold standards should be tested. These ideas map cleanly onto major guidance and frameworks: the United States Health and Human Services explains HIPAA de-identification pathways (Safe Harbor and Expert Determination), National Institute of Standards and Technology’s AI Risk Management Framework organizes trustworthy AI around governance and measurable risk controls, the World Health Organization emphasizes ethics/human rights and accountability in AI for health, and the Food and Drug Administration/International Medical Device Regulators Forum Good Machine Learning Practice principles stress lifecycle thinking for AI/ML medical products, including safety and effectiveness considerations [ 22 , 23 , 24 , 25 ]. 4. AI in the Perioperative Workflow 4.1. The Patient-Clinician Relationship Beneficial AI systems integrated into head and neck surgical oncology should be designed as assistive copilots—incorporated into existing clinical systems, accessible and understandable by users, constrained by appropriate guardrails, with reliable and reproducible processes that are available for human review—so that speed and consistency improve without sacrificing accuracy, safety, or trust. In practice, that may mean clinician-facing language tools that structure a patient’s course and management (automated notes, concise history/timeline summaries, suggested orders and portal message replies) and patient-facing tools that convert the same plan into clear, accessible education (plain-language instructions, chat-based navigation), while preserving clinician accountability. On the clinician-facing side, one of the most immediately practical applications of generative AI is documentation support via ambient listening (or ambient clinical intelligence) tools that draft a note for the clinician to review and sign. The clinician remains responsible for correctness, completeness, and clinical reasoning. A large multi-site survey study of ambient documentation technology (AI-drafted notes from clinician–patient conversations) suggested that such technology may reduce burnout and increase perceived well-being [ 11 ]. Because ambient tools often rely on audio capture, clinicians must be able to navigate patient communication and consent workflows: education, trust, and clear opt-out options [ 26 ]. Another clinician-facing use case is summarizing patient histories and compressing long charts into decision-ready narratives (e.g., pre-op risk summaries, tumor board note, treatment timeline, discharge summaries). One study demonstrated that hospital course summaries for discharge documentation written by an EHR-embedded LLM were rated as higher quality than physician-written ones [ 27 ]. This matters in complex longitudinal diseases like head and neck cancer, which often involves multifaceted data related to multimodality therapy, frequent recurrences, and multidisciplinary care. A separate study investigating patient inbox support showed that an EHR-integrated, HIPAA-compliant LLM that drafted inbox replies led to improvement in burnout, although message time metrics did not meaningfully change [ 28 ]. Another patient-facing investigation that had survey participants rate satisfaction with AI versus clinician responses to questions found higher satisfaction scores for AI responses [ 29 ]. In practice, this could translate into faster, more consistent first-draft answers for common post-op questions (wound/drain care, diet/swallow precautions, tracheostomy/laryngectomy care, expected radiation effects), with the clinical team ensuring safety, personalization, and appropriate escalation. One study showed LLM could transform inpatient discharge summaries into a more patient-friendly format with markedly improved readability/understandability, but also found safety concerns of omissions and inaccuracies, reinforcing that real-world use should include clinical review [ 30 ]. AI applications for post-operative patient education are attractive but must employ critical safety checks. Given that head and neck cancer surgeries often have complex discharge needs, significant issues can arise should erroneous information be communicated (e.g., airway management). Real-world use should consider clinical review and safety checks before giving content directly to patients. Finally, patient-facing chatbots (education, navigation, symptom guidance, appointment preparation) can improve access and responsiveness, but practical deployment hinges on guardrails. Some evidence suggests chatbots can produce responses judged as high quality and empathetic [ 31 ]. At the same time, accuracy and sourcing are common failure points. Studies on chatbots for mental health show that there may be value for improving access and even symptom improvement, but also concerns for variable and event inappropriate responses as well [ 32 , 33 ]. A chatbot that can be perceived as a “medical authority” is risky without retrieval, citation, and escalation logic. Keeping a degree of human-in-the-loop supervision is a key safety design decision. 4.2. Preoperative Applications The accessibility and accuracy of diagnosis, staging, and treatment planning dictate the entire treatment trajectory. A powerful early research area of AI in medicine centered on radiography, the interpretation of which can be complex, time-consuming, qualitative and subjective. AI algorithms can automatically segment primary head and neck tumors on scans with high accuracy comparable to expert radiologists [ 34 ]. AI systems have also shown expert-level sensitivity in detecting subtle lesions, aiding earlier diagnosis of cancers or precancerous conditions that might be missed on routine exam [ 35 ]. Beyond detection, radiomics can standardize image interpretation by transforming medical images into quantitative imaging features. These features can be used to develop ML and DL algorithms to improve risk stratification and predict outcomes in head and neck oncology. For head and neck oncology, where endoscopic images are commonly employed, AI can also be applied to these procedures to detect and classify cancers [ 36 , 37 , 38 ]. Additional algorithms have shown promise for automating staging in head and neck cancers based on multimodal DL pipelines [ 39 ]. Beyond this, AI is also being used to streamline clinical trial prescreening by translating inclusion and exclusion criteria into computable checks over both structured (e.g., laboratory values) and unstructured data (e.g., prose of clinical notes), reducing the manual chart-review burden that can meaningfully drive trial costs and delays [ 6 , 40 ]. This screening paradigm is particularly relevant to head and neck cancers, where trial eligibility often hinges on nuanced, frequently unstructured details such as tumor subsite and stage, prior surgery/radiation/systemic therapy, and biomarker status (p16 in oropharyngeal carcinoma). In the realm of treatment, DL models can detect subtle perturbations in pixel intensity that correlate with metastatic infiltration, thus allowing more confident assessment and planning for nodal metastasis. A landmark study demonstrated an Area Under the Curve (AUC) of 0.9177 in a training cohort and 0.7205 in an external validation cohort for predicting occult metastasis [ 2 ]. This performance outperforms conventional criteria and suggests a future of AI-assisted decisions for elective neck, although the drop in performance across cohorts highlights the importance of external validation. It is worth noting that this performance gap between training data and external validation is characteristic of many current AI models in head and neck oncology and reflects the challenges of dataset heterogeneity, differences in imaging protocols, and institution-specific annotation practices. Until models demonstrate more consistent performance across independent datasets, their clinical utility for elective neck treatment decisions should be interpreted with caution. Extranodal Extension (ENE) is a prognostic indicator that upstages a patient and often leads to aggressive multimodality treatment. Radiologists often struggle to detect ENE, with a 73% sensitivity in one meta-analysis [ 41 , 42 ]. DL algorithms trained on contrast-enhanced computed tomography scans have demonstrated superior accuracy in this task. Research indicates that AI-based algorithms applied to pretreatment scans can predict ENE with performance outperforming radiologists, identifying abstract prognostic features invisible to the naked eye [ 3 , 4 ]. Identifying ENE preoperatively permits more precise prognostication and may alter treatment planning (e.g., the use of upfront chemoradiation rather than surgery followed by adjuvant radiation with chemotherapy). Critically, while both referenced ENE studies demonstrate improved performance over radiologist assessment, neither has yet been prospectively integrated into a clinical decision pathway in a manner that demonstrably altered treatment outcomes. Their value in guiding the choice between upfront chemoradiation and surgery-first approaches remains to be established through prospective trials. AI is increasingly enhancing histopathological assessment by improving diagnostic accuracy, reproducibility, and the extraction of prognostic features. In oropharyngeal squamous cell carcinoma, DL has been used to “virtually stain” pathology images between stain types such as hematoxylin and eosin to immunohistochemistry, allowing a tissue-preserving alternative to identifying new prognostic biomarkers [ 43 ]. AI-based image analysis has also been applied to predict human papillomavirus status in oropharyngeal squamous cell carcinoma directly from histology slides, offering a potential future rapid, cost-effective adjunct to molecular testing [ 5 ]. These approaches offer objective, reproducible pathology insights that may complement or even surpass conventional histologic evaluation. Beyond diagnostic and staging applications, augmented reality (AR) is also emerging as a valuable tool in preoperative planning. Prior to surgery, AR-enabled platforms can integrate multimodal imaging data such as CT, MRI, and endoscopic video, to construct three-dimensional anatomical reconstructions that surgeons can explore and manipulate in a virtual environment [ 44 ]. This allows for rehearsal of complex resections, identification of critical structures at risk, and simulation of reconstructive options before the patient arrives in the operating room. AR-based preoperative planning has been explored across surgical specialties, where studies suggest it can improve spatial understanding of complex anatomy and may reduce intraoperative decision time [ 45 , 46 , 47 ]. In head and neck oncology specifically, preoperative AR visualization holds promise for procedures involving the skull base, parapharyngeal space, and proximity to the carotid artery, where anatomical complexity and the margin for error are highest. Integration of AI-based tumor segmentation with AR visualization further enables automatic overlay of predicted tumor boundaries and nodal basins onto three-dimensional surgical plans [ 48 ]. While the clinical maturity of these tools remains investigational and prospective outcome data are limited, they represent a logical upstream application of the same AR infrastructure being developed for intraoperative guidance. 4.3. Intraoperative Applications During surgery, AI technologies are enhancing real-time decision support. AR implementation in which digital overlays are projected onto the surgeon’s visual field can guide tumor resection and preservation of key structures. For instance, researchers are developing head-mounted AR display systems that project AI-analyzed data (such as tumor boundaries or vital structures) onto the operative field [ 49 ]. A major barrier in soft tissue surgery is the “soft tissue shift,” where the tumor moves once the skin is incised. AI algorithms utilizing deformable registration are solving this by warping preoperative models to fit the intraoperative reality, although these are still investigational. A recent study introduced a framework using both pre-resection external surfaces and post-resection cavities to model specimen thickness, improving Target Registration Error by up to 33% in tongue specimens and improved average target relocation errors from 9.8 mm to 4.8 mm, bringing AR within the realm of clinical safety [ 7 ]. Another critical intraoperative need in head and neck cancer surgery is accurate margin assessment. AI is making significant strides here through advanced optical imaging and real-time tissue analysis. Traditional frozen section pathology during surgery is labor-intensive and time-consuming, with accuracy limitations. Hyperspectral Imaging (HSI) captures light across hundreds of spectral bands to differentiate tumor from healthy tissue based on metabolic signatures [ 8 ]. Researchers have combined hyperspectral imaging (HSI) with deep learning to differentiate tumor from normal tissue during surgery without any dye or contrast [ 50 ]. In this approach, freshly resected specimens were scanned with HSI across many wavelengths and processed by CNNs. AI could then predict tumor presence with 98% accuracy and 93% sensitivity, matching closely with gold-standard histopathology. This kind of label-free, rapid optical analysis can be integrated into the surgical workflow to guide surgeons on where residual tumor might remain at the margins, potentially outperforming the 63% accuracy of frozen sections. An important limitation of the HSI studies cited is that they were performed on ex vivo surgical specimens rather than in real-time intraoperative settings with live tissue. Translation to an intraoperative workflow requires resolution of significant technical challenges, including motion artifact, tissue desiccation, and time constraints of an ongoing operation. These findings are therefore best understood as proof-of-concept data rather than clinical-readiness evidence. Other modalities like fluorescence imaging, Raman spectroscopy, and optical coherence tomography can also be paired with AI algorithms to identify cancerous tissue in real time [ 51 , 52 , 53 , 54 , 55 , 56 ]. AR can then display these AI-identified cancer areas to the surgeon. As these tools mature, surgeons could have immediate feedback on margin status, enabling more resections with negative margins, as well as real-time, more precise re-excisions. AI-assisted features are being explored to enhance robotic surgery in head and neck oncology by introducing elements of autonomy, precision, and safety. Traditional robotic systems, such as the da Vinci platform, rely entirely on direct human control; however, newer AI-integrated “cooperative” robotic systems can assist the surgeon in more active and intelligent ways [ 57 ]. For instance, AI models can enhance the surgical field by removing “smoke” from electrocautery in real-time or perform “denoising” and color correction to improve visual clarity during deep-cavity procedures. DL models can identify surgical planes and distinguish between healthy tissue and tumors, with some models generating heatmaps of probable cancer locations during surgery [ 58 ]. Since current robots lack haptic feedback, AI is also being used to estimate grip force and tissue tension, preventing accidental tissue damage [ 59 ]. An underexplored but clinically important intraoperative application domain is AI-assisted reconstructive planning. Reconstruction after oncologic head and neck surgery is inherently complex, requiring surgeons to evaluate defect characteristics, patient comorbidities, vascular anatomy, prior treatment history, and the spectrum of available flap options—from local and regional pedicled flaps to free tissue transfer. Traditionally, this decision-making process relies heavily on surgeon experience and centers on a relatively small set of well-established reconstructive algorithms. AI-based decision support tools have the potential to systematize this process by analyzing defect type, geometry, and patient-specific factors to suggest feasible reconstructive options ranked by predicted outcomes such as flap viability, functional restoration, and donor site morbidity. Rather than replacing the reconstructive surgeon’s judgment, such tools could serve as a cognitive aid, particularly for complex or unusual defects where the optimal reconstructive strategy is less clearly defined. Furthermore, integrating AI-driven reconstructive planning with preoperative imaging and tumor segmentation could eventually allow surgeons to simulate both oncologic resection and reconstruction in a single planning workflow. Furthermore, such planning could be used as a patient counseling and trainee education tool. At present, clinical evidence for AI in reconstructive decision support remains limited, and prospective validation is needed before such tools can be responsibly integrated into routine care. 4.4. Postoperative Applications After initial treatment, AI plays a growing role in predicting patient outcomes and informing surveillance strategies in head and neck oncology. Numerous studies have applied ML algorithms to clinical, pathological, and imaging data to forecast outcomes such as survival, disease recurrence, and treatment complications [ 9 , 10 , 60 ]. For instance, predictive models have been built to estimate overall survival and disease-free survival based on patterns in patients’ demographics, tumor features, and therapy details [ 9 ]. Integrating radiomic features with clinical data improved prediction of 2-year survival in head and neck cancer compared to clinical data alone [ 61 ]. Beyond cancer control outcomes, AI can be used to foresee postoperative complications and resource needs. Models have been created to predict wound complications, gastrostomy dependence, free flap failure, or postoperative length of hospital stay based on pre-surgical factors, intraoperative data, and tumor stage [ 62 , 63 , 64 ]. Most published complication prediction models in head and neck surgery have been developed and validated within single institutions, limiting their generalizability. External performance is rarely reported, and the clinical threshold at which a model’s predicted risk would change management has not been defined. Prospective implementation studies that embed these tools into preoperative decision-making and measure their impact on patient outcomes (rather than simply their predictive accuracy) remain an important unmet need. Integration of these predictive capabilities into clinical workflows may allow for targeted interventions for vulnerable populations, including but not limited to patient education, or enhanced rehabilitation and nursing services. AI chatbots also hold promise to aid patients in their postoperative recovery, potentially augmenting patient education while decreasing workload on healthcare teams [ 60 ]. The postoperative phase in head and neck oncology also involves long-term surveillance for recurrence. Here, AI is augmenting traditional follow-up by analyzing subtle signals earlier than humans might detect. One frontier is liquid biopsy monitoring: AI can interpret trends in circulating tumor DNA or other blood biomarkers over time to signal minimal residual disease and impending recurrence, although prospective evidence linking AI-interpreted predictions to improved survival does not yet exist [ 65 ]. Patient-generated health data can also play a role in postoperative monitoring. For instance, AI-assisted analysis of voice recordings or swallowing metrics post-laryngectomy or oropharyngeal surgery may detect functional deterioration indicating a tumor recurrence or second primary [ 66 ]. By continuously learning from each patient’s post-treatment profile, AI systems may improve lead time detection of cancer recurrence for possibly earlier intervention. 5. Challenges For decades, the fundamental tenets of head and neck surgical oncology have remained consistent: complete oncologic resection, management of the cervical lymphatics, and functional reconstruction. While technological advances have refined these tasks, the cognitive burden of decision-making and the sensory limitations of the human operator have remained constant constraints. AI promises augmentation for the head and neck surgical oncologist; however, several barriers must be addressed before routine adoption ( Figure 4 ). Figure 4. Open in a new tab Key Knowledge Gaps and Future Directions for the Clinical Translation of Artificial Intelligence (AI) in Head and Neck Surgery. The framework identifies six primary domains. Red panels indicate current systemic bottlenecks, while green panels denote the requisite advancements for successful clinical implementation. 5.1. Data Quality, Representativeness, and Bias AI systems are only as reliable as the data on which they are trained [ 67 ]. Imaging, pathology, and clinical data require accurate annotations and formatting (e.g., imaging standards, standardized pathology labels) to allow effective training and comparison of algorithms. Bias remains a central concern, as training datasets may underrepresent certain racial, socioeconomic, or geographic populations, which could lead to AI algorithms propagating inequities in prognostic estimation or treatment recommendations [ 68 ]. Harmonized, representative datasets with intentional bias mitigation are critical [ 69 ]. Furthermore, federated learning—where models are trained across institutions without transferring raw data—offer one pathway to improve generalizability while preserving patient privacy [ 70 ]. However, such approaches require substantial technical infrastructure and coordination. An additional source of heterogeneity is the diversity of imaging protocols across institutions, including variation in CT slice thickness, contrast timing, MRI field strength, and endoscopy equipment. Pathology annotation is similarly institution-dependent, with variability in reporting standards for nodal levels, margin definitions, and tumor grade. These sources of data heterogeneity compound the challenge of building generalizable AI models and must be explicitly addressed in future multicenter dataset curation efforts. 5.2. External Validation and Clinical Utility Many AI models report strong discrimination in retrospective application to internal datasets, but performance declines in external validation cohorts. Prospective validation in real-world clinical settings, particularly in a randomized fashion, remains limited. Critically, AI tools have not yet demonstrated improvement in outcomes, such as survival, margin status, complications reduction, or quality of life. The transition from proof-of-concept to demonstrated clinical value will require prospective implementation studies, careful assessment of workflow impact, and acceptability to the physicians and/or the patients who use these AI tools. 5.3. Workflow Integration and Surgeon Trust Even technically robust and validated models may fail if poorly integrated into the clinical workflow. Tools at the point-of-care, with accessible user interface design, and interpretable and actionable outputs are most likely to gain traction with physicians. Clinicians often express mistrust or lack of understanding of AI “black boxes.” If an algorithm cannot explain its reasoning, it is hard for surgical oncologists to fully trust its recommendations over their own expertise. Currently, many advocate for a “human-in-the-loop” model combined with transparent, interpretable outputs (e.g., heatmaps highlighting image regions of concerns, confidence intervals, and traceable source) [ 71 ]. 5.4. Regulatory and Medicolegal Considerations AI systems deployed in clinical care are increasingly regulated as medical devices, yet jurisprudence regarding liability in AI-assisted patient care is still evolving [ 72 ]. No consensus framework has emerged for liability in the setting of AI-related error. Clinicians, developers, and healthcare institutions may each bear responsibility [ 73 ]. Liability may depend in part on the degree of algorithmic autonomy, with clinicians likely assuming greater liability when AI is used in a decision-support role, and liability shifting toward developers or institutions with increasing algorithmic independence [ 74 ]. As AI tools become more widely validated, the standard of care may evolve to incorporate their use. Adoption can be risky. Failure to adopt improved technologies and inappropriate reliance on flawed outputs may both expose surgeons to liability. The “black box” nature of many AI systems adds further complexity to liability attribution, as limited explainability may hinder a surgeon’s ability to justify AI-influenced decisions, particularly in high-stakes settings such as tumor margin assessment or treatment de-escalation [ 75 ]. This raises important considerations for informed consent, as physicians may be unable to fully explain the basis of AI-generated recommendations, potentially limiting patient autonomy [ 76 ]. While the use of AI in clinical decision-making should likely be disclosed to patients, the appropriate extent of disclosure remains uncertain given concerns that overemphasis on AI involvement may undermine patient trust when these tools function primarily as adjuncts to clinician judgment [ 76 ]. Additional concerns arise in the use of patient data for AI development, where evolving regulatory frameworks, challenges in obtaining meaningful consent, and risks of data breaches necessitate robust data governance, transparency, and privacy protections [ 77 ]. Clear documentation of decision-making rationale, incorporation of human oversight, and implementation of audit trails and post-deployment surveillance will be essential to mitigate medicolegal risk as AI becomes increasingly integrated into head and neck surgical oncology. 5.5. Economic and Operational Realities Development, validation, and maintenance of AI systems require significant investment in infrastructure, data engineering, cybersecurity, and personnel expertise [ 78 , 79 , 80 ]. Cost-effectiveness analyses remain limited, particularly in surgical oncology. 6. Conclusions This review provides a comprehensive synthesis of emerging applications of artificial intelligence across head and neck surgical oncology, organized along the continuum of care from preoperative planning to intraoperative decision-making and postoperative outcomes and surveillance. Importantly, this work aims to bridge the gap between technical development and clinical practice by equipping surgeons with a framework to understand, evaluate, and responsibly integrate AI tools into patient care. Across these domains, we highlight both the breadth of current applications and the variability in performance, validation, and clinical readiness, while identifying key barriers to clinical translation. As clinical practice becomes increasingly integrated with AI systems, rigorous validation and oversight will be required to establish where AI truly adds value and where human judgment remains irreplaceable. AI will not replace surgeons, but surgeons who understand and appropriately deploy AI tools will likely outperform those who do not. As with prior technological shifts in head and neck oncology—from endoscopy to microvascular reconstruction to transoral robotics—the surgeons who critically engage with these tools will shape how they are safely integrated into practice. Acknowledgments We thank Hannah J. Craven, from the Ruth Lilly Medical Library for assistance on evidence synthesis design and strategy. Abbreviations The following abbreviations are used in this manuscript: AI Artificial intelligence ANN Artificial neural networks AR Augmented reality AUC Area under the curve CNN Convolutional neural networks DL Deep learning EHR Electronic health record ENE Extranodal extension GAN Generative adversarial networks HIPAA Health Insurance Portability and Accountability Act HSI Hyperspectral imaging LLM Large language model ML Machine learning NLP Natural language processing PHI Protected health information RAG Retrieval-augmented generation Open in a new tab Author Contributions Conceptualization, S.X.C. and J.L.F.; methodology, S.X.C. and J.L.F.; writing—original draft preparation, S.X.C.; writing—review and editing, S.X.C., M.F., J.L.F.; visualization, S.X.C., A.B., J.L.F.; supervision, J.L.F.; All authors have read and agreed to the published version of the manuscript. Institutional Review Board Statement Not applicable. Informed Consent Statement Not applicable. Data Availability Statement No new data were created or analyzed in this study. Data sharing is not applicable to this article. Conflicts of Interest The authors declare no conflicts of interest. Funding Statement This research received no external funding. Footnotes Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. References 1. Varghese C., Harrison E.M., O’Grady G., Topol E.J. Artificial Intelligence in Surgery. Nat. Med. 2024;30:1257–1268. doi: 10.1038/s41591-024-02970-3. [ DOI ] [ PubMed ] [ Google Scholar ] 2. Lan T., Kuang S., Liang P., Ning C., Li Q., Wang L., Wang Y., Lin Z., Hu H., Yang L., et al. MRI-Based Deep Learning and Radiomics for Prediction of Occult Cervical Lymph Node Metastasis and Prognosis in Early-Stage Oral and Oropharyngeal Squamous Cell Carcinoma: A Diagnostic Study. Int. J. Surg. Lond. Engl. 2024;110:4648–4659. doi: 10.1097/JS9.0000000000001578. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 3. Dayan G.S., Hénique G., Bahig H., Nelson K., Brodeur C., Christopoulos A., Filion E., Nguyen-Tan P.-F., O’Sullivan B., Ayad T., et al. Artificial Intelligence Model for Imaging-Based Extranodal Extension Detection and Outcome Prediction in Human Papillomavirus−Positive Oropharyngeal Cancer. JAMA Otolaryngol. Neck Surg. 2026;152:7–17. doi: 10.1001/jamaoto.2025.3225. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Kann B.H., Likitlersuang J., Bontempi D., Ye Z., Aneja S., Bakst R., Kelly H.R., Juliano A.F., Payabvash S., Guenette J.P., et al. 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;5:e360–e369. doi: 10.1016/S2589-7500(23)00046-8. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Klein S., Wuerdemann N., Demers I., Kopp C., Quantius J., Charpentier A., Tolkach Y., Brinker K., Sharma S.J., George J., et al. Predicting HPV Association Using Deep Learning and Regular H&E Stains Allows Granular Stratification of Oropharyngeal Cancer Patients. npj Digit. Med. 2023;6:152. doi: 10.1038/s41746-023-00901-z. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Unlu O., Varugheese M., Shin J., Subramaniam S.M., Stein D.W.J., St Laurent J.J., Mailly C.J., McPartlin M.J., Wang F., Oates M.F., et al. Manual vs AI-Assisted Prescreening for Trial Eligibility Using Large Language Models—A Randomized Clinical Trial. JAMA. 2025;333:1084–1087. doi: 10.1001/jama.2024.28047. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. Yang Q., Li F., Xu J., Liu Z., Sridhar S., Jin W., Du J., Heiselman J., Miga M., Topf M., et al. Augmented Reality-Based Guidance with Deformable Registration in Head and Neck Tumor Resection. In: Gee J.C., Alexander D.C., Hong J., Iglesias J.E., Sudre C.H., Venkataraman A., Golland P., Kim J.H., Park J., editors. Proceedings of the Medical Image Computing and Computer Assisted Intervention—MICCAI 2025. Springer Nature; Cham, Switzerland: 2026. pp. 45–54. [ Google Scholar ] 8. Halicek M., Dormer J.D., Little J.V., Chen A.Y., Myers L., Sumer B.D., Fei B. Hyperspectral Imaging of Head and Neck Squamous Cell Carcinoma for Cancer Margin Detection in Surgical Specimens from 102 Patients Using Deep Learning. Cancers. 2019;11:1367. doi: 10.3390/cancers11091367. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. Karadaghy O.A., Shew M., New J., Bur A.M. Development and Assessment of a Machine Learning Model to Help Predict Survival Among Patients with Oral Squamous Cell Carcinoma. JAMA Otolaryngol. Neck Surg. 2019;145:1115–1120. doi: 10.1001/jamaoto.2019.0981. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. Formeister E.J., Baum R., Knott P.D., Seth R., Ha P., Ryan W., El-Sayed I., George J., Larson A., Plonowska K., et al. Machine Learning for Predicting Complications in Head and Neck Microvascular Free Tissue Transfer. Laryngoscope. 2020;130:E843–E849. doi: 10.1002/lary.28508. [ DOI ] [ PubMed ] [ Google Scholar ] 11. You J.G., Dbouk R.H., Landman A., Ting D.Y., Dutta S., Wang J.C., Centi A.J., Macfarlane M., Bechor E., Letourneau J., et al. Ambient Documentation Technology in Clinician Experience of Documentation Burden and Burnout. JAMA Netw. Open. 2025;8:e2528056. doi: 10.1001/jamanetworkopen.2025.28056. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 12. Jordan M.I., Mitchell T.M. Machine Learning: Trends, Perspectives, and Prospects. Science. 2015;349:255–260. doi: 10.1126/science.aaa8415. [ DOI ] [ PubMed ] [ Google Scholar ] 13. LeCun Y., Bengio Y., Hinton G. Deep Learning. Nature. 2015;521:436–444. doi: 10.1038/nature14539. [ DOI ] [ PubMed ] [ Google Scholar ] 14. Derry A., Krzywinski M., Altman N. Convolutional Neural Networks. Nat. Methods. 2023;20:1269–1270. doi: 10.1038/s41592-023-01973-1. [ DOI ] [ PubMed ] [ Google Scholar ] 15. Derry A., Krzywinski M., Altman N. Neural Networks Primer. Nat. Methods. 2023;20:165–167. doi: 10.1038/s41592-022-01747-1. [ DOI ] [ PubMed ] [ Google Scholar ] 16. Ahn J., Kang B.G., Chang M., Yoon S. Applications and Future Perspectives of Large Language Models in Otolaryngology-Head and Neck Surgery: A Comprehensive Survey. Clin. Exp. Otorhinolaryngol. 2025;18:283–295. doi: 10.21053/ceo.2025-00121. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Teo Z.L., Thirunavukarasu A.J., Elangovan K., Cheng H., Moova P., Soetikno B., Nielsen C., Pollreisz A., Ting D.S.J., Morris R.J.T., et al. Generative Artificial Intelligence in Medicine. Nat. Med. 2025;31:3270–3282. doi: 10.1038/s41591-025-03983-2. [ DOI ] [ PubMed ] [ Google Scholar ] 18. Asgari E., Montaña-Brown N., Dubois M., Khalil S., Balloch J., Yeung J.A., Pimenta D. A Framework to Assess Clinical Safety and Hallucination Rates of LLMs for Medical Text Summarisation. npj Digit. Med. 2025;8:274. doi: 10.1038/s41746-025-01670-7. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 19. Amugongo L.M., Mascheroni P., Brooks S., Doering S., Seidel J. Retrieval Augmented Generation for Large Language Models in Healthcare: A Systematic Review. PLoS Digit. Health. 2025;4:e0000877. doi: 10.1371/journal.pdig.0000877. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 20. Yang R., Ning Y., Keppo E., Liu M., Hong C., Bitterman D.S., Ong J.C.L., Ting D.S.W., Liu N. Retrieval-Augmented Generation for Generative Artificial Intelligence in Health Care. npj Health Syst. 2025;2:2. doi: 10.1038/s44401-024-00004-1. [ DOI ] [ Google Scholar ] 21. Young T., Au Yeung J., Sambasivan K., Adjogatse D., Kong A., Petkar I., Reis Ferreira M., Lei M., King A., Teo J., et al. Natural Language Processing to Extract Head and Neck Cancer Data from Unstructured Electronic Health Records. Clin. Oncol. 2025;41:103805. doi: 10.1016/j.clon.2025.103805. [ DOI ] [ PubMed ] [ Google Scholar ] 22. Tabassi E. Artificial Intelligence Risk Management Framework (AI RMF 1.0) National Institute of Standards and Technology (U.S.); Gaithersburg, MD, USA: 2023. [ Google Scholar ] 23. U.S. Department of Health and Human Services . Guidance on De-Identification of Protected Health Information. U.S. Department of Health and Human Services; Washington, DC, USA: 2012. [ Google Scholar ] 24. World Health Organization . Ethics and Governance of Artificial Intelligence for Health: WHO Guidance. 1st ed. World Health Organization; Geneva, Switzerland: 2021. [ Google Scholar ] 25. U.S. Food and Drug Administration . Good Machine Learning Practice for Medical Device Development: Guiding Principles. FDA; Silver Spring, MD, USA: 2025. [ Google Scholar ] 26. Lawrence K., Kuram V.S., Levine D.L., Sharif S., Polet C., Malhotra K., Owens K. Informed Consent for Ambient Documentation Using Generative AI in Ambulatory Care. JAMA Netw. Open. 2025;8:e2522400. doi: 10.1001/jamanetworkopen.2025.22400. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Small W.R., Austrian J., O’Donnell L., Burk-Rafel J., Hochman K.A., Goodman A., Zaretsky J., Martin J., Johnson S., Major V.J., et al. Evaluating Hospital Course Summarization by an Electronic Health Record–Based Large Language Model. JAMA Netw. Open. 2025;8:e2526339. doi: 10.1001/jamanetworkopen.2025.26339. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 28. Garcia P., Ma S.P., Shah S., Smith M., Jeong Y., Devon-Sand A., Tai-Seale M., Takazawa K., Clutter D., Vogt K., et al. Artificial Intelligence–Generated Draft Replies to Patient Inbox Messages. JAMA Netw. Open. 2024;7:e243201. doi: 10.1001/jamanetworkopen.2024.3201. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 29. Kim J., Chen M.L., Rezaei S.J., Liang A.S., Seav S.M., Onyeka S., Lee J.J., Vedak S.C., Mui D., Lal R.A., et al. Perspectives on Artificial Intelligence–Generated Responses to Patient Messages. JAMA Netw. Open. 2024;7:e2438535. doi: 10.1001/jamanetworkopen.2024.38535. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 30. Zaretsky J., Kim J.M., Baskharoun S., Zhao Y., Austrian J., Aphinyanaphongs Y., Gupta R., Blecker S.B., Feldman J. Generative Artificial Intelligence to Transform Inpatient Discharge Summaries to Patient-Friendly Language and Format. JAMA Netw. Open. 2024;7:e240357. doi: 10.1001/jamanetworkopen.2024.0357. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Ayers J.W., Poliak A., Dredze M., Leas E.C., Zhu Z., Kelley J.B., Faix D.J., Goodman A.M., Longhurst C.A., Hogarth M., et al. Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Public Social Media Forum. JAMA Intern. Med. 2023;183:589–596. doi: 10.1001/jamainternmed.2023.1838. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 32. Campbell L.O., Babb K., Lambie G.W., Hayes B.G. An Examination of Generative AI Response to Suicide Inquires: Content Analysis. JMIR Ment. Health. 2025;12:e73623. doi: 10.2196/73623. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 33. Zhang Q., Zhang R., Xiong Y., Sui Y., Tong C., Lin F.-H. Generative AI Mental Health Chatbots as Therapeutic Tools: Systematic Review and Meta-Analysis of Their Role in Reducing Mental Health Issues. J. Med. Internet Res. 2025;27:e78238. doi: 10.2196/78238. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 34. van Dijk L.V., Fuller C.D. Artificial Intelligence and Radiomics in Head and Neck Cancer Care: Opportunities, Mechanics, and Challenges. Am. Soc. Clin. Oncol. Educ. Book. 2021;41:e225–e235. doi: 10.1200/EDBK_320951. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 35. Tobias M.A.S., Nogueira B.P., Santana M.C.S., Pires R.G., Papa J.P., Santos P.S.S. Artificial Intelligence for Oral Cancer Diagnosis: What Are the Possibilities? Oral Oncol. 2022;134:106117. doi: 10.1016/j.oraloncology.2022.106117. [ DOI ] [ PubMed ] [ Google Scholar ] 36. Azam M.A., Sampieri C., Ioppi A., Africano S., Vallin A., Mocellin D., Fragale M., Guastini L., Moccia S., Piazza C., et al. Deep Learning Applied to White Light and Narrow Band Imaging Videolaryngoscopy: Toward Real-Time Laryngeal Cancer Detection. Laryngoscope. 2022;132:1798–1806. doi: 10.1002/lary.29960. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 37. Bur A.M., Zhang T., Chen X., Kavookjian H., Kraft S., Karadaghy O., Farrokhian N., Mussatto C., Penn J., Wang G. Interpretable Computer Vision to Detect and Classify Structural Laryngeal Lesions in Digital Flexible Laryngoscopic Images. Otolaryngol. Head Neck Surg. 2023;169:1564–1572. doi: 10.1002/ohn.411. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 38. Wellenstein D.J., Woodburn J., Marres H.A.M., van den Broek G.B. Detection of Laryngeal Carcinoma during Endoscopy Using Artificial Intelligence. Head Neck. 2023;45:2217–2226. doi: 10.1002/hed.27441. [ DOI ] [ PubMed ] [ Google Scholar ] 39. Turki A., Alshabrawy O., Woo W.L. Multimodal Deep Learning for Stage Classification of Head and Neck Cancer Using Masked Autoencoders and Vision Transformers with Attention-Based Fusion. Cancers. 2025;17:2115. doi: 10.3390/cancers17132115. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 40. Unlu O., Shin J., Mailly C.J., Oates M.F., Tucci M.R., Varugheese M., Wagholikar K., Wang F., Scirica B.M., Blood A.J., et al. Retrieval-Augmented Generation–Enabled GPT-4 for Clinical Trial Screening. NEJM AI. 2024;1:AIoa2400181. doi: 10.1056/AIoa2400181. [ DOI ] [ Google Scholar ] 41. Dankbaar J.W., Pameijer F.A. Radiological Assessment of Extranodal Extension in Patients with Head and Neck Squamous Cell Carcinoma. J. Cancer Metastasis Treat. 2021;7:56. doi: 10.20517/2394-4722.2021.112. [ DOI ] [ Google Scholar ] 42. Park S.I., Guenette J.P., Suh C.H., Hanna G.J., Chung S.R., Baek J.H., Lee J.H., Choi Y.J. The Diagnostic Performance of CT and MRI for Detecting Extranodal Extension in Patients with Head and Neck Squamous Cell Carcinoma: A Systematic Review and Diagnostic Meta-Analysis. Eur. Radiol. 2021;31:2048–2061. doi: 10.1007/s00330-020-07281-y. [ DOI ] [ PubMed ] [ Google Scholar ] 43. Aggarwal A., Jana M., Singh A., Dam T., Maurya H., Pathak T., Orsulic S., Yang K., Chute D., Bishop J.A., et al. Artificial Intelligence-Based Virtual Staining Platform for Identifying Tumor-Associated Macrophages from Hematoxylin and Eosin-Stained Images. Eur. J. Cancer. 2025;220:115390. doi: 10.1016/j.ejca.2025.115390. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Li Y., Luijten G., Gsaxner C., Grunert K., Bader A., Hölzle F., Röhrig R., de la Fuente M., Egger J., Xie K., et al. Usability Study of Augmented Reality Visualization Modalities on Localization Accuracy in the Head and Neck: Randomized Crossover Trial. JMIR Serious Games. 2026;14:e75962. doi: 10.2196/75962. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 45. Shi Z., Peng Y., Gao X., Chen S., Chen G., Pan G., Liang Z., Wang Z. Translating High-Precision Mixed Reality Navigation from Lab to Operating Room: Design and Clinical Evaluation. BMC Surg. 2025;25:331. doi: 10.1186/s12893-025-03096-0. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 46. Qi Z., Jin H., Wang Q., Gan Z., Xiong R., Zhang S., Liu M., Wang J., Ding X., Chen X., et al. The Feasibility and Accuracy of Holographic Navigation with Laser Crosshair Simulator Registration on a Mixed-Reality Display. Sensors. 2024;24:896. doi: 10.3390/s24030896. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 47. Su Y., Thieringer F.M., Fernandes R., Parmar S. Editorial: Virtual Surgical Planning and 3d Printing in Head and Neck Tumor Resection and Reconstruction. Front. Oncol. 2022;12:960545. doi: 10.3389/fonc.2022.960545. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 48. Prasad K., Miller A., Sharif K., Colazo J.M., Ye W., Necker F., Baik F., Lewis J.S., Rosenthal E., Wu J.Y., et al. Augmented-Reality Surgery to Guide Head and Neck Cancer Re-Resection: A Feasibility and Accuracy Study. Ann. Surg. Oncol. 2023;30:4994–5000. doi: 10.1245/s10434-023-13532-1. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 49. Aweeda M., Adegboye F., Yang S.F., Topf M.C. Enhancing Surgical Vision: Augmented Reality in Otolaryngology-Head and Neck Surgery. J. Med. Ext. Real. 2024;1:124–136. doi: 10.1089/jmxr.2024.0010. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 50. Pertzborn D., Nguyen H.-N., Hüttmann K., Prengel J., Ernst G., Guntinas-Lichius O., von Eggeling F., Hoffmann F. Intraoperative Assessment of Tumor Margins in Tissue Sections with Hyperspectral Imaging and Machine Learning. Cancers. 2022;15:213. doi: 10.3390/cancers15010213. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 51. Badhey A.K., Schwarz J.S., Laitman B.M., Veremis B.M., Westra W.H., Yao M., Teng M.S., Genden E.M., Miles B.A. Intraoperative Use of Wide-Field Optical Coherence Tomography to Evaluate Tissue Microstructure in the Oral Cavity and Oropharynx. JAMA Otolaryngol. Neck Surg. 2023;149:71–78. doi: 10.1001/jamaoto.2022.3763. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 52. Esperouz F., Ciavarella D., Lorusso M., Santarelli A., Lo Muzio L., Campisi G., Lo Russo L. Critical Review of OCT in Clinical Practice for the Assessment of Oral Lesions. Front. Oncol. 2025;15:1569197. doi: 10.3389/fonc.2025.1569197. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 53. Auner G.W., Koya S.K., Huang C., Broadbent B., Trexler M., Auner Z., Elias A., Mehne K.C., Brusatori M.A. Applications of Raman Spectroscopy in Cancer Diagnosis. Cancer Metastasis Rev. 2018;37:691–717. doi: 10.1007/s10555-018-9770-9. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 54. Chadokiya J., Chang K., Sharma S., Hu J., Lill J.R., Dionne J., Kirane A. Advancing Precision Cancer Immunotherapy Drug Development, Administration, and Response Prediction with AI-Enabled Raman Spectroscopy. Front. Immunol. 2025;15:1520860. doi: 10.3389/fimmu.2024.1520860. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 55. van Schaik J.E., Halmos G.B., Witjes M.J.H., Plaat B.E.C. An Overview of the Current Clinical Status of Optical Imaging in Head and Neck Cancer with a Focus on Narrow Band Imaging and Fluorescence Optical Imaging. Oral Oncol. 2021;121:105504. doi: 10.1016/j.oraloncology.2021.105504. [ DOI ] [ PubMed ] [ Google Scholar ] 56. Wu C., Gleysteen J., Teraphongphom N.T., Li Y., Rosenthal E. In-Vivo Optical Imaging in Head and Neck Oncology: Basic Principles, Clinical Applications and Future Directions. Int. J. Oral Sci. 2018;10:10. doi: 10.1038/s41368-018-0011-4. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 57. Knudsen J.E., Ghaffar U., Ma R., Hung A.J. Clinical Applications of Artificial Intelligence in Robotic Surgery. J. Robot. Surg. 2024;18:102. doi: 10.1007/s11701-024-01867-0. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 58. Marsden M., Weyers B.W., Bec J., Sun T., Gandour-Edwards R.F., Birkeland A.C., Abouyared M., Bewley A.F., Farwell D.G., Marcu L. Intraoperative Margin Assessment in Oral and Oropharyngeal Cancer Using Label-Free Fluorescence Lifetime Imaging and Machine Learning. IEEE Trans. Biomed. Eng. 2021;68:857–868. doi: 10.1109/TBME.2020.3010480. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 59. Yan Y.-L., Ren T., Ding L., Sun T., Huang S. Robust Prediction of Tool-Tissue Interaction Force Using ISSA-Optimized BP Neural Networks in Robotic Surgery. BMC Surg. 2025;25:368. doi: 10.1186/s12893-025-03121-2. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 60. Salehi A., Wang L., Coates P.J., Norberg Spaak L., Gu X., Sgaramella N., Nylander K. Reiterative Modeling of Combined Transcriptomic and Proteomic Features Refines and Improves the Prediction of Early Recurrence in Squamous Cell Carcinoma of Head and Neck. Comput. Biol. Med. 2022;149:105991. doi: 10.1016/j.compbiomed.2022.105991. [ DOI ] [ PubMed ] [ Google Scholar ] 61. Chinnery T., Arifin A., Tay K.Y., Leung A., Nichols A.C., Palma D.A., Mattonen S.A., Lang P. Utilizing Artificial Intelligence for Head and Neck Cancer Outcomes Prediction from Imaging. Can. Assoc. Radiol. J. 2021;72:73–85. doi: 10.1177/0846537120942134. [ DOI ] [ PubMed ] [ Google Scholar ] 62. Goshtasbi K., Yasaka T.M., Zandi-Toghani M., Djalilian H.R., Armstrong W.B., Tjoa T., Haidar Y.M., Abouzari M. Machine Learning Models to Predict Length of Stay and Discharge Destination in Complex Head and Neck Surgery. Head Neck. 2021;43:788–797. doi: 10.1002/hed.26528. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 63. Mascarella M.A., Muthukrishnan N., Maleki F., Kergoat M.-J., Richardson K., Mlynarek A., Forest V.-I., Reinhold C., Martin D.R., Hier M., et al. Above and Beyond Age: Prediction of Major Postoperative Adverse Events in Head and Neck Surgery. Ann. Otol. Rhinol. Laryngol. 2022;131:697–703. doi: 10.1177/00034894211041222. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 64. Tighe D., McMahon J., Schilling C., Ho M., Provost S., Freitas A. Machine Learning Methods Applied to Risk Adjustment of Cumulative Sum Chart Methodology to Audit Free Flap Outcomes after Head and Neck Surgery. Br. J. Oral Maxillofac. Surg. 2022;60:1353–1361. doi: 10.1016/j.bjoms.2022.09.007. [ DOI ] [ PubMed ] [ Google Scholar ] 65. Hernando-Calvo A., Yang S.Y.C., Vila-Casadesús M., Han M., Liu Z.A., Berman A.H.K., Spreafico A., Razak A.A., Lheureux S., Hansen A.R., et al. Combined Transcriptome and Circulating Tumor DNA Longitudinal Biomarker Analysis Associates with Clinical Outcomes in Advanced Solid Tumors Treated with Pembrolizumab. JCO Precis. Oncol. 2024;8:e2400100. doi: 10.1200/PO.24.00100. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 66. Fang S.-H., Tsao Y., Hsiao M.-J., Chen J.-Y., Lai Y.-H., Lin F.-C., Wang C.-T. Detection of Pathological Voice Using Cepstrum Vectors: A Deep Learning Approach. J. Voice. 2019;33:634–641. doi: 10.1016/j.jvoice.2018.02.003. [ DOI ] [ PubMed ] [ Google Scholar ] 67. Rao K.N., Fernandez-Alvarez V., Guntinas-Lichius O., Sreeram M.P., de Bree R., Kowalski L.P., Forastiere A., Pace-Asciak P., Rodrigo J.P., Saba N.F., et al. The Limitations of Artificial Intelligence in Head and Neck Oncology. Adv. Ther. 2025;42:2559–2568. doi: 10.1007/s12325-025-03198-4. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 68. Obermeyer Z., Powers B., Vogeli C., Mullainathan S. Dissecting Racial Bias in an Algorithm Used to Manage the Health of Populations. Science. 2019;366:447–453. doi: 10.1126/science.aax2342. [ DOI ] [ PubMed ] [ Google Scholar ] 69. Novi S.L., Navarathna N., D’Cruz M., Brooks J.R., Maron B.A., Isaiah A. Deep Learning in Otolaryngology: A Narrative Review. JAMA Otolaryngol. Neck Surg. 2026;152:71–80. doi: 10.1001/jamaoto.2025.3911. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 70. Rieke N., Hancox J., Li W., Milletarì F., Roth H.R., Albarqouni S., Bakas S., Galtier M.N., Landman B.A., Maier-Hein K., et al. The Future of Digital Health with Federated Learning. npj Digit. Med. 2020;3:119. doi: 10.1038/s41746-020-00323-1. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 71. For Trustworthy AI, Keep the Human in the Loop. Nat. Med. 2025;31:3207. doi: 10.1038/s41591-025-04033-7. [ DOI ] [ PubMed ] [ Google Scholar ] 72. Martín-Noguerol T., López-Úbeda P., Luna A. AI in Radiology: Legal Responsibilities and the Car Paradox. Eur. J. Radiol. 2024;175:111462. doi: 10.1016/j.ejrad.2024.111462. [ DOI ] [ PubMed ] [ Google Scholar ] 73. Cestonaro C., Delicati A., Marcante B., Caenazzo L., Tozzo P. Defining Medical Liability When Artificial Intelligence Is Applied on Diagnostic Algorithms: A Systematic Review. Front. Med. 2023;10:1305756. doi: 10.3389/fmed.2023.1305756. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 74. Habli I., Lawton T., Porter Z. Artificial Intelligence in Health Care: Accountability and Safety. Bull. World Health Organ. 2020;98:251–256. doi: 10.2471/BLT.19.237487. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 75. Xu H., Shuttleworth K.M.J. Medical Artificial Intelligence and the Black Box Problem: A View Based on the Ethical Principle of “Do No Harm”. Intell. Med. 2024;4:52–57. doi: 10.1016/j.imed.2023.08.001. [ DOI ] [ Google Scholar ] 76. Duffourc M.N., Cahill R.A. Legal Implications of AI Standard of Care Integration on Patients’ Informed Consent: Lessons from Surgery. npj Digit. Med. 2025;9:4. doi: 10.1038/s41746-025-02157-1. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 77. Moulaei K., Akhlaghpour S., Fatehi F. Patient Consent for the Secondary Use of Health Data in Artificial Intelligence (AI) Models: A Scoping Review. Int. J. Med. Inf. 2025;198:105872. doi: 10.1016/j.ijmedinf.2025.105872. [ DOI ] [ PubMed ] [ Google Scholar ] 78. El Arab R.A., Al Moosa O.A. Systematic Review of Cost Effectiveness and Budget Impact of Artificial Intelligence in Healthcare. npj Digit. Med. 2025;8:548. doi: 10.1038/s41746-025-01722-y. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 79. Khanna N.N., Maindarkar M.A., Viswanathan V., Fernandes J.F.E., Paul S., Bhagawati M., Ahluwalia P., Ruzsa Z., Sharma A., Kolluri R., et al. Economics of Artificial Intelligence in Healthcare: Diagnosis vs. Treatment. Healthcare. 2022;10:2493. doi: 10.3390/healthcare10122493. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 80. Wong A., Nallamothu B.K., Longhurst C.A., Singh K. Leveraging AI to Reduce Operational Healthcare Costs: Lessons from Other Industries. npj Health Syst. 2026;3:9. doi: 10.1038/s44401-026-00070-7. [ DOI ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Data Availability Statement No new data were created or analyzed in this study. Data sharing is not applicable to this article. Articles from Journal of Clinical Medicine are provided here courtesy of Multidisciplinary Digital Publishing Institute (MDPI) ACTIONS View on publisher site PDF (2.0 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top