AI-Assisted Endoscopy for Upper Aerodigestive Tract Lesions (H&NANCE)

Head&Neck Application of Novel Computer-assisted Endoscopy

This is a prospective observational clinical study designed to evaluate the performance of artificial intelligence (AI) algorithms applied to upper aerodigestive tract (UADT) video-endoscopy. The study assesses three main tasks: lesion detection (localization), classification (benign vs malignant), and segmentation of tumor margins.

AI algorithms will be applied to endoscopic video data acquired during routine clinical practice without influencing clinical decision-making. The system will process images in real time and store data for subsequent analysis. AI outputs will be compared with physician assessment and reference standard histopathology to evaluate diagnostic performance.

Study Overview

• Status: Not yet recruiting
• Conditions: Leukoplakia; Squamous Cell Carcinoma; Vocal Fold Polyp; Vocal Fold Nodules; Upper Aerodigestive Tract Neoplasms; Upper Aero-digestive Tract (UADT) Neoplasm; Vocal Cord Disease; Head and Neck Cancer (H&N); Upper AerodigestiveTract Cancer

Detailed Description

The artificial intelligence algorithms developed will be employed in the analysis of laryngeal lesions for 3 tasks:

• Task 1: Computer aided diagnosis (CADx): the algorithm provides a differential diagnosis between benign and malignant neoplasms (binary classification) and the exact histology (multiclass classification). During the UADT video-endoscopy in the outpatient clinic, the physician performs the video-endoscopy and selects and captures n.3 WL and n.3 NBI significant frames of the lesion. The AI model records the classification output of the algorithm that the physician cannot access. The predicted pathologic results will be finally displayed as two different classifications along with the probability of each prediction (0% to 100%) as estimated by the AI algorithm: a first binary classification "neoplastic" or "non-neoplastic," and a second multiclass classification with the exact histology. The physician subsequently, based on the endoscopic examination, will write the suspected diagnosis (benign vs. malignant lesion and the actual histology) in the appropriate patient chart. Next, the physician reviews the screenshot taken and makes sure the lesion is visible in every one of them. Retrospectively, an investigator (blinded to the physician's assessment) will review the AI processed frames with the resulting CADx classifications and mark the AI-processed diagnosis in the patient chart. Once biopsied, the final histology of the lesion analyzed by definitive histopathological examination is recorded in the patient chart by the investigator. The investigators will finally compare the two recorded diagnoses (CADx and physician) with the definitive histology.
• Task 2: Computer aided detection (CADe): the algorithm, through the representation of a rectangle (bounding box), localizes the lesion during the video-endoscopy in the outpatient clinic in real-time. During the UADT video-endoscopy, the physician performs the video-endoscopy as for standard-of-care procedure. In parallel, the AI model processes in real-time the endoscopic video and records the output of the algorithm (which the physician cannot access). The physician captures n.3 WL and n.3 NBI significant frames of the lesion. Moreover, n.3 frames where no lesions are visible are captured as negative controls. Later, the physician reviews the screenshot taken and makes sure to label the frames where the lesion is visible as "positive cases" and the frame where the lesion is not visible as "negative cases". The investigators will finally assess if the lesion was detected by the CADe system to define a "true positive". Similarly, to define a true negative, the CADe system should have not output a bounding box in the majority of the "negative cases" frames.
• Task 3: Computer aided segmentation (CASe): the algorithm analyzes the neoplasm margins and provides a delineation mask. In the operating room setting, once the lesion to be resected is identified with a 0° telescope, the surgeon captures n.1 WL and n.1 NBI close-up photographs that exemplify the superficial lesion margins. The same procedure is repeated with a 70° optics and other two photographs are acquired. The frames taken are then saved and analyzed by the AI algorithm, which will perform the segmentation task. The surgeon will be blinded to the AI prediction. Later, the surgeon will draw the margins of the lesion according to her/his evaluation of each captured frame. The annotated frame will be saved so that it can be analyzed at a later time. Afterwards, in cases where positive superficial margins are identified by histopathologic examination, the surgeon-designed margins and the AI model ones will be compared to see if there was any difference in the suggested margin.

• Study Type: Observational
• Enrollment (Estimated): 283

Contacts and Locations

• Study Contact: Name: Leonardo De Mattos, PhD; Phone Number: +39 010 2898 270; Email: leonardo.demattos@iit.it

Study Locations

• Belgium
  • Flemish Brabant
    • Leuven, Flemish Brabant, Belgium, 3000
      • UZ Leuven
      • Contact: Vincent Vander Poorten, Prof. and ENT Surgeon; Phone Number: +32 16 33 23 42; Email: vincent.vanderpoorten@uzleuven.be
      • Principal Investigator: Vincent Vander Poorten, MD PhD
• Italy
  • GE
    • Genova, GE, Italy, 16131
      • IRCCS Ospedale Policlinico San Martino
      • Contact: Francesco Mora, Otorhinolaryngologist and Prof; Phone Number: +39 010-5557479; Email: Francesco.Mora@unige.it
      • Principal Investigator: Francesco Mora, MD PhD
• Spain
  • Barcelona
    • Barcelona, Barcelona, Spain, 08036
      • Hospital Clinic De Barcelona
      • Contact: Claudio Sampieri, Otorhinolaryngologist; Phone Number: +34 93 227 57 66; Email: claudio.sampieri@outlook.com
      • Principal Investigator: Claudio Sampieri, MD PhD

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Non-Probability Sample

Study Population

The sample size for a single-arm, prospective cohort study, where the same group of patients undergoes testing with an AI model and a human physician and then they are compared with the gold standard reference test (biopsy) was calculated based on information from previous studies

Description

Inclusion Criteria:

• Age > 18 years
• Injury originating from the upper aero-digestive tract
• Recording of the video-endoscopic examination
• Patient known to undergo a biopsy of the lesion or clinical follow-up for lesion with known biopsy (e.g. laryngeal papillomatosis) or suffering from Reinke's edema (in this pathology, in fact, biopsy is not necessary since the diagnosis is clinical)
• Or patients undergoing transoral lesion excision

Exclusion Criteria:

• Submucosal lesion
• Patients with previous operations on the upper aero-digestive tract
• Patients with previous radiotherapy of the head and neck district
• Poor compliance on endoscopic examination
• Unavailability of CADe/CADx or CASe data logging note

Study Plan

How is the study designed?

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Negative Predictive Value of the CADx Algorithm for Malignant or Premalignant Upper Aerodigestive Tract Lesions	Negative Predictive Value (NPV) of the computer-aided diagnosis (CADx) algorithm for classifying UADT lesions as malignant/premalignant versus benign/non-neoplastic, using definitive histopathology as the reference standard. The CADx final classification will be based on the majority rule across selected white-light and narrow-band imaging frames. NPV = true negatives / (true negatives + false negatives). The pre-specified performance target is NPV ≥ 90%.	From index outpatient UADT video-endoscopy until definitive histopathology result is available, assessed up to 60 days after endoscopy.
Sensitivity of the CADe Algorithm for Localization of Upper Aerodigestive Tract Lesions	Sensitivity of the computer-aided detection (CADe) algorithm for localizing UADT lesions with a bounding box. A true positive is defined as localization of the lesion area by a bounding box in the majority of physician-labeled lesion-positive captured frames. Sensitivity = true positives / (true positives + false negatives).	At index outpatient UADT video-endoscopy, with blinded post-processing assessment performed up to 30 days after endoscopy.
Median Intersection Over Union Between CASe Segmentation and Surgeon-Drawn Lesion Margins	Median overlap between the AI-generated segmentation mask and the lesion margin area drawn by the surgeon on intraoperative endoscopic images. Intersection over Union (IoU) = area of overlap / area of union. Values range from 0 to 1; higher values indicate greater agreement.	At intraoperative pre-resection endoscopy, with assessment performed after image annotation up to 30 days after surgery.
Median Dice Similarity Coefficient Between CASe Segmentation and Surgeon-Drawn Lesion Margins	Median Dice Similarity Coefficient (DSC) between the AI-generated segmentation mask and the lesion margin area drawn by the surgeon on intraoperative endoscopic images. Dice Similarity Coefficient = 2 × area of overlap / (AI segmented area + surgeon-drawn area). Values range from 0 to 1; higher values indicate greater agreement.	At intraoperative pre-resection endoscopy, with assessment performed after image annotation up to 30 days after surgery.

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
WL-NPV vs. NBI-NPV of CADx classification	Negative Predictive Value (NPV) of CADx classification calculated using only the three selected white-light frames, compared with definitive histopathology, vs. NPV of CADx classification calculated using only the three selected narrow-band imaging frames, compared with definitive histopathology. The final AI-result will be calculated based on the majority rule of the 3 WL and 3 NBI frames computed separately.	From index outpatient UADT video-endoscopy until definitive histopathology result is available, assessed up to 60 days after endoscopy.
Clinician-Reported Usability Score for the AI Endoscopy System	Usability of the AI endoscopy system assessed using standardized usability questionnaires administered to clinicians after use of the AI system. Questionnaire scoring will be interpreted according to the selected questionnaire manual, with higher scores indicating greater usability.	Assessed after clinician use of the AI system during study procedures, up to 20 months after study initiation.
Sensitivity, Specificity and Accuracy of CADx histology prediction	Sensitivity, Specificity and Accuracy of the CADx algorithm for histology prediction, compared with definitive histopathology.	From index outpatient UADT video-endoscopy until definitive histopathology result is available, assessed up to 60 days after endoscopy.
F1 Score of CADx Classification	F1 score of the CADx classification output compared with definitive histopathology.	From index outpatient UADT video-endoscopy until definitive histopathology result is available, assessed up to 60 days after endoscopy.
Area Under the Receiver Operating Characteristic Curve of CADx Classification	AUC of the ROC curve for CADx classification of UADT lesions compared with definitive histopathology.	From index outpatient UADT video-endoscopy until definitive histopathology result is available, assessed up to 60 days after endoscopy.
Sensitivity, Specificity and Accuracy of human physician histology prediction	Sensitivity, Specificity an Accuracy of the treating physician's suspected diagnosis compared with definitive histopathology.	From index outpatient UADT video-endoscopy until definitive histopathology result is available, assessed up to 60 days after endoscopy.
Specificity of the CADe Algorithm for Localization of Upper Aerodigestive Tract Lesions	Proportion of physician-labeled lesion-negative frames/cases in which the CADe algorithm does not output a bounding box in the majority of negative-control frames.	At index outpatient UADT video-endoscopy, with blinded post-processing assessment performed up to 30 days after endoscopy.
Accuracy of the CADe Algorithm for Localization of Upper Aerodigestive Tract Lesions	Overall proportion of correctly classified lesion-positive and lesion-negative cases/frames by the CADe algorithm.	At index outpatient UADT video-endoscopy, with blinded post-processing assessment performed up to 30 days after endoscopy.
Positive Predictive Value of the CADe Algorithm for Localization of Upper Aerodigestive Tract Lesions	Positive predictive value of CADe bounding-box output for lesion localization. PPV = true positives / (true positives + false positives).	At index outpatient UADT video-endoscopy, with blinded post-processing assessment performed up to 30 days after endoscopy.
Negative Predictive Value of the CADe Algorithm for Localization of Upper Aerodigestive Tract Lesions	Negative predictive value of CADe absence of bounding-box output for lesion localization. NPV = true negatives / (true negatives + false negatives).	At index outpatient UADT video-endoscopy, with blinded post-processing assessment performed up to 30 days after endoscopy.
Percentage of Positive Superficial Margin Cases in Which the AI-Predicted Tumor Area Is Wider Than the Surgeon-Drawn Area	Among cases with positive superficial margins on final histopathology, percentage of cases in which the AI-predicted tumor area extends beyond the surgeon-drawn margin at the affected margin.	At intraoperative pre-resection endoscopy, with assessment performed after image annotation up to 30 days after surgery.

Collaborators and Investigators

• Sponsor: Istituto Italiano di Tecnologia
• Collaborators: Universitaire Ziekenhuizen KU Leuven; Hospital Clinic of Barcelona; Universita degli Studi di Genova; CRO "Centro Clinical Trials" IRCCS Ospedale Policlinico San Martino; Università degli Studi 'G. d'Annunzio' Chieti e Pescara; Scuola Superiore Sant'Anna di Pisa

Publications and helpful links

General Publications

• Rex DK, Kahi C, O'Brien M, Levin TR, Pohl H, Rastogi A, Burgart L, Imperiale T, Ladabaum U, Cohen J, Lieberman DA. The American Society for Gastrointestinal Endoscopy PIVI (Preservation and Incorporation of Valuable Endoscopic Innovations) on real-time endoscopic assessment of the histology of diminutive colorectal polyps. Gastrointest Endosc. 2011 Mar;73(3):419-22. doi: 10.1016/j.gie.2011.01.023.
• Dunham ME, Kong KA, McWhorter AJ, Adkins LK. Optical Biopsy: Automated Classification of Airway Endoscopic Findings Using a Convolutional Neural Network. Laryngoscope. 2022 Feb;132 Suppl 4:S1-S8. doi: 10.1002/lary.28708. Epub 2020 Apr 28.
• Piazza C, Peretti G, Vander Poorten V. Editorial: Advances in Transoral Approaches for Laryngeal Cancer. Front Oncol. 2018 Oct 17;8:455. doi: 10.3389/fonc.2018.00455. eCollection 2018. No abstract available.
• Paderno A, Piazza C, Del Bon F, Lancini D, Tanagli S, Deganello A, Peretti G, De Momi E, Patrini I, Ruperti M, Mattos LS, Moccia S. Deep Learning for Automatic Segmentation of Oral and Oropharyngeal Cancer Using Narrow Band Imaging: Preliminary Experience in a Clinical Perspective. Front Oncol. 2021 Mar 24;11:626602. doi: 10.3389/fonc.2021.626602. eCollection 2021.
• Azam MA, Sampieri C, Ioppi A, Africano S, Vallin A, Mocellin D, Fragale M, Guastini L, Moccia S, Piazza C, Mattos LS, Peretti G. Deep Learning Applied to White Light and Narrow Band Imaging Videolaryngoscopy: Toward Real-Time Laryngeal Cancer Detection. Laryngoscope. 2022 Sep;132(9):1798-1806. doi: 10.1002/lary.29960. Epub 2021 Nov 25.
• Ren J, Jing X, Wang J, Ren X, Xu Y, Yang Q, Ma L, Sun Y, Xu W, Yang N, Zou J, Zheng Y, Chen M, Gan W, Xiang T, An J, Liu R, Lv C, Lin K, Zheng X, Lou F, Rao Y, Yang H, Liu K, Liu G, Lu T, Zheng X, Zhao Y. Automatic Recognition of Laryngoscopic Images Using a Deep-Learning Technique. Laryngoscope. 2020 Nov;130(11):E686-E693. doi: 10.1002/lary.28539. Epub 2020 Feb 18.
• Kim DH, Kim Y, Kim SW, Hwang SH. Use of narrowband imaging for the diagnosis and screening of laryngeal cancer: A systematic review and meta-analysis. Head Neck. 2020 Sep;42(9):2635-2643. doi: 10.1002/hed.26186. Epub 2020 May 4.

Study record dates

Study Major Dates

• Study Start (Estimated): May 1, 2026
• Primary Completion (Estimated): December 1, 2027
• Study Completion (Estimated): February 1, 2028

Study Registration Dates

• First Submitted: May 5, 2026
• First Submitted That Met QC Criteria: May 12, 2026
• First Posted (Actual): May 19, 2026

Study Record Updates

• Last Update Posted (Actual): May 19, 2026
• Last Update Submitted That Met QC Criteria: May 12, 2026
• Last Verified: May 1, 2026

More Information

Terms related to this study

• Additional Relevant MeSH Terms: Neoplasms by Site; Pathological Conditions, Anatomical; Neoplasms by Histologic Type; Neoplasms, Glandular and Epithelial; Carcinoma; Neoplasms, Squamous Cell; Precancerous Conditions; Pathological Conditions, Signs and Symptoms; Neoplasms; Carcinoma, Squamous Cell; Head and Neck Neoplasms; Leukoplakia
• Other Study ID Numbers: IIT_AIRCARE_H&NANCE

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: UNDECIDED

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No