Skip to main content
AAO-HNSF 2024 Annual Meeting & OTO EXPO Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Back

Scientific Oral Presentations (Best of Scientific Oral)

Care Equity, Quality, and Safety

Session: Best of Scientific Oral Presentations 1

A Novel Artificial Intelligence Model for Optimizing Treatment of Salivary Gland Malignancies

Saturday, September 28, 2024
1:06 PM – 1:12 PM EDT
Location: MBCC, 222
Disclosure(s):
Andrew G. Girgis, BS:

faculty for this accredited education activity has no relationship(s) with ineligible companies to disclose.

Bishoy M. Galoaa, MS:

faculty for this accredited education activity has no relationship(s) with ineligible companies to disclose.

Anand K. Devaiah, MD:

faculty for this accredited education activity has no relationship(s) with ineligible companies to disclose.

Introduction: Artificial intelligence (AI) is increasingly accessible in healthcare, and holds potential for Otolaryngology treatment optimization. Given the challenges in treating salivary gland malignancy, the present study focuses on optimizing treatment for these cancers.

Methods:

Methods: A proprietary AI model for data analysis, harmonization, and treatment optimization was developed, trained, and tested using the SEER dataset. The AI optimized treatment by tumor histology, stage, location, survival, and personalized by demographics (age, sex, race, income). Multiple one-way-ANOVA were used to assess for model bias from demographic substratification.

Results:

Results: 20,174 salivary gland malignancy patients were analyzed by the AI to optimize treatment for each histology and stratified by demographics to personalize treatment. Overall AI-optimized survival was better than the cohort (t-test 3.16, p=0.0016). The AI optimization improved survival for adenoid cystic (n=7398, t-test: 2.750, p-value: 0.006), and less common tumors (n=16918, t-test= 2.811, p=0.0049). There was no difference for mucoepidermoid (n=7396, t-test 0.549, p= 0.583) nor acinic cell (n=3596, t-test 0.668, p=0.504).

In testing for bias, there were no significant differences when comparing AI-optimized and the non-optimized cohorts using ANOVA (Race: F=1.52, p=0.194; Age: F=1.05, p=0.395; Sex: F=2.89, p=0.089; and Income: F=1.18, p=0.300).

Conclusions:

Conclusion: AI can be used for large scale data analytics to optimize and personalize treatment, as seen in the present study with our AI model. Using demographics to personalize treatment does not introduce significant bias. Further validation, training, and clinical correlation is recommended.

    Presenting Author(s)

  • Andrew G. Girgis, BS

    Medical Student
    Boston University School of Medicine
    Belmont, MA, United States

    Disclosure(s):

    faculty for this accredited education activity has no relationship(s) with ineligible companies to disclose.

  • Bishoy M. Galoaa, MS

    Computer Vision Researcher ACLab Northeastern
    College Of Engineering Electrical and Computer Engineering Department, Northeastern University
    Boston, MA, United States

    Disclosure(s):

    faculty for this accredited education activity has no relationship(s) with ineligible companies to disclose.

    • Senior Author(s)

  • Anand K. Devaiah, MD

    Professor, Department of Otolaryngology, Neurosurgery, and Ophthalmology
    Boston Med Ctr
    Boston, Massachusetts, United States

    Disclosure(s):

    faculty for this accredited education activity has no relationship(s) with ineligible companies to disclose.