Skip to main content
AAO-HNSF 2023 Annual Meeting & OTO Experience Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Back
Rate ePoster

Head and Neck Surgery

(140) Diagnostic Performance of Radiomics Modeling in Predicting HPV Status

Monday, October 2, 2023
2:45 PM – 3:45 PM East Coast USA Time
Has Audio

    Presenting Author(s)

  • Derek J. Vos, BS

    Medical Student
    Case Western Reserve University School of Medicine
    Cleveland, Ohio, United States

    • Co-Author(s)

  • NY

    Noah Max Yaffe, BS

    Case Western Reserve University School of Medicine, USA

  • CC

    Claudia I. Cabrera, MD, MS

    Clinical Research Fellow
    University Hospitals Cleveland Medical Center
    Cleveland, Ohio, United States

  • Nicole M. Fowler, MD

    Associate Professor
    University Hospitals Cleveland Medical Center
    Cleveland, Ohio, United States

    • Senior Author(s)

  • BD

    Brian D. D'Anza, MD

    Associate Professor, Medical Director
    University Hospitals Cleveland Medical Center
    Cleveland, OH, United States

Disclosure(s):
Derek J. Vos, BS: No relevant relationships to disclose.
Claudia I. Cabrera, MD, MS: No relevant relationships to disclose.
Nicole M. Fowler, MD: No relevant relationships to disclose.


Introduction: Radiomics is a branch of artificial intelligence that uses high-throughput extraction of radiologic imaging features to develop predictive models. This study seeks to assess the diagnostic performance and methodological quality of studies utilizing radiomics for the prediction of HPV status in patients with head and neck squamous cell carcinoma.

Methods: A comprehensive literature search of PubMed, OVID, Cochrane, Web of Science, and Scopus until June 7, 2022 was performed to identify eligible studies. Eligible studies were defined as those which applied machine learning algorithms to pre-treatment imaging data for the prediction of HPV status in patients with confirmed head and neck squamous cell carcinoma. Diagnostic performance of radiomics models were assessed using sensitivity, specificity, and area under the ROC curve (AUC), a commonly reported measurement of binary classification performance. Statistical analysis was performed using JMP Pro 16 and methodological quality of studies was assessed using the Radiomics Quality Score (RQS).

Results: A total of 17 studies (2,368 patients) were included in this systematic review. The mean AUC of included studies was 0.815, with a range of 0.7 to 0.953. Sensitivity ranged from 0.65 to 0.88 with a mean of 0.78. Specificity ranged from 0.53 to 0.93 with a mean of 0.78. A total of 5 studies included models combining clinical and radiomics features but the addition of clinical features did not have a significant effect on AUC compared to models using only radiomics features, t(13) = 0.7, p=.49. There was no significant difference in AUC for models using internal validation cohorts (M = 0.80, SD=0.05) compared to models using external validation cohorts (M = 0.82, SD=0.05), t(11) = -0.57, p = .58. The mean total RQS score for each study evaluated was 13.12 (SD = 1.03), with an intraclass correlation coefficient (ICC) of 0.889.

Conclusions: Radiomics modeling may serve as a diagnostic indicator for HPV status in head and neck cancer patients. However, the methodological quality of such studies in the literature remains a limitation for further clinical implementation.