Simeng Zhu, MD

Tailoring Treatment for Head and Neck Cancer
 

Simeng Zhu, MD, and Farzan Siddiqui, MD, are using artificial intelligence (AI) to personalize treatment for patients with head and neck cancer. The standard treatment for oropharyngeal cancer (OPC) is radiation therapy (RT) with chemotherapy. Although many patients with OPC have a normal life expectancy following treatment, they frequently experience significant toxicities from RT and chemotherapy. Identifying which patients are least likely to experience either locoregional recurrence near the original tumor site or more distant metastases could help determine when treatment could be de-escalated to reduce toxicities. Additionally, treatment, surveillance and follow-up could be intensified for patients with the highest risk of recurrence.

The team is developing an innovative AI-based model that uses pre-treatment imaging and clinical data to predict the risks of locoregional recurrence and distance metastases more precisely. Current methods for selecting OPC treatment options vary, but factors usually include human papilloma virus (HPV) status, tumor stage and other clinical characteristics. Past studies have shown that HPV-positive patients with OPC have more favorable overall survival and lower locoregional recurrence, but distant metastases develop at a similar rate regardless of HPV status, indicating the need for additional markers. By combining tumor image features with existing clinical characteristics in their new AI-based model, Dr. Zhu and Dr. Siddiqui aim to improve risk stratification that can support more personalized decisions about the best treatment options. With support from ROI, the team is:

• Building on prior work to enlarge their curated dataset of over 1,300 patients with OPC that were treated with RT with or without chemotherapy.
• Using convolutional neural networks to construct baseline models for imaging-based OPC outcome prediction.
• Improving the model prediction accuracy and interpretability to help foster trust in clinical AI models.

The anticipated result of the team's research is the development of a clinical decision support tool that would provide a better estimate of the risk of locoregional recurrence or distant metastases for patients with OPC. If successful, their next steps would include conducting a prospective trial of the tool and developing strategies to implement the tool in clinical workflows. Eventually, their novel AI-based approach that leverages information from noninvasive imaging could help personalize treatment decisions and improve outcomes for patients with OPC.

Presentations
 

• Imaging-Based Prognostic Artificial Intelligence Model for Oropharyngeal Carcinoma After Radiation Therapy was presented at the 2024 Multidisciplinary Head and Neck Cancers Symposium.