A Radiomics Approach to Identify Immunologically Active Tumor in Patients with Head and Neck Squamous Cell Carcinomas

Area of Science:

• Oncology
• Radiology
• Genomics

Background:

• A gene-expression-based HOT score identifies head and neck squamous cell carcinoma (HNSCC) hot/cold phenotype, correlating with immunotherapy response.
• The study aimed to assess if radiomic features from computed tomography (CT) scans can differentiate between hot and cold HNSCC phenotypes.

Purpose of the Study:

• To determine if radiomic profiling from CT scans can distinguish hot and cold HNSCC phenotypes.
• To develop a non-invasive method for predicting immunotherapy response in HNSCC patients.

Main Methods:

• 113 HNSCC patients from The Cancer Genome Atlas (TCGA) and 20 from Groupe Hospitalier Pitié-Salpêtrière (GHPS) with pre-treatment CT scans were analyzed.
• Radiomic features were extracted using IBEX software, and robust features were selected based on intraclass correlation coefficient (ICC).
• Machine learning models were trained on TCGA data and validated on GHPS data using area under the receiver operator characteristic curve (AUC).

Main Results:

• 144 radiomic features with ICC >0.9 were selected.
• An XGBoost model incorporating these features achieved an AUC of 0.86 in predicting the HNSCC hot/cold phenotype in the GHPS validation cohort.

Conclusions:

• A radiomic model effectively captures the hot/cold phenotype in HNSCC.
• This non-invasive approach may assist in identifying HNSCC patients suitable for immunotherapy.