Supplementary file 1_Early toxicity prediction using longitudinal cone-beam CT delta-radio

Forty-three patients with HPV-positive oropharyngeal cancer underwent daily cone-beam CT imaging during radiotherapy. The dataset includes 43 radiomic features extracted from the supraglottic region, processed into weekly averages and ratio-to-baseline delta values to predict 6-12 month post-treatment dysphagia. Authored by Levent Sensoy and shared under a CC-BY-4.0 license, this 11.7 MB file was last updated in May 2026.

Use Cases

• Training machine learning models for early toxicity prediction based on longitudinal delta-radiomic features.
• Investigating the predictive power of specific radiomic features like global skewness for treatment outcomes.
• Developing parsimonious predictive models using feature subsets from specific temporal windows during treatment.

Strengths

• Data is derived from a specific, clinically relevant cohort of 43 HPV-positive oropharyngeal cancer patients.
• Features were processed using a defined methodology including weekly aggregation and a ratio-to-baseline formulation.
• Model performance is quantified with specific metrics, such as a mean AUC of 0.779 for the Weeks 1-4 model.

Limitations

• The cohort size of 43 patients is described as limited by the authors.
• Row count and column-level documentation are unknown, which may limit suitability assessment.
• The findings are exploratory, as external validation was not performed.

Provenance

Source

figshare

Collection Method

Radiomic features were extracted from daily cone-beam CT images of patients undergoing radiotherapy, then averaged weekly and transformed.

Freshness

Last updated 2026-05-21 04:35:56; freshness should be verified.