Radial Data Mining to Identify Density-Dose Interactions That Predict Distant Failure Following SABR.

Area of Science:

• Radiation Oncology
• Medical Imaging
• Data Mining

Background:

• Lower incidental doses in lung stereotactic radiotherapy correlate with increased distant metastasis (DM).
• Pretreatment computed tomography (CT) tumor density is linked to microscopic disease (MDE) risk.
• The interaction between imaging biomarkers and incidental dose remains unexplored.

Purpose of the Study:

• To investigate density-dose interactions radially from the gross tumor volume (GTV) to predict DM.
• To determine if imaging can inform dose escalation decisions for patients at risk.
• To explore the impact of incidental dose on outcome based on tumor imaging characteristics.

Main Methods:

• Quantified dose and density in 1-mm annuli around the GTV for 199 lung cancer patients.
• Utilized Cox regressions with dose-density interactions in independent annuli.
• Created heatmaps to visualize DM prediction improvement due to interactions.

Main Results:

• Dose-density interactions significantly improved DM prediction in over 50% of bootstrap resamples.
• Tumor density variance and high peritumour density were associated with DM in patients with specific dose distributions outside the GTV.
• Identified dose-density associations were independent of the mean GTV dose.

Conclusions:

• High tumor variance and peritumour density increase DM risk with low/non-uniform incidental doses outside the GTV.
• Incidental dose plays a crucial role in controlling occult disease, independent of tumor dose.
• This methodology enables spatial dose-density interaction studies, potentially accelerating clinical implementation of imaging biomarkers.