Study Design & Imaging Protocol
Prospective data collection with standardized acquisition parameters
PROSPECTIVE STUDY DESIGN: This study was designed prospectively with predefined imaging protocols established at the initiation of data collection (2020-2024). All imaging acquisitions followed standardized protocols to ensure data quality and reproducibility.
Total Prospective Cohort
N = 1,466
Head and Neck Cancer patients treated at CMC Vellore (2020-2024)
CT Imaging Acquisition Parameters
CT Scanners: SIEMENS Biograph 6, SIEMENS SOMATOM Definition AS, GE Healthcare Discovery CT750 HD
Parameter Specification
Energy Range 100.0 - 130.0 kVp
Exposure Range 5.0 - 350.0 mAs
Slice Thickness 2.5 - 5.0 mm
In-Plane Resolution 0.78125 × 0.78125 mm² to 1.367188 × 1.367188 mm²
Contrast Protocol Contrast-enhanced and non-contrast imaging included
Primary Radiation Treatment Cohort
N = 367
Patients who received primary radiation ± chemotherapy (n = 1,099 excluded: surgery, palliative care, other treatments)
Patient Selection & Dataset Splitting
Application of inclusion/exclusion criteria to N = 367 primary radiation patients
Inclusion Criteria
  • 1. Diagnosed head and neck cancer
  • 2. Contrast-enhanced CT of head and neck available
  • 3. No treatment before CT scan (treatment-naïve imaging)
  • 4. Treatment with radiation and/or chemoradiation only
Exclusion Criteria (n = 204 excluded)
  • 1. Treatment non-completion (n = 2)
  • 2. Tumor volume unavailable - only primary disease visible (n = 13)
  • 3. Presence of significant image artifacts (n = 3)
  • 4. Did not receive radiation treatment as planned (n = 11)
  • 5. Follow-up of one year not available (n = 175)
Final Study Cohort
N = 163
Locoregional Recurrence (LRR)
55
33.7%
No Locoregional Recurrence
108
66.3%
Training Cohort
n = 130
(80% of dataset)
Used for feature selection, hyperparameter tuning, and cross-validation
Test Cohort
n = 33
(20% of dataset)
Held-out for final model evaluation