External Validation of the Proliferation Saturation Index Model in Predicting Tumor Volume Regression in Patients With Non-Small Cell Lung Cancer Undergoing Radiation Therapy.
[PURPOSE] The Proliferation Saturation Index (PSI) model is a patient-specific mathematical model designed to simulate and predict tumor volume regression (TVR) during radiation therapy (RT) using ear
- Sensitivity 77.7%
- Specificity 73.6%
APA
Barrett S, Zahid MU, et al. (2026). External Validation of the Proliferation Saturation Index Model in Predicting Tumor Volume Regression in Patients With Non-Small Cell Lung Cancer Undergoing Radiation Therapy.. International journal of radiation oncology, biology, physics. https://doi.org/10.1016/j.ijrobp.2026.02.213
MLA
Barrett S, et al.. "External Validation of the Proliferation Saturation Index Model in Predicting Tumor Volume Regression in Patients With Non-Small Cell Lung Cancer Undergoing Radiation Therapy.." International journal of radiation oncology, biology, physics, 2026.
PMID
41724248
Abstract
[PURPOSE] The Proliferation Saturation Index (PSI) model is a patient-specific mathematical model designed to simulate and predict tumor volume regression (TVR) during radiation therapy (RT) using early treatment response dynamics. This study validates the PSI model in an independent external cohort of patients with non-small cell lung cancer using previously derived model parameters.
[METHODS AND MATERIALS] In a cohort of 71 patients, treated with definitive RT alone (55 Gy/20#) tumor volume measurements were extracted from 6 cone beam computed tomography (CBCT) scans acquired on days 1, 2, and 3 and weekly thereafter. Model predictions of TVR at final CBCT (CBCT6) were made using tumor volume dynamics recorded from CBCT 1-3, CBCT 1-4, and CBCT 1-5. Agreement between predicted and actual volumes were assessed using the coefficient of determination (R) and Pearson correlation coefficient (PCC).
[RESULTS] Model predictions showed strong agreement with measured volumes, improving with additional input data: R increased from 0.81 (3 inputs) to 0.94 (5 inputs), and PCC rose from 0.90 to 0.97. Prediction using data from approximately day 10 of RT (CBCT4) yielded R = 0.91 and PCC = 0.95. The model demonstrated strong performance in identifying poor responders (TVR ≤10%) with sensitivity of 77.7%, specificity of 73.6%, and a negative predictive value of 90.7%. Sensitivity analysis confirmed model stability with under ±20% parameter variation.
[CONCLUSIONS] This external validation confirms the PSI model's reproducibility and robustness in predicting TVR in patients with non-small cell lung cancer treated with definitive RT, supporting its integration into personalized RT planning strategies.
[METHODS AND MATERIALS] In a cohort of 71 patients, treated with definitive RT alone (55 Gy/20#) tumor volume measurements were extracted from 6 cone beam computed tomography (CBCT) scans acquired on days 1, 2, and 3 and weekly thereafter. Model predictions of TVR at final CBCT (CBCT6) were made using tumor volume dynamics recorded from CBCT 1-3, CBCT 1-4, and CBCT 1-5. Agreement between predicted and actual volumes were assessed using the coefficient of determination (R) and Pearson correlation coefficient (PCC).
[RESULTS] Model predictions showed strong agreement with measured volumes, improving with additional input data: R increased from 0.81 (3 inputs) to 0.94 (5 inputs), and PCC rose from 0.90 to 0.97. Prediction using data from approximately day 10 of RT (CBCT4) yielded R = 0.91 and PCC = 0.95. The model demonstrated strong performance in identifying poor responders (TVR ≤10%) with sensitivity of 77.7%, specificity of 73.6%, and a negative predictive value of 90.7%. Sensitivity analysis confirmed model stability with under ±20% parameter variation.
[CONCLUSIONS] This external validation confirms the PSI model's reproducibility and robustness in predicting TVR in patients with non-small cell lung cancer treated with definitive RT, supporting its integration into personalized RT planning strategies.