Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-0622
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-08-02T06:07:07.748025Z
status: published
lens: g_space
sector: health
headline: Local Recurrence Prediction After Carbon-ion Radiotherapy for Early-stage Non-small Cell Lung Cancer Using Machine Learning
dek: Background/aim Predicting local recurrence remains challenging in carbon-ion radiotherapy (CIRT) for non-small cell lung cancer (NSCLC). In this study, we aimed to develop and validate a machine learning model to predict local recurrence after CIRT for early-stage peripheral NSCLC. Patients and methods We retrospectively analyzed patients treated with CIRT at our institution between 2010 and 2020. An Extreme Gradient Boosting classifier using clinical parameters was developed to predict local recurrence within 2…
gain_title: Clinical-parameter XGBoost model stratified patients into low-risk and high-risk groups with 94.0% vs 65.0% 2-year local control after carbon-ion radiotherapy for early-stage peripheral NSCLC.
problem_title: (none)
trace_subject: (none)
gain_reading: Clinical-parameter XGBoost model stratified patients into low-risk and high-risk groups with 94.0% vs 65.0% 2-year local control after carbon-ion radiotherapy for early-stage peripheral NSCLC.
gain_evidence: the 2-year local control rates were 94.0% in the low-risk group and 65.0% in the high-risk group | A machine learning model based on clinical parameters may predict 2-year local recurrence after CIRT for early-stage peripheral NSCLC
problem_reading: (none)
problem_evidence: (none)
quick_read: Between 2010 and 2020, 124 patients with early-stage peripheral non-small cell lung cancer treated with carbon-ion radiotherapy at a single institution were analyzed retrospectively to develop a machine learning predictor of local recurrence within 24 months. An Extreme Gradient Boosting classifier trained on clinical parameters with nested threefold cross-validation achieved ROC-AUC 0.622 and PR-AUC 0.145, and separated patients into low-risk and high-risk groups.

The ability to flag a high-risk subgroup with 65.0% versus 94.0% 2-year local control could help prioritize closer surveillance or adjuvant strategies after CIRT, but the modest discrimination, low precision-recall performance, and small event count of 10 recurrences (8.1%) leave uncertainty about reliability and generalizability beyond this institution.
limitation: Retrospective single-institution study with only 124 patients and 10 recurrence events, and modest discriminative performance limits generalizability.
tag: Evidence-backed gain
key_points: Retrospective analysis of 124 patients treated with CIRT between 2010 and 2020, with 10 (8.1%) experiencing local recurrence within two years. | Extreme Gradient Boosting classifier developed using nested threefold cross-validation optimizing ROC-AUC to predict recurrence within 24 months. | Model achieved ROC-AUC of 0.622 and PR-AUC of 0.145, with risk stratification showing 94.0% vs 65.0% 2-year local control. | Interpretability explored using SHapley Additive exPlanations (SHAP) and evaluation included survival analysis with log-rank test.
rundown: Researchers retrospectively analyzed 124 patients with early-stage peripheral NSCLC treated with carbon-ion radiotherapy between 2010 and 2020, with median follow-up of 44.9 months and 2-year local control rate of 91.0%. An Extreme Gradient Boosting classifier using clinical parameters was trained with nested threefold cross-validation optimizing ROC-AUC to predict recurrence within 24 months.

The model yielded ROC-AUC 0.622 and PR-AUC 0.145, but stratified patients into groups with 94.0% versus 65.0% 2-year local control. Authors used precision-recall evaluation, survival analysis, and SHAP for interpretability, concluding the clinical-parameter model may predict recurrence.
sources:
- peer_reviewed | Anticancer Research | https://doi.org/10.21873/anticanres.18299 | 2026-08-01
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

82418ef08b2d18ce128075939e8a1261ae004440469448998dc67f0de9311cf9

previous

0000000000000000000000000000000000000000000000000000000000000000