Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1979
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-20T06:54:21.131369Z
status: published
lens: trace
sector: health
headline: Predicting lymph node metastasis in non-small cell lung cancer using clinical and PET/CT-derived morphological features: An interpretable machine-learning study
dek: Background Accurate prediction of lymph node metastasis (LNM) in non-small cell lung cancer (NSCLC) is crucial for staging and treatment planning. This study aimed to establish a standardized machine learning workflow based on routinely available clinical, serological, and PET/CT-derived parameters to enhance preoperative LNM risk stratification. Methods This retrospective study included 255 patients with histologically confirmed NSCLC, all of whom underwent systematic lymph-node dissection. Patients were random…
gain_title: A machine-learning model using four routine clinical and imaging predictors identified lymph node metastasis in NSCLC with moderate discrimination and a high negative predictive value in a held-out test cohort.
problem_title: The same model had limited positive predictive value and did not significantly outperform a simple logistic regression reference, so it needs external multicenter validation before clinical use.
trace_subject: Machine-learning prediction of lymph node metastasis in non-small cell lung cancer
gain_reading: A machine-learning model using four routine clinical and imaging predictors identified lymph node metastasis in NSCLC with moderate discrimination and a high negative predictive value in a held-out test cohort.
gain_evidence: moderate discrimination and a high negative predictive value for pathological LNM in NSCLC
problem_reading: The same model had limited positive predictive value and did not significantly outperform a simple logistic regression reference, so it needs external multicenter validation before clinical use.
problem_evidence: its limited positive predictive value and the absence of significant differences among the candidate algorithms support cautious interpretation
quick_read: Researchers built a standardized machine-learning workflow to estimate the risk of lymph node metastasis before surgery in people with non-small cell lung cancer. They used routine clinical, blood marker and PET/CT-derived data from 255 patients who had systematic lymph-node dissection, and compared six algorithms.

Better preoperative risk stratification could support staging and treatment planning, but this retrospective, single-cohort result is early. The model's positive predictions were not reliable enough and a simple regression model performed comparably, so its clinical value is unproven until tested at other centers.
limitation: The study is retrospective with a small internal test cohort, the model showed limited positive predictive value, and the authors say external multicenter evaluation is required before clinical implementation.
tag: Dual reading
key_points: The study retrospectively analyzed 255 NSCLC patients, split into a training cohort of 178 and a held-out test cohort of 77. | Four predictors were retained: central tumor location, elevated carcinoembryonic antigen, elevated squamous cell carcinoma antigen, and solid nodule status. | None of the quantitative PET metabolic parameters remained independently associated with lymph node metastasis. | XGBoost had a cross-validated AUC of 0.735 and a test AUC of 0.747, versus 0.733 for logistic regression. | The logistic regression model had a Brier score of 0.152 versus 0.198 for XGBoost.
rundown: The team screened candidate predictors with univariable and multivariable logistic regression in the training cohort, then trained six machine-learning algorithms using repeated nested five-fold cross-validation. XGBoost had the highest cross-validated AUC and was named the primary model.

After development, all six models were applied unchanged to the test cohort and their AUCs compared with paired DeLong tests and Holm correction. The primary model was also assessed for calibration, clinical utility, and feature contribution using SHAP.

A conventional logistic regression using the same four predictors served as a reference. Its discrimination did not differ significantly from XGBoost and its Brier score was lower, indicating better calibration in this comparison.
sources:
- peer_reviewed | PLOS One | https://doi.org/10.1371/journal.pone.0361043 | 2026-10-15
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

4c343d9705692b1537f5d00af4c2cf5b849e1806697a6d8cd4c0c2b8070bbfc9

previous

0000000000000000000000000000000000000000000000000000000000000000