SCHC

SCHC classifies pulmonary nodules as benign or malignant using platelet-derived features and machine learning to improve diagnostic discrimination.

Key Features:

• Algorithm: eXtreme Gradient Boosting (XGBoost) was used to construct the predictive model.
• Input data: Computed tomographic data, clinical information, and platelet-related metrics were incorporated.
• Platelet predictors: Included age, platelet counts in platelet-rich plasma (pPLT), plateletcrit in platelet-rich plasma (pPCT), plateletcrit in whole blood (bPCT), and nodule size.
• Study design and cohorts: Developed using prospective/observational real-world data with a development cohort of 419 participants and an external validation cohort of 62 participants.
• Evaluation metrics: Model performance was assessed with Receiver Operating Characteristic (ROC) curves, continuous net reclassification improvement (NRI), integrated discrimination improvement (IDI), and net benefit (NB).
• Comparative performance: Demonstrated superior discrimination and reclassification improvements versus VA, MC, and BU models in the development cohort and subgroups, with reduced robustness on external validation.

Scientific Applications:

• Diagnostic discrimination: Differentiating early-stage malignant from benign pulmonary nodules using integrated platelet features and imaging.
• Biomarker integration: Incorporating common laboratory platelet metrics (pPLT, pPCT, bPCT) into oncological diagnostic models.
• Clinical decision support: Informing pulmonary nodule management and risk stratification through model-derived probability estimates.

Methodology:

Model development employed XGBoost trained on computed tomographic data, clinical variables, and platelet metrics (age, pPLT, pPCT, bPCT, nodule size), with performance evaluated by ROC curves, continuous NRI, IDI, and NB and validated on an external cohort.