-
1
The study developed and internally evaluated an architecture-diverse deep-learning ensemble to classify CT images as adenocarcinoma, large cell carcinoma, squamous cell carcinoma, or normal lung.
-
2
The analysis assessed validation-based model selection, weighted ensembling, temperature scaling, class-specific decision scaling, calibration, class-specific errors, and predictive entropy.
-
3
The intended use was assistive classification and uncertainty-informed prioritization for expert review, not replacement of histopathology, molecular testing, or radiologist interpretation.
-
4
The publicly available dataset contained 1,000 images labeled in four categories, and the repository labels served as the operational reference standard.
-
5
The dataset's original diagnostic verification methods and patient-level independence between partitions could not be determined.
