A Calibrated Deep Learning Framework Integrating Spatial Annotations and Clinical Metadata for Safe Three-Class Bone Lesion Classification on Radiographs

Bonferroni and Holm corrections across the family of eight pooled tests (four metrics × two contrasts) yielded adjusted p = 0.0002 for every real-versus-permuted comparison, so the causal effect of the metadata vector remains highly significant after multiplicity adjustment ( Supplementary Table S5 ). 3.13.

also in 132,142 other papers

Clinical metadata contributed a marginal +0.14 pp improvement in balanced accuracy (paired t -test: t = 0.26, p = 0.810; Supplementary Table S1 ), which did not reach statistical significance, an expected result given the limited statistical power of n = 5 paired observations for such a small effect size (Cohen’s d = 0.13).

also in 111,027 other papers