Cargando…
Classification performance bias between training and test sets in a limited mammography dataset
OBJECTIVES: To assess the performance bias caused by sampling data into training and test sets in a mammography radiomics study. METHODS: Mammograms from 700 women were used to study upstaging of ductal carcinoma in situ. The dataset was repeatedly shuffled and split into training (n=400) and test c...
| Autores principales: | , , , , |
|---|---|
| Formato: | Online Artículo Texto |
| Lenguaje: | English |
| Publicado: |
Cold Spring Harbor Laboratory
2023
|
| Materias: | |
| Acceso en línea: | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9980247/ https://www.ncbi.nlm.nih.gov/pubmed/36865183 http://dx.doi.org/10.1101/2023.02.15.23285985 |