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...

Descripción completa

Detalles Bibliográficos
Autores principales: Hou, Rui, Lo, Joseph Y., Marks, Jeffrey R., Hwang, E. Shelley, Grimm, Lars J.
Formato: Online Artículo Texto
Lenguaje:English
Publicado: Cold Spring Harbor Laboratory 2023
Materias:
Article
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

Internet

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

Ejemplares similares