Comparison of performance of multiple standard convolutional neural network (CNN) models for binary patch classification on mammography.

Woo, MinJae; Zhang, Linglin; Brown-Mulry, Beatrice; Hwang, InChan; Gichoya, Judy Wawira; Gastounioti, Aimilia; Banerjee, Imon; Seyyed-Kalantari, Laleh; Trivedi, Hari

doi:10.1371/journal.pdig.0000811.s001

Comparison of performance of multiple standard convolutional neural network (CNN) models for binary patch classification on mammography.

journal contribution

posted on 2025-04-08, 17:29 authored by MinJae Woo, Linglin Zhang, Beatrice Brown-Mulry, InChan Hwang, Judy Wawira Gichoya, Aimilia Gastounioti, Imon Banerjee, Laleh Seyyed-Kalantari, Hari Trivedi

Comparison of performance of multiple standard convolutional neural network (CNN) models for binary patch classification on mammography.

History

Usage metrics

Keywords

rads density c help develop fair field digital mammograms conducted using univariate considering confounding factors classification model achieved false negative rate false negative predictions mammography may underperform deep learning model false positive rate multivariate regression models understand performance gaps patches matched based deep learning models certain patient subgroups 910 patches ),mammography deep learning negative patches based classification false positives making models specific subgroups patient demographics multivariate analysis create positive confounding effects 390 patches 050 ),011 ),xlink ">white patients ultimately enhancing subgroup evaluation subgroup analysis study underscores study evaluates significantly associated significant predictors resnet152v2 demonstrating never biopsied interpretable decision imaging features identify imaging findings suggest f1 score demographic characteristics best performance architectural distortion 931 patients

Licence

CC BY 4.0

Comparison of performance of multiple standard convolutional neural network (CNN) models for binary patch classification on mammography.

History

Usage metrics

Categories

Keywords

Licence

Exports