Skip to main content

Table 6 Logistic regression model for predicting masking with predicted PMD based on boosting

From: Using automated texture features to determine the probability for masking of a tumor on mammography, but not ultrasound

Variable Coefficient (standard error)
Baseline –0.906 (1.308)
Age (year) –0.018 (0.014)
BMI (kg/m2) –0.080 (0.033)
Previous breast surgery  
 Noa 0
 Yes 0.502 (0.286)
Menopausal and HRT status  
 Premenopausala 0
 Postmenopausal and no HRT –0.530 (0.357)
 Postmenopausal and HRT 0.208 (0.355)
Imaging technique  
 Analoga 0
 Digital 0.416 (0.223)
Predicted PMD 0.032 (0.009)
  1. The model is fitted on the complete dataset. To estimate a patient’s risk for masking, the following steps are necessary: texture features values are calculated from the mammogram, the boosting regression model is applied to obtain the predicted PMD, and patient characteristics and predicted PMD are linearly combined with the logistic regression coefficient to obtain interim value z. Finally, exp (z)/(1 + exp (z)) is the predicted risk for masking
  2. aReference category