Skip to main content

Table 3 Breast cancer prediction capacity of automated characterization of the parenchymal patterns

From: Beyond breast density: a review on the advancing role of parenchymal texture analysis in breast cancer risk assessment

Study Dataset Model Discriminatory capacity (AUC)
Year Participating Institutions A B m   $Texture $PD $Texture + PD ^Texture ^PD ^Texture + PD
Distinguishing between cancer cases and healthy women
 Wu et al. (2008) [76] University of Michigan 128C 549 No LDACV 0.73    
 Manduca et al. (2009) [66] Mayo Clinic, Moffitt 246P 522 Yes LRCV [0.58, 0.60] 0.58   Age, BMI
[0.61, 0.62] 0.60 [0.62, 0.63]
 Wei et al. (2011) [73] University of Michigan 136C 246 No LDA 0.74* 0.61 0.76 Age, BMI, family history of breast cancer, #of previous breast biopsies
    0.78
 Nielsen et al. (2011) [61] University of Copenhagen, Nordic Bioscience, Delft University of Technology, RadboudUMC, Mayo Clinic 245P 250 No LRCV 0.63 0.60 0.66*    
 Brandt et al. (2011) [74] University of Copenhagen, RadboudUMC, Synarc Imaging Technologies 245P 245 Yes kNNCV 0.63 0.56     
 Häberle et al. (2012) [56] Erlangen University Hospital, Fraunhofer Institute for Integrated Circuits IIS, IMPRS, UCLA 864C 418 Yes LRCV 0.75 0.51 0.75 Age, BMI, family history of breast cancer, parity, age at first term pregnancy
0.79 0.66 0.79
 Li et al. (2012) [84] University of Chicago 75C 328 No BANNCV 0.73      
 Li et al. (2012) [84] University of Chicago 67C 268 Yes BANNCV 0.70      
 Chen et al. (2014) [75] University of Manchester 50C 50 No LR 0.71 0.62 0.68    
 Nielsen et al. (2014) [64] UCPH, Nordic Bioscience, Biomediq, RadboudUMC, Mayo Clinic 245P 250 No LRCV     Age, BMI, menopause, hormonal use
0.60 0.63 0.66
 Nielsen et al. (2014) [64] UCPH, Nordic Bioscience, Biomediq, RadboudUMC, Mayo Clinic 226C 442 Yes LRCV     Age, BMI, menopause, hormonal use
0.61 0.63  
 Li et al. (2014) [71] University of Chicago 67C 268 Yes BANNCV 0.70* 0.57 0.68    
 Karemore et al. (2014) [90] UCPH, RadboudUMC 245P 250 Yes kNNCV 0.59      
 Zheng et al. (2015) [51] University of Pennsylvania 106C 318 Yes LRCV 0.85* 0.59 0.86    
 Sun et al. (2015) [53] University of Texas, China Northeastern University, University of Oklahoma, TTUHS, Guiyang Medical University 141P 199 No SVMCV 0.73    Age, BMI, family history of breast cancer, hormonal use, age at first term pregnancy
0.77   
 Tan et al. (2015) [77] University of Texas, University of Oklahoma, University of Pittsburgh     ANNCV     age   
812P 1084 No 0.71    0.78   
 Tan et al. (2015) [78] University of Oklahoma, University of Pittsburgh 430P 440 No ANNCV [0.64, 0.73]      
Predicting the risk of carrying a high-risk genetic mutation
 Huo et al. (2000) [80] University of Chicago 15 143 No LDA [0.59, 0.82]      
 Huo et al. (2000) [80] University of Chicago 15 30 Yes LDA [0.53, 0.87]      
 Huo et al. (2002) [55] University of Chicago, University of Pennsylvania 30 142 No LDA 0.91      
 Huo et al. (2002) [55] University of Chicago, University of Pennsylvania 30 60 Yes LDA 0.92      
 Li et al. (2004) [54] University of Chicago, University of Pennsylvania 30 60 Yes LDACV [0.69, 0.92]      
 Li et al. (2005) [81] University of Chicago 30 142 No ROCA [0.66, 0.86]      
 Li et al. (2005) [81] University of Chicago 30 60 Yes ROCA [0.67, 0.86]      
 Li et al. (2007) [82] University of Chicago 30 142 No ROCACV [0.74, 0.93]      
 Li et al. (2007) [82] University of Chicago 30 60 Yes ROCACV [0.77, 0.91]      
 Li et al. (2008) [83] University of Chicago 30 142 No ROCA 0.90      
 Li et al. (2008) [83] University of Chicago 30 60 Yes ROCA 0.89      
 Li et al. (2012) [84] University of Chicago 53 328 No BANNCV 0.82      
 Li et al. (2012) [84] University of Chicago 34 136 Yes BANNCV 0.81      
 Li et al. (2014) [71] University of Chicago 34 136 Yes BANNCV 0.81* 0.53 0.81    
 Gierach et al. (2014) [85] University of Chicago 137 100 No BANNCV 0.68 0.59 0.72    
 Gierach et al. (2014) [85] University of Chicago, NCI-NIH, Washington Radiology Associates, Genentech, USUHS, UCL, WRNMC, Westat Inc. 126 89 Yes BANNCV 0.71 0.55 0.72    
  1. Area under the ROC curve (AUC) achieved by risk assessment models when fed with mammographic texture and/or density measures
  2. IMPRS International Max Planck Research School for Optics and Imaging, Moffitt Moffitt Cancer Center and Research Institute, NCI-NIH National Cancer Institute, National Institutes of Health, RadboudUMC Radboud University Nijmegen Medical Centre, TTUHS Texas Tech University Health Sciences, UCL University College London, UCLA University of California at Los Angeles, USUHS Uniformed Services University of the Health Sciences, WRNMC Walter Reed National Military Medical Center
  3. Dataset: A cancer cases (Pprior unaffected images, Cimages from the contralateral, unaffected, breast at the time of cancer diagnosis) or other high-risk population (i.e., BRCA1/2 carriers), B controls, m matched subgroups; Model: LDA Linear Discriminant Analysis, LR Logistic Regression, kNN k-nearest neighbors, BANN Bayesian Artificial Neural Network, ANN Artificial Neural Network, ROCA Receiver Operating Characteristic Analysis. PD percent density. CVcross-validated models; $unadjusted models; ^models adjusted for established risk factors; *statistically significant from $PD at < 0.05