Table 1: Summary of ANN studies in mammography interpretation and diagnostic decision making.

StudyTypeANN structureInputDataset and training/testing strategyResults and findings

Stafford et al. (1993) [33]CADeA committee of four three-layer BP-ANNsPixel information167 mammograms with pathologies and 89 without pathologies.
50% for training and 50% for testing.
Test on 20 out of 128 mammograms covering microcalcification size-range of
50250 μm:
0.9% FP at 85% TP,
2.4% FP at 100% TP.
50–2,000 μm:
25% FP at 84% TP,
40% FP at 100% TP.

Zhang et al. (1994) [30]CADeThe Shift-Invariant ANN (SI-ANN)Pixel information168 ROIs from 34 digitized mammograms.
50%-50% cross-validation.
ROC index: ,
45% FP at 100% TP.

Chan et al. (1995) [35]CADeThe Convolution Neural Network (CNN) Pixel information52 mammograms
Group 1:
110 TP and 116 FP.
Group 2:
108 TP and 116 FP.
Two-fold cross-validation.
ROC index: .  
FP rate:
0.1 cluster per image at 100% TP (for obvious cases),
1.5 cluster per image at 90% TP (for average and subtle cases).

Nagel et al. (1998) [36]CADeSI-ANNFeatures extracted from image196 TPs  and  1,252 FPs.
Leave-one-out cross-validation.
The number of FPs per image at 83% TP:
1.6 for ANN,
0.8 for ANN and rule-based method.
Average ROC index:
(stdev = 0.04) for ANN,
(stdev = 0.07) for ANN + rule-based method ( ).

Wu et al. (1992) [34]CADeBP-ANNPixel information56 positive, 56 negative, and 56 FP ROIs, respectively.
50%-50% cross-validation.
For individual microcalcifications:
.
For clustered microcalcifications:
; 50% FP at 95% TP.

Jiang et al. (1996) [45]CADxBP-ANNComputer-extracted morphological features40 malignant and 67 benign cases from 100 images.
Leave-one-out cross-validation.
Identified 100% malignant and 82% of the benign cases.
Significantly better than radiologists without computer aid ( ).

Jiang et al. (1999) [46]CADxBP-ANNComputer-extracted morphological features46 malignant and 58 benign cases.
Leave-one-out cross-validation.
By 10 radiologists:
, sensitivity = 73.5%, specificity = 31.6%.
With aid of ANN:
( ), sensitivity = 87.4%, specificity = 41.9%.

Huo et al. (1998) [47]CADxBP-ANNMorphological features characterizing margin and density38 benign and 57 malignant cases from 65 patients.
Leave-one-out cross-validation.
ANN:
,
19.2% specificity at 100% sensitivity.
Hybrid method (rule-based + ANN):
,
69.2% specificity at 100% sensitivity.

Kallergi (2004) [22]CADxBP-ANNMorphological and distributional descriptors50 benign and 50 malignant cases.
Leave-one-out cross-validation.
,
85% specificity at 100% sensitivity.

Chan et al. (1997) [48]CADxBP-ANNTexture features SGLD matrices41 malignant and 45 benign cases from 54 patients.
Leave-one-out cross-validation.
With best subset of features:
Mammogram-by-mammogram:
,
24% specificity at 100% sensitivity.
Patient-by-patient:
39% specificity at 100% sensitivity.

Baker et al. (1995) [41]CADxBP-ANNBI-RADS lesion descriptors and medical history variables133 benign and 73 malignant cases.
Leave-one-out cross-validation.
PPV:
61% (ANN) versus 35% (radiologists).
ROC index: (ANN) versus (radiologists), .  
Specificity:
62% (ANN) versus 30% (radiologists) at 100% sensitivity ( ).

Lo et al. (1999) [42]CADxBP-ANNBI-RADS lesion descriptors, age, and history variables326 benign and 174 malignant cases.
Leave-one-out cross-validation.
Only BI-RADS features:
, 6% specificity at 100% sensitivity.
BI-RADS + age: , 30% specificity at 100% sensitivity.
All features: , 22% specificity at 100% sensitivity.
Age variable significantly improves the ( ).

Ayer et al. (2010) [43]CADxBP-ANNDemographic, mammographic features, and BI-RADS categories510 malignant and 61,709 benign cases.
10-fold cross-validation.
(ANN) versus 0.939 (radiologists), .
Specificity at 85% sensitivity: 94.5% (ANN) versus 88.2% (radiologists), .

Jesneck et al. (2007) [49]CADxBP-ANNMammographic features, sonographic features, and history features296 malignant and 507 benign cases.
500 for training and validation, 303 for testing.
Training and validation set:
,
Testing set:
.

Tourassi et al. (2003) [51]CADxCSNNBI-RADS features, age and historyTraining set: 174 malignant and 326 benign cases.
Testing set: 358 malignant and 672 benign cases.
On training set:   
On testing set:   
CSNN is also capable to impute missing data.

Orr (2001) [54]CADx and risk estimationBP-ANNAge and radiographic features185 malignant and 1,103 benign cases.
490 for training and the rest for testing.
(surgeons) versus 0.86 (ANN), .
ANN is possible for risk stratification.

CADe: computer-aided detection, CADx: computer-aided diagnosis, ANN: artificial neural network, BP-ANN: back-propagation artificial neural network, FP: false positive, TP: true positive, ROI: region of interest, SGLD: spatial grey level dependence, PPV: positive prediction value, BI-RADS: the breast imaging reporting and data system, CSNN: constraint satisfaction neural network, and SI-ANN: shift-invariant artificial neural network.