Table 4.
DL-based mammography for breast tumor detection.
| Reference | Year | Method | Database | Number of images | Accuracy | AUC | Sensitivity | Specificity |
|---|---|---|---|---|---|---|---|---|
| (88) | 2016 | Deep CNN | DDSM | 600 | 96.7% | NA | NA | NA |
| (89) | 2016 | AlexNet | FFDM | 607 | NA | 86% | NA | NA |
| (90) | 2016 | CNN | BCDR-F03 | 736 | NA | 82% | NA | NA |
| (91) | 2016 | SNN | UCI, DDSM | NA | 89.175%, 86% | NA | NA | NA |
| (92) | 2016 | ML-NN | ED(US) | NA | 98.98% | 98% | NA | NA |
| (93) | 2016 | DBN | ED(US-SW E) | NA | 93.4% | 94.7% | 88.6% | 97.1% |
| (94) | 2017 | Deep CNN | FFDM | 3185 | 82% | 88% | 81% | 72% |
| (95) | 2017 | CNN (COM) | INbreast | 115 | 95% | 91% | NA | NA |
| (96) | 2017 | Deep CNN | SFM, DM | 2242 | NA | 82% | NA | NA |
| (97) | 2017 | CNN-CT | IRMA | 2796 | 83.74% | 83.9% | 79.7% | 85.4% |
| (97) | 2017 | CNN-WT | IRMA | 2796 | 81.83% | 83.9% | 78.2% | 83.3% |
| (98) | 2017 | VGG19 | FFDM | 245 | NA | 86% | NA | NA |
| (99) | 2017 | Custom CNN | FFDM | 560 | NA | 79% | NA | NA |
| (100) | 2017 | VGG16 | IRMA | 2795 | 100% | 100% | NA | NA |
| (101) | 2017 | SNN | DDSM | 480 | 79.5% | NA | NA | NA |
| (101) | 2017 | CNN (COM) | MIAS, CBIS-INBreast | NA | 57% | 77% | NA | NA |
| (96) | 2017 | Multitask DNN | ED(Mg),DD SM | 1057 malignant, 1397 benign | 82% | NA | NA | NA |
| (102) | 2017 | CNN (COM) | ED (HP) | NA | 95.9% (2 classes), 96.4% (15 classes) | NA | NA | NA |
| (103) | 2017 | ImageNet | BreakHis | NA | 93.2% | NA | NA | NA |
| (104) | 2018 | GoogLeNet | BCDR-F03 | 736 | 81% | 88% | NA | NA |
| (104) | 2018 | AlexNet | BCDR-F03 | 736 | 83% | 79% | NA | NA |
| (104) | 2018 | Shallow CNN | BCDR-F03 | 736 | 73% | 82% | NA | NA |
| (105) | 2018 | Faster R-CNN | INbreast | 115 | NA | 95% | NA | NA |
| (105) | 2018 | Faster R-CNN | DREAM | 82,000 | NA | 85% | NA | NA |
| (106) | 2018 | ROI based CNN | DDSM | 600 | 97% | NA | NA | NA |
| (107) | 2018 | Inception V3 | DDSM | 5316 | 97.35% ( ± 0.80) | 98% | NA | NA |
| (107) | 2018 | Inception V3 | INbreast | 200 | 95.50% ( ± 2.00) | 97% | NA | NA |
| (107) | 2018 | Inception V3 | BCDR-F03 | 600 | 96.67% ( ± 0.85) | 96% | NA | NA |
| (107) | 2018 | VGG16 | DDSM | 5316 | 97.12% ( ± 0.30) | NA | NA | NA |
| (107) | 2018 | ResNet50 | DDSM | 5316 | 97.27% ( ± 0.34) | NA | NA | NA |
| (108) | 2018 | Deep CNN | MIAS | 120 | 96.7% | NA | NA | NA |
| (109) | 2018 | AlexNet, Transfer Learning | University of Pittsburgh | 20,000 | NA | 98.82% | NA | NA |
| (110) | 2018 | Faster R-CNN | DDSM, INbreast & Semmelweis University data | 2620,115, 847 | NA | 95% | NA | NA |
| (111) | 2018 | CNN | FFDM | 78 | NA | 81% | NA | NA |
| (112) | 2018 | MV-DNN | BCDR-F03 | 736 | 85.2% | 89.1% | NA | NA |
| (113) | 2018 | Deep CNN | MIAS | 322 | 65% | NA | NA | NA |
| (114) | 2018 | SDAE | ED (HP) | 58 | 98.27% (Benign), 90.54% (Malignant) | NA | 97.92% (Benign), 90.17% (Malignant) | NA |
| (115) | 2018 | CNN (UDM) | BreakHis | NA | 96.15%, 98.33% (2 Classes), 83.31-88.23% (8 Classes) | NA | NA | NA |
| (116) | 2018 | CNN-CH | BreakHis | 400× (× represents magnificati on factor) | 96% | NA | 97.79% | 90.16% |
| (116) | 2018 | CNN-CH | BreakHis | 400× (× represents magnificati on factor) | 97.19% | NA | 98.20% | 94.94% |
| (117) | 2019 | CNN | DDSM | 190 | 93.24% | NA | 91.92% | 91.92% |
| (117) | 2019 | CNN based LBP | DDSM | 190 | 96.32% | 97% | 96.81% | 95.83% |
| (118) | 2020 | InceptionV3 | DDSM | 2620 | 79.6% | NA | 89.1% | NA |
| (118) | 2020 | ResNet 50 | DDSM | 2620 | 85.7% | NA | 87.3% | NA |
| (119) | 2020 | ResNet50 | DDSM patch | 10713 | 75.1% | NA | NA | NA |
| (119) | 2020 | Mobile Net | DDSM patch | 10713 | 77.2% | NA | NA | NA |
| (119) | 2020 | MVGG16 | DDSM patch | 10713 | 80.8% | NA | NA | NA |
| (119) | 2020 | MVGG16 + ImageNet | DDSM patch | 10713 | 88.3% | 93.3% | NA | NA |
| (71) | 2020 | GAN and CNN | DDSM | 292 | 80% | 80% | NA | NA |
| (120) | 2021 | Optimal Multi-Level Thresholding-based Segmentation with DL enabled Capsule Network (OMLTS-DLCN) | Mini-MIAS dataset and DDSM dataset | NA | 98.5% for Mini-MIAS, 97.55% for DDSM | NA | NA | NA |
| (121) | 2021 | Inception-ResNet-V2 | BreastScreen Victoria dataset | 28,694 | 0.8178 | 0.8979 | NA | NA |
| (122) | 2021 | AI-powered imaging biomarker | 2,058 | NA | 0.852 | NA | NA | |
| (123) | 2022 | DualCoreNet | DDSM | NA | NA | 0.85 | NA | NA |
| (123) | 2022 | DualCoreNet | INbreast | NA | NA | 0.93 | NA | NA |
N/A, Not Applicate.