Table 3.
Deep learning approaches in mammography for breast lesion segmentation.
| Year | Model | Evaluation Dataset | Noise remove method | Performance Metrics (results) |
|---|---|---|---|---|
| 2015 (59) | CRF | INbreast and DDSM-BCRP | NA | The method achieved an 89.0% Dice index in 0.1. |
| 2018 (60) | adversarial deep structured net | INbreast and DDSM-BCRP | NA | The method achieved a segmentation rate of 97.0%. |
| 2018 (61) | deep learning using You-Only-Look-Once | INbreast | NA | The method achieved detection rate of 98.96%, Matthews correlation coefficient (MCC) of 97.62%, and F1 score of 99.24%. |
| 2018 (62) | CRU-Net | INbreast and DDSM-BCRP | NA | The CRU-Net achieved a Dice Index DI of 93.66% for INbreast and a DI of 93.32% for DDSM-BCRP. |
| 2019 (63) | MS-ResCU-Net and ResCU-Net | INbreast | NA | The MS-ResCU-Net achieved an accuracy of 94.16%, sensitivity of 93.11%, specificity of 95.02%, DI of 91.78%, Jac of 85.13%, and MCC of 87.22%, while ResCU-Net correspondingly achieved 92.91%, 91.51%, 94.64%, 90.50%, 83.02%, and 84.99%. |
| 2019 (64) | dense U-Net with AGs | DDSM | NA | The method achieved 82.24% F1 score, 77.89% sensitivity, and overall accuracy of 78.38%. |
| 2019 (65) | RU-Net | DDSM, BCDR-01, and INbreast | cLare filter | The proposed model achieved a mean test pixel accuracy of 98.00%, a mean Dice coefficient index (DI) of 98.00%, and mean IOU of 94.00%. |
| 2019 (66) | modified U-Net | DDSM | Laplacian filter | The method produced 98.50% of the F-measure and a 97.80% Dice score, Jaccard index of 97.40%, and average accuracy of 98.20%. |
| 2020 (67) | mammographic CAD based on pseudocolour mammograms and mask RCNN | INbreast | morphological filters | The DSI achieved for mass segmentation was 0.88Â ± 0.10, and GMs and mask RCNN yielded an average TPR of 0.90Â ± 0.05. |
| 2020 (68) | FrCN | INbreast | NA | FrCN achieved an overall accuracy of 92.97%, 85.93% for MCC, 92.69% for Dice, and 86.37% for the Jaccard similarity coefficient. |
| 2020 (69) | U-Net | CBIS-DDSM, INbreast, UCHCDM, and BCDR-01 | adaptive median filter | The U-Net model achieved a mean Dice coefficient index of 95.10% and a mean IOU of 90.90%. |
| 2020 (70) | cGAN | INbreast | median filter | The cGAN achieved an accuracy of 98.0%, Dice coefficient of 88.0%, and Jaccard index of 78.0%. |
| 2020 (71) | cGAN | DDSM and INbreast | Morphological operations | The proposed cGAN model achieved a Dice coefficient of 94.0% and an intersection over union (IoU) of 87.0% |
| 2020 (72) | mask RCNN and DeepLab | MIAS and DDSM | Savitzky Golay filter | The mask RCNN achieved an AUC of 98.00%, DeepLab achieved an AUC of 95.00%. |
| 2020 (73) | AUNet | CBIS-DDSM and INbreast | NA | produced an average Dice similarity coefficient of 81.80% for CBIS-DDSM and 79.10% for INbreast |
| 2020 (74) | mask RCNN-FPN | training on DDSM and testing on the INbreast database | NA | The model achieved a mean average precision of 84.0% for multidetection and 91.0% segmentation accuracy. |
| 2020 (75) | U-Net | DDSM | NA | The model achieved a sensitivity of 92.32%, specificity of 80.47%, accuracy of 85.95%, Dice coefficient index of 79.39%, and AUC of 86.40%. |
| 2021 (76) | modified CNN based on U-Net model | DDSM-400 and CBIS-DDSM | NA | The method achieved a diagnostic performance of 89.8% and AUC of 86.20% based on ground-truth segmentation maps and a maximum of 88.0% and 86.0% for U-Net-based segmentation for DDSM-400 and CBIS-DDSM, respectively. |
| 2021 (77) | deeply supervised U-Net | DDSM and INbreast | cLare filter | The method achieved 82.70% of Dice, 85.70% of Jaccard coefficient, 99.70% accuracy, 83.10% sensitivity, and 99.80% specificity. |
| 2021 (78) | modified U-Net | MIAS, DDSM, and CBIS-DDSM | NA | The method achieved accuracy of 98.87%, AUC of 98.88%, sensitivity of 98.98%, precision of 98.79%, and F1 score of 97.99% on the DDSM datasets |
| 2023 (79) | Tubule-U-Net | 30820 polygonal annotated tubules in 8225 patches | NA | achieved 95.33%, 93.74%, and 90.02%, dice, sensitivity, and specificity scores, respectively |
N/A, Not Applicate.