Table 1 Deep learning in breast cancer diagnosis and prognosis
Aspect | Study | Details | Performance | Ref |
|---|---|---|---|---|
Diagnosis | Â | Â | Â | Â |
Tumor Identification and Classification | Cruz-Roa et al. | CNN to classify WSI for invasive ductal carcinoma | F1 score: 76% | [39] |
Tumor Identification and Classification | Han et al. | Classify 8 classes of breast tumors using BreaKHis dataset | Accuracy: 93.2% | [40] |
Diagnosis of Lymph Node Metastasis | Bejnori et al. | DL algorithms vs. pathologists | AUC: 0.99 vs. 0.88 | [50] |
Histologic Grading | Veta et al. | Mitosis detection challenge | F1 score: 0.61 | [52] |
Histologic Grading | Tellez et al. | PHH3 stains with CNN annotations | Improved consistency of pathologists | [53] |
Breast Cancer Lesion and Cell Nucleus Segmentation | Veta et al. | Suppressed non-CNN algorithms | Efficient segmentation | [54] |
Tumor Identification and Classification | Rexhepaj et al. | Quantified ER- and PR-expressed cells | Correlation: 0.9 | [55] |
Tumor Identification and Classification | Couture et al. | Feature-based DL model on H&E tissue microarray images | Accuracy: 84% | [56] |
Tumor Identification and Classification | Shamai et al. | Predict 19 biomarker statuses including ER and PR | Accuracy: 92% | [57] |
TNM Staging | Chen et al. | Detect metastatic cancer cells from lymph node images | AUC: 0.99 | [58] |
Prognosis | Â | Â | Â | Â |
Recurrence Risk Prediction | Whitney et al. | ER-positive breast tumors; nuclear shape and texture features | Accuracy: 0.85 | [65] |
Prognostic Value of TILs | Makhlouf et al. | High sTIL associated with shorter survival | HR: 1.6 (discovery), 2.5 (validation) | [67] |
Response to Neoadjuvant Chemotherapy | Choi et al. | High sTIL tumors associated with better response | Odds ratio: 1.28 | [68] |
Histological Grading | Wang et al. | DeepGrade model for NHG 2 patients | HR: 2.94 | [69] |
Lymph Node Metastasis | Verghese et al. | smuLymphNet model for axillary lymph node analysis | HR: 0.28 | [71] |
Lymph Node Metastasis | Zheng et al. | Ultrasound imaging for ALN status prediction | AUC: 0.902 | [72] |
Recurrence Risk Prediction | Klimov et al. | WSI and clinical data for recurrence prediction | Accuracy: 87% | [73] |
HRD Prediction | Lazard et al. | DL method for HRD prediction using H&E slides | AUC: 0.86 | [74] |
Multi-Omics Integration | Yu et al. | DCE-MRI data for ALN metastasis prediction | Accuracy: 0.89 | [75] |