Table 3.
Summary of Key Studies on Imaging Characterization of Breast Lesions, Including Detection, Diagnosis, Biologic Characterization, and Predicting Prognosis and Treatment Response
| REFERENCE | APPLICATION | NO. OF CASES | IMAGING MODALITY | MACHINE LEARNING ALGORITHM (IF APPLICABLE) | IMAGING/RADIOMIC FEATURE TYPE | RESULTS | |
|---|---|---|---|---|---|---|---|
| Detection | |||||||
| Zhang 1994135 | Microcalcification detection | 34 | Mammography | Convolutional neural networks | Deep learning characterization followed by conventional image analysis | AUC, 0.91 | |
| Karssemeijer 2006136 | Mass lesions | 500 | Mammography | Engineered algorithms | Engineered algorithms | Performance similar to radiology | |
| Reiner 2006137 | Mass lesions | 21 | Breast tomosynthesis | Engineered algorithms | Engineered algorithms | Sn, 90% | |
| Sahiner 2012138 | Microcalcifications | 72 | Breast tomosynthesis | Engineered algorithms | Engineered algorithms | Sn, 90% | |
| Diagnosis | |||||||
| Gilhuijs 1998139 | Mass lesions | 27 | DCE‐MRI | Engineered algorithms | Size, shape, kinetics | AUC, 0.96 | |
| Jiang 1999140 | Microcalcifications | 104 | Mammography | Engineered algorithms | Size and shape of individual microcalcifications and clusters | AUC, 0.75 | |
| Chen 2007141 | Mass lesions | 121 | DCE‐MRI | Engineered algorithms | Uptake heterogeneity in cancer tumors via 3D texture analysis | 3D better compared with 2D analysis | |
| Booshan 2010142 | Differentiate benign vs DCIS vs IDC | 353 | DCE‐MRI | Bayesian neural networks | Size, shape, margin morphology, texture (uptake heterogeneity), kinetics, variance kinetics | AUC, 0.79‐0.85 | |
| Jamieson 2010143 | Mass lesions | 1126 | Multimodality: Mammography, breast ultrasound, and breast DCE‐MRI | t‐SNE followed by Bayesian neural networks | Multiradiomic features in nonsupervised data mining | AUC, 0.88 | |
| Nielsen 2011144 | Breast cancer risk | 495 | Mammography | — | Texture analysis | AUC, 0.57‐0.66 | |
| Huynh 2016145 | Mass lesions | 219 | Mammography | Deep learning | Feature extracted from transfer learning from pretrained CNN | AUC, 0.81 | |
| Andropova 2017146 | Mass lesions | 1125 | Multimodality: Mammography, breast ultrasound, and breast DCE‐MRI | Deep learning | Fusion of human‐engineered computer features and those feature extracted from transfer learning from pretrained CNN | AUC: DCE‐MRI, 0.89; FFDM, 0.86; ultrasound, 0.90 | |
| Biologic characterization | |||||||
| Gierach 2014147 | BRCA1/2 mutation status | 237 | Mammography | Bayesian artificial neural network | Texture analysis | AUC, 0.68‐0.72 | |
| Li 2016148 | Molecular subtype classification | 91 (from TCGA) | DCE‐MRI | Engineered features, linear discriminant analysis | Multiradiomic tumor signature, including size, shape, margin morphology, texture (uptake heterogeneity), kinetics, variance kinetics | AUC, 0.65‐0.89 | |
| Li 2017149 | BRCA1/2 mutation status | 456 | Mammography | CNNs, computerized radiographic texture analysis, SVM | Texture analysis and deep learning | AUC, 0.73‐0.86 | |
| Predicting treatment response and prognosis | |||||||
| Drukker 2018150 | Prediction of recurrence‐free survival | 284 (from ACRIN 6657) | DCE‐MRI | — | .Most‐enhancing tumor volume | HR, 2.28‐4.81 | |
Abbreviations: 2D, 2‐dimensional; 3D, 3‐dimensional; ACC, accuracy; ACRIN, American College of Radiology Imaging Network; AUC, area under the curve; CNN, convolutional neural networks; DCE‐MRI, dynamic contrast‐enhanced magnetic resonance imaging; DCIS, ductal carcinoma in situ; FFDM, full‐field digital mammography; HR, hazard ratio; IDC, invasive ductal carcinoma; Sn, sensitivity; Sp, specificity; SVM, support vector machine; TCGA, The Cancer Genome Atlas; t‐SNE, t‐distributed stochastic neighbor embedding.