Table 3.
Summary of key studies on the role of AI in breast MRI.
| n | Task | Algorithms | No. of Cases | Results | Ref. |
|---|---|---|---|---|---|
| 1 | detect, characterize and categorize lesions | a supervised-attention model with deep learning | 335 | AUC=81.6% | (60) |
| 2 | classify lesions | radiomic analysis and CNN | 1294 | AUC=98% | (62) |
| 3 | characterize and classify lesions | the combination of unsupervised dimensionality reduction and embedded space clustering followed by a supervised classifier | 792 | AUC=81% | (63) |
| 4 | classify breast tumors | QuantX | 111 | AUC=76% | (67) |
| 5 | assess and diagnose contralateral BI-RADS 4 lesions | MRI radiomics-based machine learning | 178 | AUC=77% | (69) |
| ACC=74.1% | |||||
| 6 | assess tumor extent and multifocality | CADstream software (version 5.2.8.591) | 86 | AUC = 88.8% | (70) |
| Spe=92.1% | |||||
| PPV=90.0% | |||||
| 7 | early predict pathological complete response to neoadjuvant chemotherapy and survival outcomes | linear support vector machine, linear discriminant analysis, logistic regression, random forests, stochastic gradient descent, decision tree, adaptive boosting and extreme gradient boosting | 38 | AUC=86% | (71) |
AI, artificial intelligence; MRI, magnetic resonance imaging; AUC, the area under the receiver operating characteristic curve; CNN, convolutional neural network; BI-RADS, Breast Imaging Reporting and Data System; ACC, accuracy; CAD, computer-aided detection; Spe, specificity; PPV, positive predictive value.