Objective:To investigate the difference of MRI and clinicopathological characteristics between patients with breast cancer susceptibility gene (BRCA)-positive and BRCA-negative breast cancer.Methods:The study was a cross-sectional study. MRI images and clinicopathological data of breast cancer patients with postoperative pathologically confirmed and determined BRCA gene status in Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College from January 2012 to April 2021 were retrospectively analyzed. A total of 120 BRCA-positive patients with 122 lesions and 120 BRCA-negative patients with 126 lesions were enrolled. All patients underwent MRI examinations, including pre-contrast and post-contrast scans. The breast MRI features of the patients were analyzed, including internal necrosis, lesion depth, degree of background parenchymal enhancement (BPE), type of lesion, shape and margin of the mass lesion, internal enhancement, and edema. Clinicopathological analysis included family history, molecular subtype, nuclear grade, and the status of human epidermal growth factor 2 (HER-2), estrogen receptor (ER), and progesterone receptor (PR). The χ2 test and Fisher exact test were performed to determine the differences in MRI features and clinicopathological manifestations of BRCA-positive and BRCA-negative breast cancers. Significant features obtained by univariate analysis were included in the multivariate logistic regression analyses to select independent influencing factors for predicting BRCA mutation in breast cancer. Results:Statistically significant differences were observed between patients with BRCA-positive and BRCA-negative breast cancers in the following features: family history, molecular subtype, ER or PR-positive status, HER-2-negative status, nuclear grade, BPE, internal necrosis, edema, and the shape and margin of the mass lesion( P?0.05). The multivariate logistic regression analyses showed that HER-2 negativity ( OR=3.277, 95% CI 1.087-9.875, P=0.035), round or oval shape ( OR=2.688,95% CI 1.143-6.320, P=0.023), circumscribed margin ( OR=3.001,95% CI 1.374-6.554, P=0.006), edema ( OR=4.407,95% CI 2.100-9.244, P<0.001), and the minimal or mild degree of BPE ( OR=2.520,95% CI 1.328-4.782, P=0.005) were significant independent factors in predicting BRCA gene mutations. Conclusions:There are differences in MRI features and clinicopathological manifestations of BRCA-positive and BRCA-negative breast cancers. HER-2 status, the shape and margin of lesions, edema and the degree of BPE could independently predict the BRCA status of breast cancer.

Precision management of breast cancer is essential for improving patient outcomes.Radiomics enables high-throughput extraction of quantitative imaging features that reflect tumor biological behavior and heterogeneity from medical imaging data.With the rapid advancement of artificial intelligence(AI),particularly deep learning,feature extraction capability and model performance in radiomics have been significantly enhanced.Based on multimodal imaging modalities such as MRI,ultrasound and mammography,AI-and radiomics-based models can address key clinical challenges including tumor diagnosis,staging,molecular subtype prediction and therapeutic response assessment,thereby facilitating precision breast cancer care.This review summarizes recent advances in AI-driven radiomics for breast cancer diagnosis and treatment and discusses future application prospects and challenges toward clinical translation.

Objective:To investigate the difference of MRI and clinicopathological characteristics between patients with breast cancer susceptibility gene (BRCA)-positive and BRCA-negative breast cancer.Methods:The study was a cross-sectional study. MRI images and clinicopathological data of breast cancer patients with postoperative pathologically confirmed and determined BRCA gene status in Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College from January 2012 to April 2021 were retrospectively analyzed. A total of 120 BRCA-positive patients with 122 lesions and 120 BRCA-negative patients with 126 lesions were enrolled. All patients underwent MRI examinations, including pre-contrast and post-contrast scans. The breast MRI features of the patients were analyzed, including internal necrosis, lesion depth, degree of background parenchymal enhancement (BPE), type of lesion, shape and margin of the mass lesion, internal enhancement, and edema. Clinicopathological analysis included family history, molecular subtype, nuclear grade, and the status of human epidermal growth factor 2 (HER-2), estrogen receptor (ER), and progesterone receptor (PR). The χ2 test and Fisher exact test were performed to determine the differences in MRI features and clinicopathological manifestations of BRCA-positive and BRCA-negative breast cancers. Significant features obtained by univariate analysis were included in the multivariate logistic regression analyses to select independent influencing factors for predicting BRCA mutation in breast cancer. Results:Statistically significant differences were observed between patients with BRCA-positive and BRCA-negative breast cancers in the following features: family history, molecular subtype, ER or PR-positive status, HER-2-negative status, nuclear grade, BPE, internal necrosis, edema, and the shape and margin of the mass lesion( P?0.05). The multivariate logistic regression analyses showed that HER-2 negativity ( OR=3.277, 95% CI 1.087-9.875, P=0.035), round or oval shape ( OR=2.688,95% CI 1.143-6.320, P=0.023), circumscribed margin ( OR=3.001,95% CI 1.374-6.554, P=0.006), edema ( OR=4.407,95% CI 2.100-9.244, P<0.001), and the minimal or mild degree of BPE ( OR=2.520,95% CI 1.328-4.782, P=0.005) were significant independent factors in predicting BRCA gene mutations. Conclusions:There are differences in MRI features and clinicopathological manifestations of BRCA-positive and BRCA-negative breast cancers. HER-2 status, the shape and margin of lesions, edema and the degree of BPE could independently predict the BRCA status of breast cancer.

Precision management of breast cancer is essential for improving patient outcomes.Radiomics enables high-throughput extraction of quantitative imaging features that reflect tumor biological behavior and heterogeneity from medical imaging data.With the rapid advancement of artificial intelligence(AI),particularly deep learning,feature extraction capability and model performance in radiomics have been significantly enhanced.Based on multimodal imaging modalities such as MRI,ultrasound and mammography,AI-and radiomics-based models can address key clinical challenges including tumor diagnosis,staging,molecular subtype prediction and therapeutic response assessment,thereby facilitating precision breast cancer care.This review summarizes recent advances in AI-driven radiomics for breast cancer diagnosis and treatment and discusses future application prospects and challenges toward clinical translation.

Objective:To investigate the value of preoperative prediction of Ki-67 expression status in breast cancer based on multi-phase enhanced MRI combined with clinical imaging characteristics prediction model.Methods:This study was retrospective. A total of 213 breast cancer patients who underwent surgical treatment at Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College between June 2016 and May 2017 were enrolled. All patients were female, aged 24-78 (51±10) years, and underwent routine breast MRI within 2 weeks prior to surgery. According to the different Ki-67 expression of postoperative pathological results, patients were divided into high expression group (Ki-67≥20%, 153 cases) and low expression group (Ki-67<20%, 60 cases). The radiomic features of breast cancer lesions were extracted from phase 2 (CE-2) and phase 7 (CE-7) images of dynamic contrast enhanced (DCE)-MRI, and all cases were divided into training and test sets according to the ratio of 7∶3. The radiomic features were first selected using ANOVA and Wilcoxon signed-rank test, followed by the least absolute shrinkage and selection operator method regression model. The same method of parameters selection was applied to clinical information and conventional imaging features [including gland classification, degree of background parenchymal enhancement, multifocal/multicentric, lesion location, lesion morphology, lesion long diameter, lesion short diameter, T 2WI signal characteristics, diffusion-weighted imaging (DWI) signal characteristics, apparent diffusion coefficient (ADC) values, time-signal intensity curve type, and axillary lymph nodes larger than 1 cm in short axis]. Support vector machine (SVM) was then used to construct prediction models for Ki-67 high and low expression states. The predictive performance of the models were evaluated using receiver operating characteristic (ROC) curves and area under cueve(AUC). Results:Totally 1 029 radiomic features were extracted from CE-2 and CE-7 images, respectively, and 9 and 7 best features were obtained after selection, respectively. And combining the two sets of features for a total of 16 features constituted the CE-2+CE-7 image best features. Five valuable parameters including lesion location, lesion short diameter, DWI signal characteristics, ADC values, and axillary lymph nodes larger than 1 cm in short axis, were selected from all clinical image features. The SVM prediction models obtained from the radiomic features of CE-2 and CE-7 images had a high AUC in predicting Ki-67 expression status (>0.70) in both the training set and the test set. The models were constructed by combining the CE-2, CE-7, and CE-2+CE-7 radiomic features with clinical imaging features, respectively, and the corresponding model performance in predicting Ki-67 expression status was improved compared with the models obtained by using the CE-2, CE-7, and CE-2+CE-7 radiomic features alone. The SVM prediction model obtained from CE-2+CE-7 radiomic features combined with clinical imaging features had the best prediction performance, with AUC of 0.895, accuracy of 84.6%, sensitivity of 87.9%, and specificity of 76.2% for predicting Ki-67 expression status in the training set and AUC of 0.822, accuracy of 70.3%, sensitivity of 76.1%, and specificity of 55.6% in test sets.Conclusion:The SVM prediction model based on DCE-MRI radiomic features can effectively predict Ki-67 expression status, and the combination of radiomic features and clinical imaging features can further improve the model prediction performance.

Objective:To compare magnetic resonance imaging (MRI)-based and computed tomography (CT)-based target volume delineation and dose coverage in partial breast irradiation (PBI) for patients with breast cancer, aiming to explore the application value of MRI localization in PBI after breast-conserving surgery.Methods:Twenty-nine patients with early breast cancer underwent simulating CT and MRI scans in a supine position. The cavity visualization score (CVS) of tumor bed (TB) was evaluated. The TB, clinical target volume (CTV), planning target volume (PTV) were delineated on CT and MRI images, and then statistically compared. Conformity indices (CI) between CT- and MRI-defined target volumes were calculated. PBI treatment plan of 40 Gy in 10 fractions was designed based on PTV-CT, and the dose coverage for PTV-MRI was evaluated.Results:The CVS on CT and MRI images was 2.97±1.40 vs. 3.10±1.40( P=0.408). The volumes of TB, CTV, PTV on MRI were significantly larger than those on CT, (24.48±16.60) cm 3vs. (38.00±19.77) cm 3, (126.76±56.81) cm 3vs. (168.42±70.54) cm 3, (216.63±81.99) cm 3vs. (279.24±101.55) cm 3, respectively, whereas the increasing percentage of CTV and PTV were significantly smaller than those of TB. The CI between CT-based and MRI-based TB, CTV, PTV were 0.43±0.13, 0.66±0.11, 0.70±0.09( P<0.001), respectively. The median percentage of PTV-MRI receiving 40 Gy dose was 81.9%(62.3% to 92.4%), significantly lower than 95.6%(95.0%~97.5%) of PTV-CT. Conclusions:The CVS between CT and MRI is not significantly different, but the MRI-based TB, CTV, PTV are significantly larger than CT-based values. The PTV-MRI is of underdose if PBI treatment plan is designed for PTV-CT. As a supplement of CT scan, MRI can enhance the accuracy of TB delineation after breast-onserving surgery.

Objective:To explore the diagnostic value of synthetic magnetic resonance imaging (syMRI) quantitative parameters for benign and malignant breast lesions.Methods:From September 2018 to March 2019, a total of 43 cases of breast lesions which were confirmed by surgery and pathology in Cancer Hospital, Chinese Academy of Medical Sciences were enrolled in this study. All patients underwent syMRI sequence scans before and after enhancement except for conventional T2WI, DWI, and enhancement scans. GE AW4.7 workstation was used to generate syMRI parameter maps (T1, T2, proton density mappings), and ITK-SNAP software was used to delineate the volume of interest. The T1, T2, PD values before and after dynamic contrast enhanced (DCE) were obtained, and the change values of each parameter were calculated. Meanwhile, the apparent diffusion coefficient (ADC) and time intensity curve (TIC) of the lesions were measured. The differences of each parameter value were compared between benign and malignant breast lesions, and the receiver operating characteristic (ROC) curve was used to analyze the diagnostic performance of each parameter.Results:Among the 43 enrolled cases, 13 were benign and 30 were malignant. Among the syMRI parameters, the pre-enhancement parameters including T1pre (median 1 663.07 ms), T2pre (median 103.33 ms), post-enhancement parameters ΔT1 (median 1 022.68 ms) and ΔT2 (median 27.67 ms) of benign group, significantly higher than those of the malignant group (the medians were 1 141.74, 92.53, 664.95, and 16.19 ms, respectively, P<0.05). The ADC value of the benign group (median 1.66×10 -3mm 2/s) was significantly higher than that of the malignant group (median 1.00×10 -3mm 2/s, P<0.05). The benign group included 6 cases of TIC curve type Ⅰ, 5 cases of type Ⅱ, and 2 cases of type Ⅲ. The malignant group included 2 cases of TIC curve type Ⅰ, 17 cases of type Ⅱ, and 11 cases of type Ⅲ. The difference between the two groups was statistically significant ( P<0.05). The area under the ROC curve (AUC) of T1pre before DCE was 0.869, higher than 0.806 of ADC and 0.697 of TIC. When the best cut-off value of 1 282.94 ms was chosen, the sensitivity and specificity of diagnosis were 76.9% and 93.3%, respectively. The combination of T1pre and T2pre can further improve the diagnostic performance (AUC=0.908). Conclusions:Among the syMRI quantitative parameters, T1pre, T2pre, ΔT1 and ΔT2 have good value for the differential diagnosis of benign and malignant breast lesions. T1pre has the best diagnostic performance, and the combination of T1pre and T2pre can further improve the diagnostic performance.

Objective:To explore the diagnostic value of synthetic magnetic resonance imaging (syMRI) quantitative parameters for benign and malignant breast lesions.Methods:From September 2018 to March 2019, a total of 43 cases of breast lesions which were confirmed by surgery and pathology in Cancer Hospital, Chinese Academy of Medical Sciences were enrolled in this study. All patients underwent syMRI sequence scans before and after enhancement except for conventional T2WI, DWI, and enhancement scans. GE AW4.7 workstation was used to generate syMRI parameter maps (T1, T2, proton density mappings), and ITK-SNAP software was used to delineate the volume of interest. The T1, T2, PD values before and after dynamic contrast enhanced (DCE) were obtained, and the change values of each parameter were calculated. Meanwhile, the apparent diffusion coefficient (ADC) and time intensity curve (TIC) of the lesions were measured. The differences of each parameter value were compared between benign and malignant breast lesions, and the receiver operating characteristic (ROC) curve was used to analyze the diagnostic performance of each parameter.Results:Among the 43 enrolled cases, 13 were benign and 30 were malignant. Among the syMRI parameters, the pre-enhancement parameters including T1pre (median 1 663.07 ms), T2pre (median 103.33 ms), post-enhancement parameters ΔT1 (median 1 022.68 ms) and ΔT2 (median 27.67 ms) of benign group, significantly higher than those of the malignant group (the medians were 1 141.74, 92.53, 664.95, and 16.19 ms, respectively, P<0.05). The ADC value of the benign group (median 1.66×10 -3mm 2/s) was significantly higher than that of the malignant group (median 1.00×10 -3mm 2/s, P<0.05). The benign group included 6 cases of TIC curve type Ⅰ, 5 cases of type Ⅱ, and 2 cases of type Ⅲ. The malignant group included 2 cases of TIC curve type Ⅰ, 17 cases of type Ⅱ, and 11 cases of type Ⅲ. The difference between the two groups was statistically significant ( P<0.05). The area under the ROC curve (AUC) of T1pre before DCE was 0.869, higher than 0.806 of ADC and 0.697 of TIC. When the best cut-off value of 1 282.94 ms was chosen, the sensitivity and specificity of diagnosis were 76.9% and 93.3%, respectively. The combination of T1pre and T2pre can further improve the diagnostic performance (AUC=0.908). Conclusions:Among the syMRI quantitative parameters, T1pre, T2pre, ΔT1 and ΔT2 have good value for the differential diagnosis of benign and malignant breast lesions. T1pre has the best diagnostic performance, and the combination of T1pre and T2pre can further improve the diagnostic performance.

Objective: To investigate the diagnostic value of diffusion kurtosis imaging (DKI) and its combination with DWI for differentiating benign and malignant breast lesions. Methods: Eighty two patients with clinically suspected breast lesions from May 2016 to February 2017 in the Cancer Hospital of Chinese Academy of Medical Sciences were prospectively enrolled in the study. Mammary MRI was performed in all the all patients (89 lesions), and the pathology results were confirmed by surgery or biopsies. All of them underwent 3.0 T MR examinations, including conventional fat-suppression imaging, DWI, DKI and dynamic contrast-enhanced MR imaging (DCE-MRI). The ADC values, mean diffusivity (MD), and mean diffusion kurtosis (MK) values of lesions were obtained, and the lesion morphology, enhancement patterns, and time-signal intensity curve (TIC) types were observed. Independent-samples t test, Mann-Whitney U test were used for the comparison of DKI and DWI between benign and malignant breast lesions. The characteristics of the lesions, enhancement manners and TIC between benign and malignant lesions were analyzed with χ² test. ROC was used to evaluate the efficacy of DKI and DWI parameters in the differential diagnosis of benign and malignant breast lesions. Results: There was no statistically difference for major axis, fiber types (P>0.05, respectively). The shape of enhancement mode and TIC had statistical significance (P<0.05, respectively). The ADC and MD values were significantly lower in malignant than in benign lesions. Conversely, MK value was significantly higher in malignant lesions than in benign ones, there was a statistically significant difference between these parameters (P<0.05, respectively). Among ADC, MD and MK values, the area under the curve (AUC) of MD for differentiating the benign lesions from the malignancy had the highest diagnosis efficiency (0.975), which was higher than the AUC of MK (0.969) and ADC (0.873). Combing the parameters of DWI and DKI, the diagnostic performance was superior to the single parameter. The AUC of the combination of MD and MK was 0.977, and the parameters of DKI combined with ADC-value had a equal diagnosis efficiency for the combination of MD and MK, which the areas under the ROC curve was 0.978. Conclusion The parameters derived from DKI can be used to distinguish benign breast lesions from malignant, and the combination of DKI and DWI could obtain a better diagnostic performance than single parameter.

Objective To investigate the value of intravoxel incoherent motion (IVIM) model of diffusion weighted MRI in assessing grades and enhancement patten of uterine cervical cancer. Methods Thirty one patients with pathologically proven cervical cancer, who underwent MRI scan preoperatively, were analyzed retrospectively and were divided into 3 groups according to their pathological grading of cacer, including 6 with G1 cancer, 17 with G2 and 8 with G3. The diameter of each lesion was≥1 cm. 10 b values (0, 30, 50, 100, 150, 200, 400, 800, 1 000, 1 500 s/mm2) were used in DWI, and DCE-MRI was performed with a time resolution of 9.8 s. Parameters of DWI (ADC, D, f, D*) and semiquantitative parameters of DCE-MRI (Slop, Maxslop, CER, Washout, AUC90) were measured. One-way ANOVA analysis of variance and Pearson correlation were used to analyze normally distributed continuous data. Kruskal-Wallis H test and Spearman correlation were used to analyze abnormally distributed continuous data. Tumor volume and all of the MRI parameters were compared as well as correlated with pathological grading.The perfusion parameters derived from IVIM were correlated with those derived from dynamic enhanced MR imaging. The sensitivity and specificity of f value to to diagnose G3 cervical cancer and the best cutoff were calculated from areas under the ROC curves.Results Tumor volume of G1,G2 and G3 cancers were(33.8±31.1),(19.6±16.9)and(31.2±29.1)cm3(F=1.147,P=0.332), respectively.ADC values of the three groups were(1.03 ± 0.11)× 10-3,(1.00 ± 0.10)× 10-3 and(0.90 ± 0.05)× 10-3mm2/s,respectively(F=4.619,P=0.018).D values of the three groups were (0.80 ± 0.11) × 10-3, (0.77 ± 0.06) × 10-3and (0.69 ± 0.06) × 10-3mm2/s ,respectively(F=5.272, P=0.011).f values of the three groups were 0.20±0.02, 0.22±0.03 and 0.24± 0.03, respectively (F=3.524, P=0.043).All of the others were of no significant difference (P>0.05).Both ADC and D correlated negatively with tumor grading (r=-0.464 and-0.493, P=0.009 and 0.005, respectively). f value correlated positively with tumor grading (r=0.436, P=0.014).Areas of ADC, D and f value under ROC curves to diagnose G3 cancers were 0.179, 0.147 and 0.690, respectively. While the cut-off value of f was 0.22, the diagnostic performance for G3 cancer was with a sensitivity of 75.0% (6/8) and a specificity of 60.9% (14/23). The value of f had weak positive correlations with Slop, Maxslop, CER and AUC90 of semiquantitative analysis of DCE-MRI (r=0.319, 0.337, 0.293 and 0.344, respectively, P<0.01). Conclusion IVIM model of multi-b value DWI may provide information in the assessment of differentiation and en hancement pattern of cervical cancer.