Intravoxel incoherent motion magnetic resonance imaging for breast cancer: A comparison with molecular subtypes and histological grades

Prediction of the Nottingham prognostic index and molecular subtypes of breast cancer through multimodal magnetic resonance imaging

PURPOSE

To explore the ability of intravoxel incoherent motion (IVIM), diffusion kurtosis imaging (DKI) and background parenchyma enhancement (BPE) to predict the Nottingham prognostic index (NPI) and molecular subtypes of breast cancer (BC).

MATERIALS AND METHODS

In this study, 93 patients with BC were included, and they all underwent DKI, IVIM and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) examinations. The corresponding mean kurtosis value (MK), pure diffusion (MD), perfusion fraction (f), pseudo diffusion coefficient (D*), true diffusion coefficient (D), and BPE were measured. We used logistic regression analysis to investigate the relevance between the NPI, molecular subtypes and variables. The diagnostic efficacy was analyzed using receiver operating characteristic curves (ROC).

RESULTS

The MD and D values of the high-level NPI group were significantly lower than those of the low-level NPI group (p < 0.01), and the f value of the high-level NPI group was obviously higher than that of low-level NPI group (p < 0.001). The area under curve (AUC) of the combined model (f + D) was 0.824. Comparing with non-Luminal subtypes, the Luminal subtypes showed obviously lower MK, f and D*, and the AUC of the combined model (MK + f + D*) was 0.785. In comparison to other subtypes, the MK and D* values of triple-negative subtype were higher than other subtypes, and the combined model (MK + D*) represented an AUC of 0.865.

CONCLUSION

The quantitative parameters of DKI and IVIM have vital value in predicting the NPI and molecular subtypes of BC, while BPE could not provide additional information. Besides, these combined models can obviously improve the prediction performance.

Multiparametric MRI model to predict molecular subtypes of breast cancer using Shapley additive explanations interpretability analysis

PURPOSE

The purpose of this study was to assess the predictive performance of multiparametric magnetic resonance imaging (MRI) for molecular subtypes and interpret features using SHapley Additive exPlanations (SHAP) analysis.

MATERIAL AND METHODS

Patients with breast cancer who underwent pre-treatment MRI (including ultrafast dynamic contrast-enhanced MRI, magnetic resonance spectroscopy, diffusion kurtosis imaging and intravoxel incoherent motion) were recruited between February 2019 and January 2022. Thirteen semantic and thirteen multiparametric features were collected and the key features were selected to develop machine-learning models for predicting molecular subtypes of breast cancers (luminal A, luminal B, triple-negative and HER2-enriched) by using stepwise logistic regression. Semantic model and multiparametric model were built and compared based on five machine-learning classifiers. Model decision-making was interpreted using SHAP analysis.

RESULTS

A total of 188 women (mean age, 53 ± 11 [standard deviation] years; age range: 25-75 years) were enrolled and further divided into training cohort (131 women) and validation cohort (57 women). XGBoost demonstrated good predictive performance among five machine-learning classifiers. Within the validation cohort, the areas under the receiver operating characteristic curves (AUCs) for the semantic models ranged from 0.693 (95% confidence interval [CI]: 0.478-0.839) for HER2-enriched subtype to 0.764 (95% CI: 0.681-0.908) for luminal A subtype, inferior to multiparametric models that yielded AUCs ranging from 0.771 (95% CI: 0.630-0.888) for HER2-enriched subtype to 0.857 (95% CI: 0.717-0.957) for triple-negative subtype. The AUCs between the semantic and the multiparametric models did not show significant differences (P range: 0.217-0.640). SHAP analysis revealed that lower iAUC, higher kurtosis, lower D*, and lower kurtosis were distinctive features for luminal A, luminal B, triple-negative breast cancer, and HER2-enriched subtypes, respectively.

CONCLUSION

Multiparametric MRI is superior to semantic models to effectively predict the molecular subtypes of breast cancer.

Diffusion kurtosis imaging and intravoxel incoherent motion imaging parameters in breast lesions: Effect of radiologists' experience and region-of-interest selection

PURPOSE

To investigate the influence of ROI placement methods and radiologists' experience on diffusion kurtosis imaging (DKI) and intravoxel incoherent motion (IVIM) parameters' diagnostic performance in differentiating benign and malignant lesions based on the mass and non-mass enhancement (NME).

METHODS

We evaluated 138 lesions in 131 patients retrospectively. The IVIM and DKI parameter values were measured by three radiologists with different experiences independently using two different ROI placement methods. IVIM parameters include diffusion coefficient (ADCstand), true diffusion coefficient (ADCslow), pseudo-diffusion coefficient (ADCfast) and perfusion fraction (f). DKI parameters include mean diffusivity (MD) and mean kurtosis (MK). Each radiologist measured the lesions twice with a 3-month interval. We utilized intra-class correlation (ICC) to determine the inter- and intra-reader agreement for mass and NME, respectively. ROC analysis compared the diagnostic performance of parameters between different radiologists, ROI methods, and between mass and NME.

RESULTS

In mass lesions, inter- and intra-observer agreement were perfect for all parameters (ICC: 0.800-989). In NME, the inter-observer agreement was substantial to perfect for all parameters(ICC: 0.703-877), the intra-observer agreement of the senior and intermediate radiologists was substantial to perfect(ICC: 0.748-931) and the intra-observer agreement of the junior radiologist was moderate to substantial(ICC: 0.569-784). The diagnostic performance of ADCslow (Z = 2.209, P = 0.023), MD (mean diffusivity) (Z = 2.887, P = 0.004), and MK (mean kurtosis) (Z = 2.080, P = 0.038) in the small ROI measured by the senior radiologist was better than that of the junior radiologist for NME. The diagnostic performance of ADCslow in the large ROI measured by the senior radiologist (Z = 2.281, P = 0.023) and intermediate radiologist (Z = 2.867, P = 0.0041) was better than the junior radiologist for mass lesions. The diagnostic performance of ADCslow, ADCstand, MD, and MK did not show a significant difference between the two ROI placement methods (P > 0.05).

CONCLUSION

The observers' experience can influence the ROI selection and the diagnostic performance of ADCslow, ADCstand, MD, and MK measured using different methods show equal diagnostic performance.

Breast Cancer Subtypes and Quantitative Magnetic Resonance Imaging: A Systemic Review

Magnetic resonance imaging (MRI) is the most sensitive imaging modality for breast cancer detection. This systematic review investigated the role of quantitative MRI features in classifying molecular subtypes of breast cancer. We performed a literature search of articles published on the application of quantitative MRI features in invasive breast cancer molecular subtype classification in PubMed from 1 January 2002 to 30 September 2021. Of the 1275 studies identified, 106 studies with a total of 12,989 patients fulfilled the inclusion criteria. Bias was assessed based using the Quality Assessment of Diagnostic Studies. All studies were case-controlled and research-based. Most studies assessed quantitative MRI features using dynamic contrast-enhanced (DCE) kinetic features and apparent diffusion coefficient (ADC) values. We present a summary of the quantitative MRI features and their correlations with breast cancer subtypes. In DCE studies, conflicting results have been reported; therefore, we performed a meta-analysis. Significant differences in the time intensity curve patterns were observed between receptor statuses. In 10 studies, including a total of 1276 lesions, the pooled difference in proportions of type Ⅲ curves (wash-out) between oestrogen receptor-positive and -negative cancers was not significant (95% confidence interval (CI): [−0.10, 0.03]). In nine studies, including a total of 1070 lesions, the pooled difference in proportions of type 3 curves between human epidermal growth factor receptor 2-positive and -negative cancers was significant (95% CI: [0.01, 0.14]). In six studies including a total of 622 lesions, the pooled difference in proportions of type 3 curves between the high and low Ki-67 groups was significant (95% CI: [0.17, 0.44]). However, the type 3 curve itself is a nonspecific finding in breast cancer. Many studies have examined the relationship between mean ADC and breast cancer subtypes; however, the ADC values overlapped significantly between subtypes. The heterogeneity of ADC using kurtosis or difference, diffusion tensor imaging parameters, and relaxation time was reported recently with promising results; however, current evidence is limited, and further studies are required to explore these potential applications.

Prediction of Prognostic Factors and Genotypes in Patients With Breast Cancer Using Multiple Mathematical Models of MR Diffusion Imaging

Purpose

To explore the clinical value of apparent diffusion coefficient (ADC), intravoxel incoherent motion (IVIM), and diffusion kurtosis imaging (DKI) based on diffusion-weighted MRI (DW-MRI) for predicting genotypes and prognostic factors of breast cancer.

Materials and Methods

A total of 227 patients with breast cancer confirmed by pathology were reviewed retrospectively. Diffusion-weighted imaging (DWI), IVIM, and DKI were performed in all patients. The corresponding ADC, true diffusion coefficient (D), perfusion-related diffusion coefficient (D*), perfusion fraction (f), mean diffusion rate (MD), and mean kurtosis value (MK) were measured. Multivariate logistic regression analysis and receiver operating characteristic (ROC) curve were used to analyze the diagnostic efficacy in predicting the Nottingham prognostic index (NPI), the expression of antigen Ki-67, and the molecular subtypes of breast cancer. The nomogram of the combined genotype-prediction model was established based on the multivariate logistic regression model results.

Results

D* and MK values were significantly higher in the high-grade Nottingham group (NPI ≥ 3.4) than the low-grade Nottingham group (NPI < 3.4) (p < 0.01). When D* ≥ 30.95 × 10⁻³ mm²/s and MK ≥ 0.69, the NPI tended to be high grade (with areas under the curve (AUCs) of 0.712 and 0.647, respectively). The combination of D* and MK demonstrated the highest AUC of 0.734 in grading NPI with sensitivity and accuracy of 71.7% and 77.1%, respectively. Additionally, higher D*, f, and MK and lower ADC and D values were observed in the high Ki-67 than low Ki-67 expression groups (p < 0.05). The AUC of the combined model (D + D* + f + MK) was 0.755, being significantly higher than that of single parameters (Z = 2.770~3.244, p = 0.001~0.006) in distinguishing high from low Ki-67 expression. D* and f values in the Luminal A subtype were significantly lower than in other subtypes (p < 0.05). Luminal B showed decreased D value compared with other subtypes (p < 0.05). The HER-2-positive subtype demonstrated increased ADC values compared with the Luminal B subtype (p < 0.05). Luminal A/B showed significantly lower D, D*, MD, and MK than the non-Luminal subtypes (p < 0.05). The combined model (D + D* + MD + MK) showed an AUC of 0.830 in diagnosing the Luminal and non-Luminal subtypes, which is significantly higher than that of a single parameter (Z = 3.273~4.440, p < 0.01). f ≥ 54.30% [odds ratio (OR) = 1.038, p < 0.001] and MK ≥ 0.68 (OR = 24.745, p = 0.012) were found to be significant predictors of triple-negative subtypes. The combination of f and MK values demonstrated superior diagnostic performance with AUC, sensitivity, specificity, and accuracy of 0.756, 67.5%, 77.5%, and 82.4%, respectively. Moreover, as shown in the calibration curve, strong agreements were observed between nomogram prediction probability and actual findings in the prediction of genotypes (p = 0.22, 0.74).

Conclusion

DWI, IVIM, and DKI, as MR diffusion imaging techniques with different mathematical models showed potential to identify the prognosis and genotype of breast cancer. In addition, the combination of these three models can improve the diagnostic efficiency and thus may contribute to opting for an appropriate therapeutic approach in clinic treatment.