Breast DCE-MRI segmentation for lesion detection by multi-level thresholding using student psychological based optimization

Non-mass enhancement (NME) in breast dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) often shows subtle and diffuse enhancement, which can make accurate delineation challenging for radiologists. In this study, we developed an automated segmentation method for NME in DCE-MRI using a three-dimensional (3D) TransUNet model with a temporal attention mechanism. The attention module uses images of the difference between the pre- and early post-contrast phases to highlight regions that exhibit enhancement-related temporal changes. Our database consisted of 151 DCE-MRI cases with NME, each including both pre- and early post-contrast phases. The proposed network comprises two parallel CNN encoders (one for early post-contrast images and one for early–pre subtraction images), a temporal attention module, a transformer encoder, and a CNN decoder. The two encoders independently extract features from their respective inputs. The temporal attention module uses the features from the difference branch to weight those of the early branch. The weighted features are then fused and passed to the transformer encoder, which captures spatial and contextual dependencies within the 3D volume. The CNN decoder reconstructs the NME segmentation map through transposed convolutions and upsampling layers. Using five-fold cross-validation, the proposed method achieved a Jaccard index of 0.724 and a Dice coefficient of 0.802. Our approach significantly outperformed the baseline 3D TransUNet (Jaccard: 0.585, p < .001; Dice: 0.690, p < .001). These results demonstrate the high segmentation performance of the proposed method for NME and indicate its potential to improve diagnostic assessment in breast DCE-MRI.

Initial Experience of Applying TWIST-Dixon With Flexible View Sharing in Breast DCE-MRI

Clinical and Pathologic Features of Clinically Occult Synchronous Bilateral Breast Cancers

Quantification of breast tumor heterogeneity for ER status, HER2 status, and TN molecular subtype evaluation on DCE-MRI

Accurate and efficient automated tumor segmentation in breast dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is highly desirable for computer-aided tumor diagnosis. We propose a novel automatic segmentation framework which incorporates mean-shift smoothing, superpixel-wise classification, pixel-wise graph-cuts partitioning, and morphological refinement. A set of 15 breast DCE-MR images, obtained from the American College of Radiology Imaging Network (ACRIN) 6657 I-SPY trial, were manually segmented to generate tumor masks (as ground truth) and breast masks (as regions of interest). Four state-of-the-art segmentation approaches based on diverse models were also utilized for comparison. Based on five standard evaluation metrics for segmentation, the proposed framework consistently outperformed all other approaches. The performance of the proposed framework was: 1) 0.83 for Dice similarity coefficient, 2) 0.96 for pixel-wise accuracy, 3) 0.72 for VOC score, 4) 0.79 mm for mean absolute difference, and 5) 11.71 mm for maximum Hausdorff distance, which surpassed the second best method (i.e., adaptive geodesic transformation), a semi-automatic algorithm depending on precise initialization. Our results suggest promising potential applications of our segmentation framework in assisting analysis of breast carcinomas.

The highest number of deaths among all types of cancers in women is caused by breast cancer. Therefore, early detection and diagnosis of breast cancer are very much needed for its treatment. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) is widely used for breast cancer diagnosis. In this paper, a segmentation method using modified hard-clustering technique with multi-verse optimizer (MVO) is proposed for the detection of breast lesion in DCE-MRI. The proposed method is termed as CMVO in this paper. First, MR images are denoised, and intensity inhomogeneities are corrected in the preprocessing steps. Then clustering technique is used in segmentation of the MR images. Finally, from the segmented images, lesions are extracted in the postprocessing step. The results of CMVO are compared with that of K-means algorithm and PSO-based hard clustering. The CMVO performs better than other methods in lesion detection in breast DCE-MRI. KeywordsBreast cancerDCE-MRISegmentationClusteringMulti-Verse Optimizer

Purpose Background parenchymal enhancement (BPE) estimated from breast dynamic contrast enhanced MRI (DCE-MRI) has been correlated with breast cancer risk. Multiple time-point post-contrast-subtracted sequences are acquired in typical clinical breast DCE-MRI protocols. We quantitatively assessed BPE using fully automated computer algorithms and investigated the relationship between BPE quantified from three time-point post-contrast sequences and breast cancer risk prediction. Materials and Methods A retrospective case-control study was performed using breast DCE-MRI scans from 102 patients (mean 47.2±7.3 YO) who underwent either open surgical biopsy or core biopsy from 2009-2011: 51 women had unilateral breast cancer and 51 were age- and date-of-MRI matched controls with a unilateral biopsy-proven benign. For each MRI scan, three post-contrast-subtracted sequences (i.e., SUB 1, SUB 2 and SUB 3) acquired over a total of 7 minutes were analyzed. BPE was quantified from each of the sequences using fully automated computer algorithms on the breasts contralateral to the cancers and the contralateral (negative) breasts of the controls. For each sequence, two quantitative BPE measures were generated: the absolute BPE volume (|BPE|) and its relative amount over the whole breast volume (BPE%). Volumetric absolute and relative amounts of fibroglandular tissue (|FGT| and FGT%) were also automatically quantified, from the pre-contrast sequence. BI-RADS breast density assessment was retrieved from the mammography report (&amp;lt; 6 months) prior to cancer diagnosis (for cases) or MRI (for controls). Multivariable conditional logistic regression was performed to assess BPE measures, quantified respectively from SUB 1, SUB 2, and SUB 3, as predictors of breast cancer risk. Results Breast cancer risk odds ratios (ORs) were reported by |BPE| and BPE% (Table 1), after adjustment for breast density, |FGT|, and FGT%. Breast density did not correlate with risk for breast cancer in this study cohort: OR for BI-RADS density alone was 0.75 (95% CI: 0.35, 1.59; p=0.5). OR was 1.14 (95% CI: 0. 72, 1.81; p=0.6) for |FGT| alone and 0.70 (95% CI: 0.19, 2.52; p=0.6) for FGT% alone. Table 1: Breast cancer risk odds ratios (ORs) with 95% confidence intervals [CIs] for BPE quantified from SUB 1, SUB 2, and SUB 3, respectively. BPE from SUB1BPE from SUB2BPE from SUB3|BPE|2.0 (95% CI: 1.1-3.7; p=0.02)2.0 (95% CI: 1.2-3.3; p=0.01)1.7 (95% CI: 1.1-2.7; p=0.02)BPE%3.6 (95% CI: 1.3-10.1; p=0.01)2.8 (95% CI: 1.3-6.2; p=0.01)2.4 (95% CI: 1.2-5.0; p=0.02) Conclusions Increased BPE quantified from DCE-MRI are predictive of breast cancer risk, independent of measures of breast density and FGT. BPE estimated from three time-point SUB sequences has a similar predictive effect of breast cancer risk, while an early sequence (e.g., SUB 1) appears to have a larger magnitude of effect than that of a delayed sequence (e.g., SUB 3). Clinical Relevance Quantified BPE in breast DCE-MRI has potential for use as a biomarker of breast cancer risk and may be included to improve breast cancer risk prediction. A single post-contrast sequence (i.e., SUB 1) may be adequate for use to estimate breast cancer risk using breast MRI in the context of breast cancer screening for high-risk women. Citation Format: Shandong Wu, Wendie A Berg, Margarita L Zuley, Brenda F Kurland, Rachel C Jankowitz, Jules Sumkin, Robert M Nishikawa, David Gur. Quantitative assessment of early- and delayed DCE-MRI background parenchymal enhancement in breast cancer risk prediction [abstract]. In: Proceedings of the Thirty-Seventh Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2014 Dec 9-13; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2015;75(9 Suppl):Abstract nr P1-01-07.

Background/objectives: Dynamic Contrast Enhanced-Magnetic Resonance Imaging (DCE-MRI) is widely used in tumor diagnosis, staging and assessment of therapy response for different types of tumors, thanks to its capability to provide important functional information about tissue microvasculature. Tracer kinetic models used for estimating microcirculatory parameters can be broadly categorized as conventional compartmental (CC) or distributed- parameter (DP) models. While DP models seem to be more realistic, CC models (in particular the Tofts and the Brix models) have been widely used in clinical investigations over the past two decades. However, to date there is no direct comparison of CC vs DP models on real breast DCE-MRI data; moreover, a direct comparison between Tofts and Brix models, has not yet been reported on real breast data. Therefore, the purpose of this study was two-fold: on the one hand we analyzed the performance, on real breast DCE-MRI data, of CC vs DP models in terms of goodness-of-fit metrics; on the other hand we compared Tofts and Brix models on the basis of real breast DCE-MRI data. Methods: Three models were compared: two CC models (the Tofts and the Brix models) and one DP model (the ATH model). We gathered data in two different scenarios: DCE-MRI with high temporal resolution obtained by means of a k -space under-sampling and data sharing method known as Time-resolved angiography With Stochastic Trajectories (TWIST) and DCE-MRI with low temporal resolution obtained by means of the Spoiled Gradient-Echo k -space scheme known as Fast Low Angle Shot (FLASH). The performances of the three models were evaluated by means of three goodness-of-fit metrics: the Residual Sum of Squares, the Bayesian Information Criterion and the Akaike Information Criterion on four breast DCE-MRI examinations. Results: Although not conclusive, the results of this study suggest that the ATH model can achieve better fit in comparison to the Tofts and Brix models for TWIST data; and that the Brix model can achieve better fit with respect to the Tofts model for FLASH data. Conclusion: Given the current typical settings of clinical breast DCE-MRI examinations, there seems not to be a clear advantage, in terms of goodness-of-fit, of ATH with respect to Tofts and Brix models; moreover, at lower temporal resolution the Brix model can achieve better fit than the Tofts model.

In patients with dense breasts or at high risk of breast cancer, dynamic contrast enhanced MRI (DCE-MRI) is a highly sensitive diagnostic tool. However, its specificity is highly variable and sometimes low; quantitative measurements of contrast uptake parameters may improve specificity and mitigate this issue. To improve diagnostic accuracy, data need to be captured at high spatial and temporal resolution. While many methods exist to accelerate MRI temporal resolution, not all are optimized to capture breast DCE-MRI dynamics. We propose a novel, flexible, and powerful framework for the reconstruction of highly-undersampled DCE-MRI data: enhancement-constrained acceleration (ECA). Enhancement-constrained acceleration uses an assumption of smooth enhancement at small time-scale to estimate points of smooth enhancement curves in small time intervals at each voxel. This method is tested in silico with physiologically realistic virtual phantoms, simulating state-of-the-art ultrafast acquisitions at 3.5s temporal resolution reconstructed at 0.25s temporal resolution (demo code available here). Virtual phantoms were developed from real patient data and parametrized in continuous time with arterial input function (AIF) models and lesion enhancement functions. Enhancement-constrained acceleration was compared to standard ultrafast reconstruction in estimating the bolus arrival time and initial slope of enhancement from reconstructed images. We found that the ECA method reconstructed images at 0.25s temporal resolution with no significant loss in image fidelity, a 4x reduction in the error of bolus arrival time estimation in lesions (p < 0.01) and 11x error reduction in blood vessels (p < 0.01). Our results suggest that ECA is a powerful and versatile tool for breast DCE-MRI.

In this paper, we propose a Computer-Aided Detection (CAD) system based on the novel Multi-Classifier Fusion-based Classification Model (MCFM) of the breast lesions using textural features in Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI). The proposed system consists of four principal phases: Region Of Interest (ROI) segmentation, feature extraction, fusion and selection and finally Classification. We have extracted a complementary set of features from the DCE-MRI database on two statistical descriptors such as the Gray Level Co-occurrence Matrix (GLCM) and Local Binary Pattern (LBP) methods. Then, we have concatenated the extracted features using Serial method to exploit the complementary between these sets. Next, we have applied a features selection phase through Principal Component Analysis (PCA) and Feature Subset Selection (FSS) techniques to remove redundant information and improve the overall performance of breast DCE-MRI in a computer-aided system. For the classification, we have used a set of classifiers (K-Nearest Neighbors (K-NN), Support Vector Machines (SVM) and Multilayer Perceptron-Artificial Neural Network (MLP-ANN)). Then, a fusion Model between them using several metrics is applied. Experimental validation is performed over an MRI dataset, which is comprised of 286 patients with 143 Masses and 143 No Masses (Normal) breast tissues. These experiments report encouraging performances for mass detection, achieved using ten-fold cross-validation model, in terms of the averaged measurements of different metrics such as Sensitivity, Specificity, Area Under Curve (AUC) and Accuracy, which are 0.8972, 0.9572, 0.9630 and 0.9501, respectively. The proposed DCE-MRI analysis model can offer an efficient system for radiologists to identify the breast tissues.

Background: Inflammatory breast cancer (IBC) is unresectable at presentation, thus neoadjuvant systemic therapy (NAS) is the primary treatment for this aggressive disease. Eribulin (E) (Eisai Co., Ltd) is a synthetic analog of Halichondroin B that suppresses microtubule growth. Its properties of inducing a mesenchymal-to-epithelial transition and normalizing the vasculature of mammary tumors are particularly relevant to IBC. Pre-clinical data revealed changes in dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) parameters after exposure to E, thought to reflect tumor vasculature remodeling. We sought to examine the efficacy of E-containing NAS specifically for patients (pts) with HER2- IBC. We also characterized changes in breast DCE-MRI and diffusion weighted MRI (DWI) between baseline and 1 week (wk) after NAS. Methods: Pts with newly diagnosed HER2- IBC received 4 cycles of E 1.4 mg/m2 day (D)1, D8 every 21 D followed by 4 cycles of dose-dense (dd) doxorubicin (A) 60 mg/m2 + cyclophosphamide (C) 600mg/m2 every 2 wk followed by modified radical mastectomy (MRM) + postmastectomy radiation (PMRT) on a phase II prospective study (Cohort A). Primary endpoint was pathologic complete response (pCR) defined as no invasive disease in breast + axillary lymph nodes. Based on a Simon 2-stage design, if ≥ 2 of 16 pts had pCR then enrollment would continue to 25, with the regimen worthy of study if pCR ≥ 0.30 and not if pCR ≤ 0.10 (α=0.10, β=0.10). A second cohort was opened, changing the NAS sequence to ddAC followed by E (Cohort B). All pts had 2 research breast biopsies (rbx) for genomic analysis at baseline and 1 wk after first dose of NAS. An imaging study evaluated breast DCE-MRI and DWI at baseline and 1 wk after first dose of NAS. Tofts 2-compartment pharmacokinetic (PK) model calculated Ktrans, Ve, Kep, and iAUC from DCE. ADC was calculated from DWI. The change in Ktrans, Ve, Kep, iAUC, and ADC between baseline and wk 1 was evaluated using a signed-rank test. The magnitude of these changes was compared between Cohort A and Cohort B using a Wilcoxon rank-sum test. All testing was two-tailed and exact p-values were used; p-value &amp;lt; 0.05 were considered statistically significant. Results: 22 pts were enrolled, 16 pts in Cohort A and 6 pts in Cohort B before the study was closed due to slow accrual. Median age was 58 years, 3 pts (13.6%) had triple negative disease. 13 pts (59%) were postmenopausal. All but 1 had stage III disease. All 22 pts completed NAS followed by MRM + PMRT. 19 pts completed full course of NAS, 1 pt received E followed by 3 cycles of AC and 2 pts received 3 cycles of E in addition to 4 cycles of AC. In Cohort A, 1/16 pts had pCR (6.3%; 80% CI 0,22.2%). At a median follow up of 51 months, 3 pts had distant disease recurrence with 1 death; no local-regional recurrence. 7 pts experienced grade (gd) 3 or 4 adverse event (AE); most common were neutropenia and aminotransferase increase. 10 pts participated in the DCE-MRI and DWI substudy: 5 from Cohort A, 5 from Cohort B. Decrease in value from baseline to wk 1 was detected in Ktrans (median -49), Ve (-128), and iAUC (-57). The magnitude of changes was not different when comparing Cohort A to Cohort B. There was no significant change in Kep or ADC. Given the small size of each group, a lack of statistical significance should not be interpreted as suggestive of no effect. Evaluation of changes in angiogenesis- and EMT-related gene expression in rbx + residual disease is ongoing.Conclusion: A regimen of E + ddAC is tolerable NAS for HER2- IBC. All pts underwent curative surgery despite low pCR rate, commonly observed in ER+ IBC. There were changes in some DCE-MRI parameters after exposure to chemotherapy but may not be specific to E. Studies specifically designed for IBC are needed to enhance the understanding of this disease. Clinical trial: NCT02623972 Citation Format: Filipa Lynce, Eren D. Yeh, Meredith M. Regan, Lei Qin, Camden P. Bay, Ian Krop, Beth T Harrison, Faina Nakhlis, Jennifer Bellon, Beth Overmoyer. A phase 2 study of neoadjuvant systemic therapy with eribulin followed by doxorubicin and cyclophosphamide for HER2-negative inflammatory breast cancer [abstract]. In: Proceedings of the 2021 San Antonio Breast Cancer Symposium; 2021 Dec 7-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2022;82(4 Suppl):Abstract nr P2-12-18.

Breast cancer causes the highest death among all types of cancers in women. Early detection and diagnosis leading to early treatment can save the life. The computer-assisted methodologies for breast dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) segmentation can help the radiologists/doctors in the diagnosis of the disease as well as further treatment planning. In this article, we propose a breast DCE-MRI segmentation method using a hard-clustering technique with a Non-dominated Sorting Genetic Algorithm (NSGA-II). The well-known cluster validity metrics namely DB-index and Dunn-index are utilized as objective functions in NSGA-II algorithm. The noise and intensity inhomogeneities in MRI are removed from MRI in the preprocessing step as these artifacts affect the segmentation process. After segmentation, the lesions are separated and finally, localized in the MRI. The devised method is applied to segment 10 Sagittal T2-Weighted fat-suppressed DCE-MRI of the breast. A comparative study has been conducted with the K-means algorithm and the devised method outperforms K-means both quantitatively and qualitatively.

Breast dynamic contrast enhanced MRI (DCE-MRI) segmentation, based on the differential enhancement of image intensities, can help the clinician detect suspicious regions. Motivated by the recent success of texture learning and segmentation, we propose a novel segmentation method based on texture properties. The segmentation method consists of generating a library of texture primitives ”textons”, and then classifying each voxel into different tissue classes using textons and vector attributes. A Markov Random Measure field (MRF) method is combined with texture information to realise the spatial coherence. To evaluate our framework, twenty patients’ MRIs from our local hospital were used for texture learning, and a further twenty patients’ MRIs were used for testing.KeywordsSegmentation MethodSegmentation ResultImage PatchSpatial CoherenceSuspicious RegionThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

&lt;div&gt;Abstract&lt;p&gt;&lt;b&gt;Purpose:&lt;/b&gt; To determine whether markers of healthy breast stroma are able to select a subgroup of patients at low risk of death or metastasis from patients considered at high risk according to routine markers of the tumor.&lt;/p&gt;&lt;p&gt;&lt;b&gt;Experimental Design:&lt;/b&gt; Patients with ER&lt;sup&gt;+&lt;/sup&gt;/HER2&lt;sup&gt;−&lt;/sup&gt; breast cancer were consecutively included for retrospective analysis. The contralateral parenchyma was segmented automatically on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), where upon the average of the top-10% late enhancement was calculated. This contralateral parenchymal enhancement (CPE) was analyzed with respect to routine prognostic models and molecular assays (Nottingham Prognostic Index, Dutch clinical chemotherapy-selection guidelines, 70-gene signature, and 21-gene recurrence score). CPE was split in tertiles and tested for overall and distant disease-free survival. CPE was adjusted for patient and tumor characteristics, as well as systemic therapy, using inverse probability weighting (IPW). Subanalyses were performed in patients at high risk according to prognostic models and molecular assays.&lt;/p&gt;&lt;p&gt;&lt;b&gt;Results:&lt;/b&gt; Four-hundred-and-fifteen patients were included, constituting the same group in which the association between CPE and survival was discovered. Median follow-up was 85 months, 34/415(8%) patients succumbed. After IPW-adjustment for patient and tumor characteristics, patients with high CPE had significantly better overall survival than those with low CPE in groups at high risk according to the Nottingham Prognostic Index [HR (95% CI): 0.08 (0.00–0.40), &lt;i&gt;P&lt;/i&gt; &lt; 0.001]; Dutch clinical guidelines [HR (95% CI): 0.22 (0.00–0.81), &lt;i&gt;P&lt;/i&gt; = 0.021]; and 21-gene recurrence score [HR (95% CI): 0.14 (0.00–0.84), &lt;i&gt;P&lt;/i&gt; = 0.030]. One group showed a trend [70-gene signature: HR (95% CI): 0.25 (0.00–1.02), &lt;i&gt;P&lt;/i&gt; = 0.054].&lt;/p&gt;&lt;p&gt;&lt;b&gt;Conclusions:&lt;/b&gt; In patients at high risk based on the tumor, subgroups at relatively low risk were identified using pretreatment enhancement of the stroma on breast DCE-MRI. &lt;i&gt;Clin Cancer Res; 23(21); 6505–15. ©2017 AACR&lt;/i&gt;.&lt;/p&gt;&lt;/div&gt;

Breast lesion segmentation is an essential step in developing Computer-aided Diagnostic (CAD) system for the early detection of breast cancer. However, accurate lesion segmentation is challenging, since the lesions may have complicated topological structures and different intensity distributions. This paper describes a novel approach for breast lesion segmentation from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) using a level set-based active contour model. Its evolution is controlled by specially designed signed pressure function that only accounts for distribution of image background intensity, so solving the weak-edge-passed problem and improving the robustness to segment different lesions. To confirm the proposed approach, it is compared with different methods based on level set. Experiment results show that the proposed method leads to better results. On a total of 38 breast DCE-MRI studies, our method yields a mean difference (MAD) of 7.2561±12.4537 pixels and Jaccard index of 0.8564±0.0493 compared to manual ground truth segmentation.