ABSTRACT
Purpose Diffusion-weighted magnetic resonance imaging (DW-MRI) is a contrast-free modality that has demonstrated ability to discriminate between pre-defined benign and malignant breast lesions. However, the ability of DW-MRI to discriminate cancer tissue from all other breast tissues in a clinical setting is unknown. Here we explore the ability to distinguish breast cancer from healthy breast tissues using signal contributions from the newly developed three-component multi-b-value DW-MRI model.
Experimental design Pathology-proven breast cancer patients from two datasets (n=81 and n=25) underwent multi-b-value DW-MRI. The three-component signal contributions C1 and C2 and their product, C1C2, were compared to the image defined on maximum b-value (DWImax), conventional apparent diffusion coefficient (ADC), and apparent diffusion kurtosis (Kapp). Ability to discriminate between cancer and healthy breast tissues was assessed by the false positive rate given sensitivity of 80% (FPR80) and receiver operating characteristic (ROC) area under the curve (AUC).
Results Mean FPR80 for both datasets was 0.016 (95%CI=0.008–0.024) for C1C2, 0.136 (95%CI=0.092–0.180) for C1, 0.068 (95%CI=0.049–0.087) for C2, 0.159 (95%CI=0.114–0.204) for DWImax, 0.731 (95%CI=0.692–0.770) for ADC and 0.684 (95%CI=0.660–0.709) for Kapp. Mean ROC AUC for C1C2 was 0.984 (95%CI=0.977–0.991).
Conclusions The three-component model yields a clinically useful discrimination between cancer and healthy breast tissues, superior to other DW-MRI methods and obliviating pre-defining lesions by radiologists. This novel DW-MRI method may serve as non-contrast alternative to standard-of-care dynamic contrast-enhanced MRI (DCE-MRI); removing the need to administer Gadolinium contrast decreases scan time and any accumulation of Gadolinium in the brain.
INTRODUCTION
Numerous studies have indicated that early breast cancer detection, with dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), has higher sensitivity than current screening programs (ultrasound and mammography)1-5. However, DCE-MRI has a number of limitations such as conflicting results regarding specificity2-6, dependency on expert radiologist readers, and the use of Gadolinium-based contrast agents that are linked to deposition in the brain7, additional scan time and additional costs8. In contrast, diffusion-weighted MRI (DW-MRI) does not require exogenous contrast and yields quantitative information of tissues microstructure by detecting diffusion of water molecules through application of varying degree of diffusion weighting (b-values).
Various diffusion models have demonstrated comparable ability to DCE-MRI in discriminating between pre-defined benign and malignant lesions in small regions of interest (ROIs) in the breast9-15. However, diffusion imaging-based tools are underexplored for discrimination of cancer tissue from all benign breast tissue (benign lesions and normal fat and fibroglandular tissue), which would increase its clinical utility and practicality in breast cancer screening, treatment evaluation, surgical planning, and surveillance, where lesions are not pre-defined by radiologists.
Multi-component modeling of the DW-MRI signal over extended ranges of b-values (typically up to 2000-3000 s/mm2) theoretically isolates the slowly diffusing water component present in cancer tissue in several organs16-20. Each component is defined by an apparent diffusion coefficient (ADC), and therefore represents a distinct pool of the water diffusion signal, with the corresponding magnitude. Vidić et al13 demonstrated that the normalized magnitude of the slowest element in a two-component model was excellent (AUC=0.99) in discriminating between benign and malignant breast lesions. Further, when optimizing the multi-component model to fit the DW-
MRI signal from all breast tissues, including cancer and healthy breast tissues, the result was a three-component model with empirical ADCs determined across patients and sites21. The main objective of the current study is to explore the ability of estimates derived from a three-component model to discriminate breast cancer from healthy breast tissues and to compare it to conventional DW-MRI imaging.
PATIENTS AND METHODS
Patients
In order to validate the discriminatory power of the three-component model across scanners and sites, two datasets of pathology-proven breast cancer patients from a United States (US) site (n=81) and a European site (n=25) were included (Table 1). Note that 49 cases from the US site and all cases from the European site were also used to determine the three-component model with fixed ADCs for breast tissue21. In addition, cases from the European site have been previously used for DW-MRI modeling of previously defined benign and malignant lesions13,22-25, linking DW-MRI signal to histological specimen26 and distortion correction techniques27.
US site
Patients with pathology-proven breast cancer with no cytotoxic regimens, chemotherapy, or ipsilateral radiation therapy for this malignancy prior to MRI scanning were eligible for this prospective study. The study was approved by the Institutional Review Board of the US site. The recruitment of patients began in December 2015 and ended in June 2019. Tumor categorization was done by histopathologic analysis of core needle and open incisional biopsies. In total, 14 patients were excluded from the study; nine patients had contralateral cancer or mastectomy, one patient had no visible cancerous tissue on DW-MRI, and in four patients image quality was low (low signal-to-noise ratio (n=2), poor fat saturation (n=1), and severe image distortion (n=1)), resulting in 81 patients.
European site
This prospective study was approved by the Regional Committees for Medical and Health Research Ethics (REC Central Norway, 2011/568). The recruitment of patients began in August 2014 and ended in August 2016. Inclusion criteria and tumor categorization were similar to those of the US site; for more details, see inclusion of malignant lesions from Vidić et al13.
MRI Acquisition
MRI data were acquired on a 3T GE scanner (MR750, DV25-26, GE Healthcare, Milwaukee, United States) and an 8-channel breast array coil with a bilateral axial imaging plane for the US dataset, while patients from the European dataset were imaged with a 3T Siemens scanner (Skyra, VD13-E11, Siemens Healthcare, Erlangen, Germany) and a 16-channel breast array coil with a unilateral sagittal imaging plane. Differences in scanner and pulse sequence parameters across sites were used to determine that the discriminatory potential of the three-component model is robust for data collected in different scanners and pulse sequence parameters. In addition to Gadolinium DCE-MRI and T2-images, both datasets included high b-value DW-MRI acquisition:
US site protocol
Bilateral axial DW-MRI was performed using reduced field of view (FOV) echo-planar imaging (EPI) including the following parameters: spectral attenuated inversion recovery (SPAIR) fat suppression, TE=82 ms, TR=9000 ms, b-values (number of diffusion directions)=0, 500 (6), 1500 (6), and 4000 (15) s/mm2, F0V=160×320 mm2, acquisition matrix=48×96, reconstruction matrix=128×128, voxel size=2.5×2.5×5.0 mm3, phase-encoding (PE) direction A/P, and no parallel imaging.
European site protocol
Unilateral sagittal DW-MRI was performed using Stejskal-Tanner spin-echo EPI including the following parameters: FatSat (n=15) and SPAIR (n=10) fat suppression, TE=88 ms, TR=10,600 ms (n=15) and 11,800 ms (n=10), b-values (number of diffusion directions)=0, 200 (6), 600 (6), 1200 (6), 1800 (6), 2400 (6), and 3000 (6) s/mm2, F0V=180×180 mm2, acquisition matrix=90×90, reconstruction matrix=90×90, voxel size=2.0×2.0×2.5 mm3, PE direction A/P, generalized auto-calibrating partially parallel acquisition (GRAPPA) with acceleration factor of 2 and 24 reference lines.
Image Processing and Analysis
Data were normalized to the 98th percentile of intensity within the b=0 s/mm2 image, indicated by the normalization factor (N). This was done to address different image intensity scaling while simultaneously preserving contribution of T2-decay to DW-MRI signal. Noise correction28 was performed to account for decreasing signal to noise ratio with increasing b-value. The observed signal (SIobs) is the mean signal across diffusion directions from one individual b-value image. Background voxels were selected by manually placing a ROI in an area in the air outside the breast on the highest b-value image, yielding the mean background intensity (SIbkg). The corrected signal intensity (SIcorr) calculated from SIobs and SIbkg is given as:
Furthermore, corrections for eddy current artifacts, motion17 and geometric distortion29 were applied for the European dataset.
Full-volume cancer and control ROIs were manually defined on DW-MRI images, guided by all available data in the exam protocol (including DCE-MRI and anatomical T2-images, Figure 1), under supervision of and validation by two breast radiologists: RRP (US site) and AØ (European site). Cancer ROIs were drawn for the lesions corresponding to pathology-proven cancer. Control ROIs were drawn for the entire contralateral breast (US site) and in a cancer-free region in the ipsilateral breast at least 10 mm away from the cancer ROI (European site), with the aim to include all representative healthy breast tissues, excluding the axillary region, large cysts (>2.5 cm), and susceptibility artifacts. Cancer and control ROIs were used to determine discriminatory performance between cancer and healthy breast tissues, respectively.
For comparison with other DW-MRI methods, the non-noise-corrected image defined on maximum b-value (DWImax), conventional apparent diffusion coefficient (ADC), and apparent diffusion kurtosis (Kapp) were estimated. DWImax was acquired at b=4000 s/mm2 for the US dataset and b=3000 s/mm2 for the European dataset. The exponential decay formulas described by Jensen et al30 and the corresponding b-value limits, <1000 s/mm2 and <2000 s/mm2, were used for computation of ADC and Kapp, respectively. Note that ADC and Kapp were normalized by signal at b=0 per voxel to eliminate T2-dependence30 rather than dividing all voxels by the same normalization factor to preserve T2 information as previously described.
To ensure that regions outside of the breast were not included in analysis, control ROIs were masked using intensity thresholding and 3D connected components (US site) or manually delineated within the breast boundary (European site) and reviewed by RRP (US site) and AØ (European site) (Figure 1 and Figure 4). Additionally, all undefined values (zero and infinite) on the image defined on b=0 s/mm2, ADC and Kapp were excluded.
Three-component modeling of DW-MRI signal
The DW-MRI signal across all available b-values was fit with a tri-exponential model where the slowest exponential had a fixed ADC of zero, yielding a three-component model with two exponentials and one constant term. The three-component model was chosen because it represents the best fit across all voxels from both healthy breast tissues and cancer tissue determined across patients and sites21, given as: where SDWI is the diffusion signal in arbitrary units, b is the b-value in s/mm2, and Ci denotes the voxel-wise unit-less signal contribution of each component. The ADC values were fixed as previously reported in21: ADC1=0 mm2/s and therefore not expressed as an exponential, ADC2=1.4×10-3 mm2/s, and ADC3=10.2×10-9 mm2/s Fixing ADCs ensures comparability of signal contributions across voxels and patients and avoids over-fitting.
The following parametric maps were estimated from the three-component model: C1C2, C1 and C2. The parameters C1 and C2 were estimated directly from the model, while C1C2 is the corresponding product. The parametric map C3 was not included due to low cancer conspicuity21.
Discriminating performance between cancer and healthy breast tissues
Clinical utility of the three-component derived parametric maps was assessed by comparing the discriminatory performance between cancer and healthy breast tissues ROIs of C1C2, C and C2 maps, DWImax, ADC, and Kapp. Because there were ~52 times more healthy breast tissue voxels than cancer voxels, performance in discriminating between cancer and healthy breast tissues was examined for all voxels by the expected false positive rate given a sensitivity of 80% (FPR80). In addition, the conventional performance measures receiver operating characteristic (ROC) area under the curve (AUC), sensitivity, specificity, and accuracy were estimated. Sensitivity, specificity, and accuracy were calculated for the threshold value providing optimal accuracy, defined as the mean sensitivity and specificity, assuming equal prevalence of cancer and healthy breast tissue voxels. C1C2, C1, C2, DWImax, and Kapp10-12 were assumed to have higher intensity for cancer compared to healthy breast tissues, while the opposite was assumed for ADC9.
RESULTS
Sample
The total number of cancer voxels and healthy tissue voxels from both datasets was 37,659 and 1,946,186, respectively.
Optimized three-component model parameters for discrimination (C1, C2, and C1C2)
Probability density colormaps for the three-component model given C1 and C2 including all voxels across patients and datasets are plotted for cancer (cancer ROIs, Figure 1A) and healthy breast tissues (control ROIs, Figure 1B). These maps display two distinct probability density distributions for cancer and healthy breast tissues. The product C1C2 discriminates cancer from healthy breast tissues voxels, where voxels low on one or two dimensions corresponds to healthy breast tissues voxels, while cancer probability increases with increased magnitude on C1 and C2.
The relationship between C1 and C2 demonstrates that voxels with high magnitude on both dimensions had the highest probability of cancer (Figure 1A); representative cases are given in Figure 1, and all cases are given in Supplementary Figure 1-106. Discrimination performance varied depending on healthy tissue composition in relation to the magnitude of C1 and C2 in cancer tissue. FPR80 was higher (indicating more false positive voxels) for C1 and DWImax in a case with abundant fibroglandular tissue and high C1-magnitude of corresponding cancer tissue (Figure 1B), compared to abundant suppressed fat tissue and high C2-magnitude of corresponding cancer (Figure 1C). The opposite was seen for C2, while C1C2 suppressed both fibroglandular and fatty tissue. This shows that the C1C2 parameter derived from the three-component model provided the optimal discrimination performance between cancer and healthy breast tissues.
Discriminatory performance of C1C2 compared to other DW-MRI methods
Mean FPR80 for both datasets was 0.016 (95%CI=0.008–0.024) for C1C2, 0.136 (95%CI=0.092–0.180) for C1, 0.068 (95%CI=0.049–0.087) for C2, 0.159 (95%CI=0.114–0.204) for DWImax, 0.731 (95%CI=0.692–0.770) for ADC and 0.684 (95%CI=0.660–0.709) for Kapp (Figure 3). C1C2 achieved the lowest FPR80 with a mean ROC AUC of 0.984 (95%CI=0.977–0.991) when compared to conventional DW-MRI; see Supplementary Table 1 for all conventional performance measures.
C1C2 had excellent performance compared to ADC and Kapp in a wide range of representative cases (Figure 4). DWImax performs well in several cases (Figure 4A-B) but underperforms compared to C1C2, overall (Figure 3) and particularly in a case with abundant fatty tissue (Figure 4C). In addition, C1C2 visually improves poor DCE-MRI specificity in a case with marked background parenchymal enhancement (Figure 4D). However, C1C2 underperforms in cases with sparse signal from cancer tissue, such as case of non-mass enhancement (NME) DCIS (Figure 4E). In this case, all DW-MRI derived maps failed to identify tumor tissue compared to healthy breast tissues. Furthermore, there was high diffusion signal from some healthy tissue components such as cysts (Figure 5B), subareolar ducts (Figure 5A), and partial-volume artifact from the interface of fibroglandular and fatty tissue (Figure 4D, Figure 5C). High diffusion signal from cysts and ducts may be defined as non-suspicious with the assistance of T2-images (Figure 5A-B).
DISCUSSION
Our study shows that breast cancer can be noninvasively discriminated from healthy breast tissues using the derived parameter C1C2 based on an advanced DW-MRI model, with results comparable to cancer detection using DCE-MRI2-6 (FPR80 mean=0.016, 95%CI=0.008–0.024 and ROC AUC mean=0.984, 95%CI=0.977–0.991). This means the C1C2 approach achieved very low false positive rates while detecting 80% or more of the defined cancer voxels. The discriminatory power of the C1C2 was superior to that of independent signal contributions, conventional DW-MRI-estimates (ADC) and other methods, including kurtosis (Kapp) and DWImax. The C1C2 approach was performed across two different sites, scanners, and acquisition protocols, suggesting potential for real-world applications. The development of this advanced DW-MRI method allows for improved conspicuity of breast cancer relative to background breast tissue. This lays the foundation for a quantitative framework specific to pathology which may serve as an alternative to DCE-MRI.
The high discriminatory performance is due to the novel C1C2 approach. In addition to malignancy, individual signal contributions from the three-component model were sensitive to the two primary components of healthy breast tissues: fat (C1) and fibroglandular (C2) tissue. Neither component was sensitive to tissues with very fast diffusion properties, such as vessels, necrosis, or edema. By combining the signal contributions of the two slowest components C1 and C2, the majority of healthy breast tissues signal was suppressed so that the output image was predominantly sensitive to cancer compared to healthy breast tissues. While optimized for cancer discrimination, the detailed relationship between the three-component model and breast microstructure remains to be studied, as it has been demonstrated for the two-component model26.
The C1C2 approach is performed on data acquired on extended imaging protocols (b-values up to 3000 s/mm2 and 4000 s/mm2) and requires at least three separate non-zero b-values. Inclusion of higher b-values likely improves discrimination by allowing better estimates of very slow diffusion characteristic of intracellular fluid within hypercellular tumors13,16-20. Furthermore, we attribute the high discriminatory performance to the normalization method, where T2-contribution to the DW-MRI signal was retained, though noise might also have been a partial contributor. The importance of T2 has been demonstrated in separating benign and malignant breast lesions, where the image at b=0 s/mm2 yielded a relatively high AUC of 0.8513.
We hypothesize that two contributing factors to the poor performance of ADC and Kapp include the restricted diffusion component from fat tissue and elimination of T2-effects that add to cancer discrimination. The FPR80 discriminatory performance of ADC and Kapp varied greatly across subjects; at best, performing around 0.2 in selected cases (Figure 4A), but overall do no better than chance. Conversely, DWImax shares the same basic properties as C1C2 (diffusion- and T2-weighting) and performs noticeably better than ADC and Kapp. However, DWImax is also prone to influence from fatty tissue. The C1C2 approach better accounts for fatty tissue, conferring a major advantage over DWImax and the other DWI-estimates (Figure 4C), as approximately 50% of women have almost entirely fatty breast tissue or scattered fibroglandular tissue31.
In order for C1C2 to be a noninvasive alternative to DCE-MRI for breast cancer detection, it must have comparable or better sensitivity and specificity. DW-MRI is known to improve detection specificity9,32, which is beneficial as lesion-level DCE-MRI specificity have been reported to range from 72-97%2-6. In our study, performance was assessed per voxel, and the patient cohort was heterogenous, consisting of a large range of tumor volumes (mean=10.6 cm3, range=0.2–105.9 cm3), not reflecting the typical patient pool in the screening or surveillance setting which typically have smaller lesions. However, the high performance of discriminating cancer from all other breast tissues in comparison to other DW-MRI-based methods is highly promising and suggests clinical utility comparable to DCE-MRI. C1C2 may be particularly useful when DCE-MRI demonstrates false positive33 and false negative34 interpretations in patients with moderate and marked background parenchymal enhancement (Figure 4D). Ideally, C1C2 may assist in a non-contrast workflow with T1- and/or T2-sequences as an alternative to DCE-MRI, where false positive findings can be defined as non-suspicious with the assistance of T2-images (Figure 5A-B). Thus, successful analysis using C1C2 removes the need to administer Gadolinium contrast, decreasing scan time and any accumulation of Gadolinium in the brain7.
Several diffusion methods aim to isolate the signal from the slowly diffusing water component from cancer tissue by utilizing broad b-value ranges 10-13,16-20. Kurtosis imaging is based on a simple mathematical representation of diffusion data where the derived parameter Kapp has proven potential utility in the breast10-12. More advanced, multi-component models with fixed ADCs have been developed to further probe the microstructure in the brain and prostate: restriction spectrum imaging (RSI, on which the C1C2 approach is based)16,17, the VERDICT model19, and the hybrid multidimensional MRI model20. A key difference between RSI/C1C2 and the hybrid multidimensional MRI model is that the hybrid model does not use global ADCs, making comparison across patients difficult; however, the hybrid model does incorporate multi-echo information not available in our study. Moreover, the T2-effects seen in RSI/C1C2 are removed from the other two modeling approaches, potentially reducing cancer conspicuity. Although the other multi-compartment models have shown promising results as cancer biomarkers in the prostate, for example, these results may be limited in breast, where fatty tissue is an important component of healthy breast tissues.
There were some limitations to our study. First, the C1C2 methodology did not correct for partial-volume artifacts which occurred at the interface between fatty and fibroglandular tissue (Figure 4D, Figure 5C). Such artifacts have the potential to be corrected, which was not investigated in this study but is an area of interest for future improvement. Another limitation concerned the definition of control regions; although we ensured that all control regions were verified as cancer-free, based on MRI review by an expert breast radiologist (both datasets) and exclusion of cases with pathology-proven contralateral cancer in the US dataset, we cannot know if occult cancer may have been included in the control region. The unilateral European dataset may have been particularly prone to this, as the control ROIs were defined in the same breast as the cancer (this also made the size of control regions dependent on the extent of cancer and thus variable from case to case in that dataset). Lastly, detection performance is commonly evaluated at the lesion level. This study used a voxel-wise false positive rate, FPR80, as its performance measure, which does not give an absolute measure comparable to other literature. However, we argue that such a measure is useful from a radiologist’s perspective, because it mimics a breast cancer examination where all voxels in the entire image are used.
In conclusion, our study is the first to demonstrate that the derived parameter C1C2, which is the product of the two slowest components of a three-component DW-MRI model, yields a clinically useful, noninvasive method for discrimination between cancer and healthy breast tissues.
The model eliminates the need for pre-defined lesions that conventional quantitative DW-MRI metrics use. Together with anatomical images such as T2, C1C2 may assist in a combined, non-contrast workflow which could serve as an alternative to DCE-MRI. The highly promising diagnostic properties were generalized across sites, scanners, and acquisition protocols, which is important for feasibility of large-scale studies for validation in routine breast cancer detection and follow-up in comparison to DCE-MRI.
Data Availability
we do not have IRB approval to share raw data with general public.
Competing Interest Statement
Dr. Dale reports that he was a Founder of and holds equity in CorTechs Labs, Inc., and serves on its Scientific Advisory Board. He is a member of the Scientific Advisory Board of Human Longevity, Inc. He receives funding through research grants from GE Healthcare to UCSD. Dr. Rakow-Penner is a consultant for Human Longevity, Inc. and receives funding through research grants from GE Healthcare. The terms of these arrangements have been reviewed by and approved by UCSD in accordance with its conflict of interest policies. Dr. Igor Vidić is employed as a consultant for Cortechs Labs, Inc. Dr. Seibert reports personal honoraria in the past three years from Varian Medical Systems, Multimodal Imaging Services Corporation, and WebMD.
Published work
Part of the study has been accepted for presentation at ISMRM, Virtual Exhibition, 8-13. August 2020 (Andreassen MMS et al, Discrimination of breast cancer from healthy breast tissues using a tri-exponential model)
Acknowledgements/Financial support
California Breast Cancer Research Program Early Career Award, GE Healthcare, Fulbright Scholarship Program, Mid-Norway Regional Health Authority #46084400, University of California Breast Cancer Screening Program #25IB-0056, NIH/NIBIB #K08EB026503.