Clinical Advances in PET/MRI for Breast Cancer

Summary

Imaging is paramount for early detection of breast cancer, clinical staging, informing management decisions, and directing therapy. Positron emission tomography/magnetic resonance imaging (PET/MRI) is a quantitative hybrid imaging technology that combines metabolic, functional PET data with anatomic detail and functional perfusion information from MRI. The clinical applications for which PET/MRI may be beneficial for breast cancer is an active area of research. This Review discusses the rationale for using PET/MRI for patients with breast cancer and summarizes the clinical evidence across the spectrum of diagnosis, staging, prognosis, tumor phenotyping, and treatment response assessment. Continued development and approval of targeted radiopharmaceuticals, together with radiomics and automated analysis tools, will further expand the opportunity for PET/MRI to provide added value for breast cancer imaging and patient care.

Introduction

Imaging serves a key role in breast cancer screening, diagnosis, staging, and management. Mammography, ultrasound, and dynamic contrast-enhanced breast magnetic resonance imaging (MRI) are the predominant breast cancer imaging modalities. For patients with suspected or known metastatic disease, whole-body imaging is performed with computed tomography (CT), bone scintigraphy, or positron emission tomography (PET)/CT. The standard radiopharmaceutical used for PET/CT imaging of patients with breast cancer is 2-deoxy-2-¹⁸F-fluoroglucose (FDG), which accumulates in tissues with high glucose metabolism. FDG PET/CT can be used for initial staging of newly diagnosed breast cancer patients to identify distant metastases, for restaging patients with suspected recurrence or metastasis, and for monitoring treatment response for patients with locally advanced and metastatic breast cancer.^1,2 MRI may also be used for focused evaluation of specific disease sites such as the brain, spine, or liver.

There has been long-standing interest in the potential synergy of combining PET with MRI for a more comprehensive evaluation. High spatial anatomic detail as well as cellular density and perfusion data is provided by MRI, while PET provides functional metabolic information. Historically, PET/MRI has been achieved by co-registration and fusion of images obtained separately from dynamic contrast-enhanced breast MRI and from PET/CT performed in the prone position. With the development of combined PET/MRI scanners in 2011, a single imaging session is now possible.^3–6 Depending on the vendor, the acquisition of PET and MRI data occurs sequentially or simultaneously. Despite different methods for attenuation correction, strong correlation of radiopharmaceutical uptake parameters have been demonstrated between PET/CT and simultaneous PET/MRI for patients with primary and metastatic breast cancer, which validates PET/MRI for quantitative analyses.^7–9

In evaluating studies of PET/MRI for breast cancer, it is important to clarify the type of acquisition used in the imaging protocol. Whole-body PET/MRI is performed with the patient in the supine position using surface radiofrequency coils, while dedicated breast PET/MRI is performed in the prone position using a dedicated breast coil (Figure 1). Four-, eight-, and 16-channel radiofrequency breast coils have been used with integrated PET/MRI scanners.^9–13 Imaging protocols can also be customized as a dedicated axillary PET/MRI exam for regional nodal staging.¹⁴ Whole-body PET/MRI can also be performed in conjunction with a dedicated prone breast PET/MRI for evaluating local tumor, regional nodal, and distant disease.¹⁵

Figure 1:

Simultaneous PET/MRI protocol. For dedicated prone breast PET/MRI, PET data is acquired at one bed-position during the MRI sequences [attenuation correction (MRAC);fluid-sensitive/T2;diffusion-weighted imaging (DWI);dynamic contrast-enhanced (DCE) T1].⁹ For supine whole-body PET/MRI, MRI sequences are acquired with PET data at each bed-position.¹⁵

The clinical applications for which PET/MRI provides added value for breast cancer patients is an active area of investigation. This Review summarizes published data regarding PET/MRI for breast cancer imaging across the spectrum of diagnosis, staging, prognosis, tumor phenotyping, and treatment response assessment.

Diagnosis

Breast MRI is the most sensitive (90–99%) modality for detecting breast cancer.^16–20 However, the relatively moderate specificity (72–89%) of breast MRI results in false-positive findings that require additional imaging and biopsy.^16–20 Thus, reducing false-positive findings from breast MRI has significant potential impact by avoiding additional biopsies, reducing cost, decreasing patient anxiety, and minimizing time to surgery.

Data from initial studies using separately acquired FDG PET/CT and breast MRI indicate that functional information provided by PET may improve the specificity of MRI. In a study of 36 women with confirmed or suspected breast cancer, Moy et al. found that, although breast MRI alone was highly sensitive (96%), fused prone FDG PET/CT with breast MRI increased the positive predictive value to 98% from 77% with MRI alone.²¹ Furthermore, fused PET/MRI increased specificity to 97% from 53% with MRI alone.²¹ Overall accuracy was 89% with fused PET/MRI and 78% with breast MRI alone.²¹

Adding advanced MRI functional parameters to PET may further improve diagnostic accuracy.^22–24 In a prospective study of 76 women with suspicious findings on conventional breast imaging, Pinker et al. found that combined data from prone FDG PET/CT and three parameters from MRI (dynamic contrast enhancement, diffusion restriction, and spectroscopy) had the highest accuracy (area under the receiver operating characteristic curve, AUC 0.935) for differentiating benign versus malignant breast lesions compared to dynamic contrast-enhanced MRI alone (AUC 0.86) and would have led to a reduction in biopsies recommended.²² Using data from prone FDG PET/CT and two parameters from MRI (contrast kinetics and diffusion restriction), Bitencourt et al. demonstrated 89.5% accuracy in 31 patients with suspicious lesions on breast MRI, reducing unnecessary biopsies without missing any cancers.²³

Based on studies using separately acquired breast MRI and PET/CT, dedicated prone breast PET/MRI was expected to have improved specificity compared to MRI alone for distinguishing malignant from benign breast lesions. A retrospective study by Botsikas et al. of 58 breast cancer patients with 101 breast lesions (83 malignant, 18 benign) found that combined breast PET/MRI had higher specificity for lesion assessment compared to MRI alone (100% vs 67%), as predicted.²⁵ However, Grueneisen et al. found no significant difference between dedicated simultaneous prone PET/MRI and MRI alone for characterization (i.e., true-positives, true-negatives, false-positives, false-negatives) of 83 lesions in 49 patients with primary breast cancer.²⁶ Similarly, Heusner et al. also demonstrated no significant difference in accuracy comparing fused breast PET/MRI (92%) with MRI alone (85%) for evaluating 58 lesions in 27 patients with breast cancer.²⁷ In the retrospective study by Garcia-Velloso et al. evaluating 107 lesions in 45 women with breast cancer, the AUC for fused PET/MRI was 0.98 and 0.95 for MRI alone, but this was not statistically significant.²⁸ There is an ongoing clinical trial (NCT03510988) directly comparing the specificity of simultaneous breast FDG PET/MRI to MRI alone which may lend additional data regarding this question.

The optimal approach for improving the specificity of breast MRI by adding PET may require a minimum threshold for lesion size. Several studies using separately acquired FDG PET/CT and breast MRI have performed analyses using a 10 mm size threshold. A direct comparison of separately acquired prone FDG PET/CT and contrast-enhanced breast MRI showed equal overall diagnostic accuracy for breast cancer detection of 93% for each modality for all 172 suspicious lesions in a study by Magometschnigg et al. However, the diagnostic accuracy decreased to 91% for each modality for lesions less than 10 mm in size.²⁹ Although not statistically significant, breast MRI was more sensitive but less specific than FDG PET/CT for lesions smaller than 10 mm resulting in equivalent overall diagnostic accuracy.²⁹ When including only masses larger than 10 mm, Bitencourt et al. demonstrated that separately acquired, fused prone PET/MRI provided 93.3% accuracy, with 95.8% sensitivity and 83.3% specificity, in a study of 60 women with 76 suspicious lesions on breast MRI who subsequently underwent FDG PET/CT in the prone position.³⁰ Furthermore in a retrospective study of 80 patients with biopsy-confirmed breast cancer and additional suspicious lesions identified with pre-operative breast MRI (37 malignant, 24 benign), Jalaguier-Coudray et al. found that the negative predictive value from a separately acquired supine FDG PET/CT was 100% for lesions larger than 1 cm.³¹ The authors suggested that biopsy could be potentially avoided if MRI lesions are at least 1 cm without FDG uptake, but this approach needs confirmation with a larger prospective study.³¹ Thus, analysis of the influence of lesion size on diagnostic accuracy should be considered for future studies using simultaneous breast PET/MRI.

In addition to lesion size, tumor glycolytic activity is also expected to influence the sensitivity of dedicated prone breast PET/MRI when using FDG. This conclusion is inferred from early studies of whole-body FDG PET that demonstrated false-negative results and reduced sensitivity for cancers with low nuclear grade, invasive lobular carcinoma, ductal carcinoma in situ, and small tumors (≤10 mm).^32,33 Thus, clinical practice guidelines do not currently recommend FDG PET-based imaging (PET/CT or PET/MRI) for primary breast cancer detection or for distinguishing benign from malignant breast lesions as a replacement for biopsy and pathological evaluation.^34–36

Initial Staging

For patents with newly diagnosed breast cancer, accurate clinical staging information is needed for prognosis and treatment planning. Anatomic stage includes the primary tumor size (T), regional lymph node status (N), and distant metastasis (M), using the American Joint Committee on Cancer TNM classification system.³⁷ Locoregional staging by imaging is conventionally achieved using mammography and ultrasound. Breast MRI can also be used for pre-operative staging, more accurately depicting primary tumor size and extent of disease and identifying otherwise occult areas of malignancy in both breasts.³⁸ For axillary nodal staging, sentinel lymph node biopsy and/or complete axillary lymph node dissection remain the gold standard. For patients with suspected distant metastases, systemic imaging with CT and bone scintigraphy or FDG PET/CT can be performed. Studies of PET/MRI have investigated its contribution in this pre-operative staging process.

For visualization of the index cancer, dedicated prone breast PET/MRI is more sensitive than whole-body supine PET/MRI.^39,40 In a study of 42 women with 48 invasive ductal carcinomas, Kong et al. reported a sensitivity of 100% (48/48) for dedicated prone simultaneous breast PET/MRI and 87.5% (42/48) for whole-body supine PET/MRI.³⁹ Sasaki et al. later showed that all 100 malignant lesions in 108 patients with primary breast cancer were visualized using diffusion-weighted MRI sequences and PET images from the dedicated prone breast PET/MRI exam.⁴⁰ However, 13% (13/100) of lesions were visually undetectable on the diffusion-weighted MRI sequences of whole-body PET/MRI and 7% (7/100) were visually undetectable on the PET sequence of whole-body PET/MRI.⁴⁰

For local tumor staging (Figure 2), breast PET/MRI is superior to conventional imaging and whole-body PET/CT and is at least equivalent to breast MRI alone. In a prospective study of 49 patients with biopsy-proven breast cancer, Grueneisen et al. found that breast PET/MRI and MRI alone yielded identical results for assessing local tumor extent (T stage).²⁶ However, breast PET/MRI was more accurate than whole-body PET/CT for local tumor staging (82% vs 68%).²⁶ A subsequent study by Goorts et al., which included 40 patients with breast cancer imaged prior to neoadjuvant chemotherapy, also found that breast PET/MRI was not of added value compared to MRI alone for determining T stage.⁴¹ Similarly in a retrospective study of 36 patients with invasive ductal carcinoma, Taneja et al. demonstrated that PET/MRI detected no additional satellite lesions that were not seen on MRI alone.⁴² However, diagnostic confidence was highest with fused PET/MRI compared to PET or MRI alone.⁴² In a retrospective study of 58 women with breast cancer by Botsikas et al., there was no significant difference in T-staging between PET/MRI and MRI alone.²⁵ However, there was a significant difference in T-staging using breast PET/MRI compared with conventional imaging and clinical assessment.²⁵ A meta-analysis of the diagnostic efficacy of PET/MRI for determining T stage reported pooled sensitivity, specificity, and AUC as 91%, 91%, and 0.96.⁴³

Figure 2:

Dedicated prone breast FDG PET/MRI demonstrating cT2cN1 clinical anatomic staging (mammography/ultrasound/palpation=cT2cN0). Right breast invasive ductal carcinoma (SUV_max=15.1;A). Right level I axillary lymph node (SUV_max=3.0;B). Pathologic stage=pT2pN1a.

For regional lymph node staging, there are differing results in the literature regarding the value of PET/MRI. Some studies show no added advantage of PET/MRI. The study by Grueneisen et al., which included 18 patients with positive ipsilateral axillary nodes, showed no significant difference between breast PET/MRI, MRI alone, and whole-body PET/CT for detection of axillary nodal metastases.²⁶ The sensitivity, specificity, positive predictive value, negative predictive value, and diagnostic accuracy were 78%, 90%, 82%, 88%, and 86% for PET/MRI; 67%, 87%, 75%, 82%, and 80% for MRI; and 78%, 94%, 88%, 88%, and 88% for PET/CT.²⁶ Similarly Botsikas et al. found comparable diagnostic performance of PET/MRI (whole-body supine and dedicated prone breast) and MRI alone for detecting axillary, internal mammary, and supraclavicular lymph node metastases (79% vs 88% sensitivity; 100% vs 98% specificity).²⁵ A systematic review concluded that PET/MRI has similar diagnostic performance as PET/CT for nodal staging.⁴⁴

However, other studies demonstrate benefit of PET/MRI specifically for axillary nodal staging. Van Nijnatten et al. developed a dedicated axillary PET/MRI protocol for evaluating lymph node status using a body coil over the ipsilateral shoulder in the prone position.¹⁴ In their feasibility study of 12 patients with clinically positive axillary disease, PET/MRI changed the clinical N stage in 40% compared to ultrasound, 40% compared to contrast-enhanced MRI, and 22% compared to supine PET/CT.¹⁴ In a prospective two-center study of 112 women with newly diagnosed breast cancer, Morawitz et al. demonstrated that supine PET/MRI of the thorax had the best diagnostic performance for detecting axillary nodal disease compared to supine MRI of the thorax, prone breast MRI, and axillary ultrasound.⁴⁵

For overall N staging (Figure 3), studies indicate that PET/MRI identifies regional nodal disease not detected on MRI alone or other conventional imaging. In the study by Goorts et al., nodal stage changed based on prone whole-body PET/MRI results in 6 of 40 patients initially considered clinically node-negative based on conventional imaging (mammography, ultrasound, and MRI).⁴¹ Furthermore, Taneja et al. found that combined PET/MRI increased the diagnostic confidence for assessing a lymph node as suspicious compared to PET or MRI alone.⁴² PET/MRI identified suspicious extra-axillary (internal mammary and supraclavicular) lymph nodes in 15 patients which were not suspected based on conventional staging.⁴² In a prospective two-center study of 104 women with newly diagnosed breast cancer, Bruckmann et al. demonstrated superior diagnostic performance of supine whole-body PET/MRI compared with MRI alone for assessing N stage.⁴⁶ Further work by Morawitz et al. demonstrated diagnostic superiority of supine whole-body PET/MRI over CT or MRI alone with the highest sensitivity in detecting lymph node metastases at all regional nodal stations (axillary levels I through III, supraclavicular, and internal mammary) in their prospective, two-center study of 182 patients with therapy-naïve invasive breast cancer.⁴⁷ A meta-analysis of the diagnostic efficacy of PET/MRI for determining N stage reported pooled sensitivity, specificity, and AUC of 94%, 90%, and 0.96.⁴³

Figure 3:

Dedicated prone breast FDG PET/MRI demonstrating cT3cN3b clinical anatomic staging (mammography/ultrasound/palpation=cT2cN1). Left internal mammary lymph node (SUV_max=6.9;A), level I axillary lymph node (SUV_max=14.4;B), left breast invasive ductal carcinoma (SUV_max=12.2;C). Pathologic stage=ypT3ypN2a.

Discovery of distant metastases during initial staging is significant since clinical management changes from curative intent to disease stabilization and palliation. Evaluation for distant metastases can be achieved via supine whole-body PET/MRI alone or in combination with dedicated prone breast PET/MRI for locoregional and systemic staging.¹⁵ Taneja et al. discovered distant metastases in 22% (8/36) of patients with invasive ductal carcinoma at initial presentation, thus changing management.⁴² Botsikas et al. identified two patients with metastatic disease in their study of 58 women (3.4%).²⁵ Two of the 40 patients (5%) in the study by Goorts et al. were found to have sternal bone metastases, which changed the treatment plan.⁴¹ In the study reported by Jena et al., distant metastases were identified in 5.8% (4/69) patients on whole-body PET/MRI.⁴⁸ For M staging, pooled sensitivity, specificity, and AUC of PET/MRI was 98%, 96%, and 0.99 in a meta-analysis.⁴³

Compared to whole-body PET/CT, most studies have demonstrated similar performance of whole-body PET/MRI for systemic staging. In a study by Pace et al., 36 breast cancer patients undergoing PET/CT for staging and follow-up were also imaged with simultaneous PET/MRI.⁷ The authors found equivalent performance for qualitative lesion detection between the two modalities.⁷ Likewise, Botsikas et al. demonstrated similar diagnostic performance, reader confidence, and inter-reader agreement in their study of 80 women undergoing PET/CT and PET/MRI for initial or recurrent breast cancer staging.⁴⁹ In a study of 30 patients with 242 distant metastatic lesions, Melsaether et al. found that whole-body PET/MRI detected metastatic disease at a similar rate as PET/CT with 50% of the radiation dose.⁵⁰ While these three studies yielded similar performance, improved staging of whole-body PET/MRI compared to PET/CT was demonstrated by Catalano et al. in their study of 51 women with newly diagnosed breast cancer.⁵¹

Improved performance of PET/MRI over PET/CT has been demonstrated for specific metastatic sites. In the aforementioned study by Botsikas et al., PET/MRI was more sensitive than PET/CT for osseous lesions, which is the most common site of distant metastases for breast cancer.⁴⁹ Similarly, Catalano et al. reported that PET/MRI detected a higher number of bone metastases compared to PET/CT in 25 breast cancer patients with osseous metastatic disease.⁵² Furthermore, PET/MRI correctly identified osseous metastatic disease in 3 patients (12%) which were missed by PET/CT.⁵² Bruckmann et al. further showed that whole-body PET/MRI is significantly better than CT and bone scintigraphy for diagnosing osseous metastases in patients with newly diagnosed breast cancer.⁵³ Melsaether et al. also demonstrated increased sensitivity of PET/MRI for detecting hepatic and possibly osseous metastases compared to PET/CT.⁵⁰ However, there was no significant difference in pulmonary metastasis detection between the two modalities.⁵⁰ It is important to note that data regarding the application of PET/MRI for systemic staging are from small, single-institutional studies which is a limitation in the level of scientific evidence. Future systematic evaluation in large multi-center studies is needed to determine whether the diagnostic performance of PET/MRI is significantly improved for distant staging compared to PET/CT.

Whole-body PET/MRI can also be performed in conjunction with a dedicated prone breast PET/MRI for comprehensive local tumor, regional nodal, and distant disease staging.¹⁵ In a prospective study of 38 patients with newly diagnosed breast cancer, Kirchner et al. compared the diagnostic performance of a one-step staging algorithm consisting of supine whole-body FDG PET/MRI with a two-step staging algorithm consisting of dedicated prone breast PET/MRI and supine whole-body PET/MRI.¹⁵ The authors demonstrated superiority of the two-step staging protocol for local and whole-body staging.¹⁵

Use of PET/MRI for initial staging of patients with breast cancer can impact clinical management. In the prospective study by Goorts et al., prone whole-body PET/MRI added value in 20% (8/40) of patients by changing the treatment plan in 10% and confirming malignancy of suspicious MRI lesions in 10%.⁴¹ For patients with newly diagnosed breast cancer and high pre-test probability for distant metastases, Kirchner et al. found that supine whole-body PET/MRI and dedicated prone breast PET/MRI led to a change in treatment compared to the conventional staging in 14% (8/56) of patients.⁵⁴ In the retrospective study by Taneja et al., whole-body PET/MRI changed management in 33% (12/36) of patients.⁴² Clinical management was directly impacted by PET/MRI in 12% (3/25) of patients in the study by Catalano et al. by adding radiation therapy, changing hormonal therapy, or introducing chemotherapy.⁵² Thus, the potential impact of PET/MRI on clinical management ranges from 12% to 33%.

Restaging

Systemic imaging may be performed when patients with a personal history of treated breast cancer present with suspected or biopsy-proven tumor recurrence to evaluate the extent and distribution of tumor burden for informing further treatment options. In a prospective study by Sawiki et al., 21 patients with suspected breast cancer recurrence underwent whole-body FDG PET/CT with iodinated contrast followed by whole-body PET/MRI with a gadolinium-based contrast agent. PET/MRI had the highest diagnostic performance compared with PET/CT, contrast-enhanced MRI, and contrast-enhanced CT.⁵⁵ Furthermore, interobserver agreement was improved with PET/MRI and PET/CT compared with MRI and CT alone.⁵⁵ In a similar prospective study by Grueneisen et al., PET/MRI provided higher accuracy for detection of recurrent breast cancer lesions in 36 patients compared to MRI alone.⁵⁶ Diagnostic accuracy of PET/MRI using a fluid-sensitive MRI sequence and either diffusion-weighted or post-contrast enhanced MRI sequences was comparable to PET/MRI using all three MRI sequences.⁵⁶ Omission of the post-contrast sequence or the diffusion-weighted sequence reduced the scan time without affecting diagnostic performance of PET/MRI.⁵⁶ A meta-analysis including 8 publications with 341 patients with recurrent, metastatic, and primary breast cancer found that the diagnostic accuracy of PET/MRI for staging/restaging was 0.96 (patient-level) and 0.95 (lesion-level).⁵⁷

Tumor Phenotyping and Prognosis

Tumor metabolic and functional parameters measured using simultaneous breast FDG PET/MRI (Table 1) have been shown to correlate with established clinical and histopathologic prognostic factors. Although non-invasive imaging methods for tumor phenotyping are unlikely to replace biopsy, it can provide supplementary whole-tumor, whole-breast, and whole-body assessment for initial therapy planning and prognosis and can be performed during therapy to assess changes in tumor biology over time. Phenotypes depicted by simultaneous FDG PET/MRI can also be used to formulate hypotheses for investigating the relationships between circulating biomarkers and tumor biology.⁵⁸

Table 1:

Quantitative PET/MRI Parameters Investigated for Breast Cancer

PET Parameters
	SUVmax	Maximum standardized uptake value
	SUVmean	Average standardized uptake value
	SUVpeak	Peak standardized uptake value determined using 1 cm3 volume sphere including the most intense uptake
	MTV	Metabolic tumor volume
	TLG	Total lesion glycolysis = MTV × SUVmean
MRI Parameters
Pharmacokinetic modeling of gadolinium-based contrast agents	Ktrans	Volume transfer constant between the extravascular extracellular space and plasma
	ve	Volume of extravascular extracellular space per unit volume of tissue
	Kep	Flux rate constant between the extravascular extracellular space and plasma = Ktrans / ve
Diffusion weighted imaging	ADCmean	Mean apparent diffusion coefficient
	ADCmin	Minimum apparent diffusion coefficient

Several studies have focused on tumor glucose metabolism represented by the standardized uptake value (SUV) from PET and tumor cellularity represented by the apparent diffusion coefficient (ADC) from diffusion-weighted MRI. In a study of 69 women with invasive ductal carcinoma, Jena et al. found that high SUV_mean, SUV_max, and total lesion glycolysis correlated with larger tumor size, higher tumor grade, higher Ki67 proliferation index, and triple-negative receptor status.⁴⁸ Sasaki et al. found in their study of 94 women that SUV_max correlated with HER2 positivity, higher Ki67 proliferation, larger tumor size, and higher nuclear grade; while ADC_mean correlated with lower nuclear grade.⁴⁰ However using multiple regression analysis, only tumor size was independently associated with SUV_max.⁴⁰ Nuclear grade and Ki67 proliferation were independently associated with ADC_mean.⁴⁰ In a study of 46 women with invasive ductal carcinoma, Kong et al. found that triple-negative receptor status and high Ki67 expression were associated with higher SUV_max values.⁵⁹ SUV_max and intratumoral metabolic heterogeneity negatively correlated with ADC_min and tumors with lymphovascular invasion had low ADC_min values.⁵⁹ Morawitz et al. similarly found a significant inverse correlation between SUV and ADC values in their study of 56 women with newly diagnosed breast cancer imaged with dedicated prone FDG breast PET/MRI.⁶⁰ SUV positively correlated with tumor grade and Ki67 proliferation but was inversely correlated with ER expression.⁶⁰ Research using advanced MRI diffusion-weighted imaging techniques, such as intravoxel incoherent motion analysis, has also been performed.⁶¹

Evidence also indicates that pharmacokinetic MRI parameters and FDG uptake measured using simultaneous breast PET/MRI may predict tumor receptor expression, molecular subtype, proliferation status, and metastatic potential. In a study of 50 patients with invasive ductal carcinoma, Incoronato et al. developed a multivariate model which included K_trans (volume transfer constant between blood plasma and interstitial space) from dynamic contrast-enhanced MRI and SUV_max from PET that correctly predicted the tumor molecular subtype (luminal A, luminal B, HER2, triple-negative) in 77.6% of patients (38/49).⁶² Previously, Catalano et al. found that PET/MRI biomarkers (Kep_mean, ADC_mean, and SUV_max) obtained from supine whole-body PET/MRI and pharmacokinetic analysis from a dedicated prone dynamic contrast-enhanced breast MRI predicted tumor molecular phenotypes in 62% (13/21) of patients with invasive ductal carcinoma.⁶³ In a study of 12 patients with newly diagnosed breast cancer imaged using simultaneous FDG PET/MRI, Margolis et al. found higher K_trans values in tumors with high Ki67 proliferation index.⁶⁴ They also reported lower k_ep (transfer constant from interstitial space to blood plasma) and higher metabolic tumor volume for patients with systemic metastases.⁶⁴ In a study of 46 patients with primary breast cancer imaged with dedicated prone breast PET/MRI, Inglese et al. reported good agreement between tumor perfusion parameters derived from two different MRI pharmacokinetic models that were significantly correlated with SUV.⁶⁵ An integrated approach of combining parameters from PET/MRI with molecular biomarker signatures could be more accurate than imaging alone for detecting distant metastases at initial presentation.⁶⁶ It is important to note that information provided by PET/MRI regarding tumor phenotyping and prognosis in these studies currently remains insufficient to consider clinical use as a reliable surrogate measure for pathologic subtype assessment.

Therapy Response Evaluation

Another potential use of PET/MRI is for evaluation and early prediction of neoadjuvant therapy response, supported by studies performing combined analyses of separately acquired whole-body PET/CT and breast MRI.^67,68 In a preliminary study by Romeo et al., the authors provide a descriptive report of several quantitative parameters regarding FDG uptake, diffusion restriction, and contrast pharmacokinetics obtained from simultaneously acquired breast PET/MRI for four patients with invasive breast cancer imaged before and after neoadjuvant therapy.⁶⁹ In another small study of nine patients, Jena et al. found that metabolic and MRI pharmacokinetic parameters obtained from simultaneous dedicated prone breast PET/MRI predicted conventional size-based tumor response in 9 of 11 lesions⁷⁰. They found maximal reductions in SUV_peak (90% decrease) and total lesion glycolysis (76% decrease) for the PET parameters and K_trans (65% decrease) for the MRI parameters.⁷⁰

Determining tumor non-response via imaging early during neoadjuvant therapy is important as it may allow a change in the type of therapy agent or prompt the decision to proceed with surgery. In meta-analyses of studies using PET or PET/CT, changes in FDG uptake of primary breast cancer after 1 to 2 cycles of neoadjuvant chemotherapy have been shown to be predictive of histopathological response.^71,72 Thus, PET/MRI would also be expected to reveal early information regarding neoadjuvant therapy response and there are a few small cohort studies published thus far focused on this application. In a study by Wang et al., fourteen women with newly diagnosed invasive breast cancer underwent simultaneous dedicated prone breast PET/MRI before neoadjuvant chemotherapy and after one or two cycles.⁷³ Combined PET/MRI parameters of change in SUV_max or change in total lesion glycolysis with changes in ADC_min predicted pathologic response more accurately than the individual PET and MRI parameters.⁷³ In a larger study by Cho et al., 26 patients with breast cancer were imaged using simultaneous dedicated prone breast PET/MRI before and after the first cycle of neoadjuvant chemotherapy.⁷⁴ The combination of changes in total lesion glycolysis and changes in the signal enhancement ratio from MRI resulted in improved sensitivity and specificity in pathologic response prediction.⁷⁴ Excellent interobserver reproducibility was demonstrated, which is critical for distinguishing true treatment response from inherent measurement variability.⁷⁴ Incorporation of deep learning techniques may further improve the accuracy of predicting neoadjuvant therapy response using PET/MRI.⁷⁵

Studies on the application of PET/MRI for therapy response evaluation published thus far are limited to patients with newly diagnosed primary breast cancer. To the best of our knowledge, there are no studies yet published on the use of whole-body PET/MRI for therapy response evaluation for patients with metastatic breast cancer. Future research is needed in this area, particularly for assessing therapy response for bone lesions which can be challenging using conventional imaging approaches.⁷⁶

Targeted Radiotracers and Technical Advances

Nearly all the published studies of PET/MRI for breast cancer used FDG for imaging glucose metabolism. Other radiopharmaceuticals exist that target well-established breast cancer biomarkers, such as estrogen receptor, progesterone receptor, human epidermal growth factor receptor 2, and proliferation.^{77–79 18}F-fluroestradiol (FES) binds to estrogen receptor and was recently approved by the United States Food and Drug Administration for patients with recurrent and metastatic breast cancer for whole-body detection of ER+ lesions to inform decisions regarding endocrine therapy. ¹⁸F-Fluorofuranylnorprogesterone (FFNP) binds to progesterone receptor and has been shown to predict response to endocrine therapy in patients with advanced/metastatic breast cancer.⁸⁰ There are ongoing clinical trials investigating FES breast PET/MRI for patients with ductal carcinoma in situ (NCT03703492) and FFNP breast PET/MRI for patients with primary invasive breast cancer (NCT03212170) (Table 2).

Table 2:

Active Studies of PET/MRI for Breast Cancer

NCT#	Title	Aim	# Patients
Glucose Metabolism Imaging with 18F-Fluorodeoxyglucose (FDG)
03510988	Dedicated Breast PET/MRI in Evaluation of Extent of Disease in Women with Newly Diagnosed Breast Cancer	To determine any incremental added benefit to breast MRI specificity by the addition of concurrent hybrid breast PET	147
01672021	Initial Assessment of FDG-PET/MRI in Determining the Extent of Systemic Disease in Breast Cancer Patients	To assess the ability of FDG-PET/MR imaging to detect systemic disease in breast cancer patients as compared with conventional FDG-PET/CT	80
04829643	Targeting the Future of Axillary Staging in Early Breast Cancer: A Comparative Study: Sentinel Node Biopsy vs PET/MRI	To compare the staging power between sentinel node biopsy and PET/MRI in detecting axillary lymph node macrometastases	247
04826211	Targeting the Future of Axillary Staging in Node Positive Breast Cancer Patients Receiving Primary Systemic Therapy. A Comparative Study Between Sentinel Node Biopsy vs PET/MRI	To compare the staging power between sentinel node biopsy or lymphadenectomy vs PET/MRI	110
03374826	Non-invasive Axillary Lymph Node Staging in Breast Cancer With PET-MRI	To determine the accuracy of dedicated axillary hybrid PET/MRI to exclude axillary lymph node metastases in clinically node negative patients	125
04273555	Monitoring Early Response to Targeted Therapy in Stage IV Human Epidermal Growth Factor Receptor 2 Positive Breast Cancer Patients with Advanced PET/MRI	To evaluate PET/MRI for monitoring treatment response	20
Hypoxia Imaging with 18F-Fluoromisonidazole (FMISO)
04332588	Monitoring HER2+ Breast Cancer Neoadjuvant Treatment with Advanced PET/MRI	To evaluate PET/MRI with FMISO for monitoring and predicting the effect of trastuzumab (Herceptin) on chemotherapy in patients diagnosed with advanced HER2 positive breast cancer	25
04861077	Monitoring Breast Cancer Immunotherapy Treatment with Advanced PET/MRI: A Pilot Study	To investigate the utility of FMISO PET/MRI in patients with triple negative stage II-IV breast cancer to monitor and predict the effect of immunotherapy	20
Estrogen Receptor Imaging with 18F-Fluoroestradiol (FES)
03703492	PET/MRI of Estrogen Receptor Expression in Non-Invasive Breast Cancer	To compare quantitative FES uptake of biopsy-proven ductal carcinoma in situ measured using PET/MRI with ER protein levels determined by immunohistochemistry	34
Progesterone Receptor Imaging with 18F-Fluorofuranylnorprogesterone (FFNP)
03212170	FFNP PET/MR Imaging of Progesterone Receptor Expression in Invasive Breast Cancer	To test the accuracy of PET/MRI imaging with FFNP for measuring progesterone receptor expression in patients with invasive breast cancer	28
HER2 Imaging with 89Zr-Df-Trastuzumab
03321045	PET Imaging With 89Zr-Trastuzumab for Prediction of HER2 Targeted Therapy Effectiveness	To measure the diagnostic quality (with standardized uptake values) of PET/MRI imaging with 89Zr-Df-Trastuzumab in patients with newly diagnosed breast cancer	10
Gastrin-Releasing Peptide Receptor (GRPR) Imaging with Gallium 68-labeled GRPR Antagonist BAY86–7548
03831711	A Pilot Study of 68Ga-RM2 PET/MRI in the Evaluation of Patients with Estrogen Receptor-Positive Breast Cancer	To evaluate the feasibility of 68Ga-RM2 PET/MRI for identification of estrogen receptor positive primary breast cancer and metastases	20

Registered on ClinicalTrials.gov, June 202

¹⁸F-Fluoromisonidazole (FMISO) is an investigational radiopharmaceutical for imaging tumor hypoxia, a common feature of solid tumors associated with poor prognosis and treatment resistance.⁸¹ A pilot study of nine women with invasive ductal carcinoma by Andrzejewski et al. demonstrated feasibility of separately acquired breast MRI and PET/CT with FDG and FMISO.⁸² The study concluded that combined imaging data may provide complementary prognostic information regarding tumor recurrence, distant metastases, and breast cancer specific death.⁸² In a larger study of 29 women with primary invasive breast cancer, Carmona-Bozo et al. demonstrated using simultaneous breast PET/MRI that FMISO influx on PET inversely correlated with K_trans from MRI, supporting the hypothesis of perfusion-driven hypoxia in breast cancer.⁸³ The authors emphasize the importance of simultaneous multimodality imaging due to the temporally variant and heterogeneous spatial relationships that exist between tumor hypoxia and perfusion.⁸³ Further studies using FMISO PET/MRI are ongoing (Table 2).

In addition to new tracers, several technical advances are expected to expand the potential opportunities of PET/MRI for breast cancer. Minimizing radiation exposure to the patient through dose reduction is an important initiative. PET images simulating the effect of one-third⁹ and one-tenth⁸⁴ the standard FDG dose have demonstrated clinically acceptable image quality. The effective dose-equivalent of using one-tenth the standard FDG injected activity (1 mCi) is 0.45 mSv, comparable to the effective dose of digital mammography (0.44–0.56 mSv) and may allow for additional clinical indications such as high-risk screening. Prospective studies confirming the diagnostic performance of low-dose FDG PET/MRI are needed.

Given the multiple sequences resulting from PET/MRI, automated analytic tools to assist radiologists may be helpful. Vogl et al. developed an automated computer aided diagnosis system for PET/MRI based on data from the FDG PET, dynamic contrast-enhanced MRI, and diffusion-weighted MRI sequences.⁸⁵ The system was developed using separately acquired FDG PET/CT and breast MRI from 34 patients (10 benign, 22 malignant lesions) for automatic lesion detection, segmentation, and classification as benign or malignant.⁸⁵ The development of accurate and reproducible automated methods⁸⁶ for defining the tumor region for quantitative analyses will facilitate use of breast PET/MRI for treatment response evaluation by eliminating time required for manual definition. Automated tools for quantitative analysis may also reveal information regarding intratumoral heterogeneity. Using a new automated voxel-based analysis approach with simultaneous FDG breast PET/MRI, Schmitz et al. demonstrated the coexistence of different phenotypes defined by ADC and SUV maps within the same tumor that spatially correlated with histologic findings.⁸⁷ To the best of our knowledge, these types of automated tools are not yet clinically available.

Radiomics is a computational approach for extracting large amounts of image features that may uncover tumor characteristics not appreciated by visual analysis. For prone FDG PET/CT, Krajnc et al. recently demonstrated in a study of 170 patients that application of radiomics and machine learning models together with advanced data pre-processing improved the diagnostic accuracy for differentiating between benign and malignant breast lesions and for identifying aggressive triple negative molecular subtypes.⁸⁸ Thus, combining radiomic features from both PET images and breast MRI may likewise improve individual predictive models. Huang et al. investigated the association of image features from separately acquired FDG PET/CT and dynamic contrast-enhanced breast MRI with tumor grade and recurrence-free survival in a retrospective study of 113 patients with breast cancer.⁸⁹ Key textural features for predicting recurrence-free survival were the inverse difference moment normalized feature from MRI and the cluster prominence feature from PET.⁸⁹ In a retrospective study of 124 patients with breast cancer imaged using dedicated prone simultaneous FDG PET/MRI, Umutlu et al. demonstrated accuracy of a multiparametric radiomics predictive model for determining molecular subtype, hormone receptor status, proliferation index, lymph node status, and distant metastases.⁹⁰ Independent, multi-center validation of PET/MRI radiomics models following standardized guidelines are needed.

New research has recently been reported incorporating artificial intelligence and machine learning algorithms with simultaneous multiparametric FDG PET/MRI. In a study by Romeo et al. of 102 patients with 120 suspicious breast lesions, the artificial intelligence model with the best diagnostic accuracy (94.8%) for discriminating between benign and malignant breast lesions was based on combined tumor cellularity, permeability, and glucose metabolism data obtained from dedicated prone simultaneous FDG PET/MRI.⁹¹ Although, there was no significant difference in overall diagnostic performance compared to clinical interpretation by expert readers, higher specificity was achieved using the artificial intelligence model supporting further potential to reduce false-positive findings and biopsy recommendations.

Contraindications

Contraindications for PET/MRI mostly include those for each individual imaging modality. Contrast-enhanced MRI is not performed for patients who are pregnant, are allergic to gadolinium-based contrast agents, have severe renal insufficiency, have severe claustrophobia, or have MRI incompatible implantable devices or ferromagnetic foreign bodies.¹⁹ PET imaging is generally not performed for patients who are pregnant.⁹² Specifically for breast PET/MRI, elevated body mass index (e.g., greater than 36 kg/m²)⁹ may limit prone patient positioning with the breast coil due to the smaller bore opening (60 cm for integrated PET/MRI scanners versus 70 cm for typical PET/CT and wide bore MRI scanners).

Conclusion

PET/MRI for breast cancer imaging has been studied across the spectrum of diagnosis, staging, prognosis, tumor phenotyping, and treatment response assessment. PET/MRI provides a comprehensive evaluation for patients with newly diagnosed breast cancer with a clinical need for determining the extent of disease in the breast and locoregional lymph nodes as well as for systemic staging at initial diagnosis or recurrence. Importantly, PET/MRI has been shown to change clinical management. Compared to PET/CT, there is significantly reduced radiation exposure to the patient with PET/MRI and thus should be considered where available for patients with clinical indications for both PET/CT and organ-specific MRI exams. Based on existing evidence, PET/MRI appears to be superior to PET/CT for detecting unsuspected extra-axillary nodal and distant disease, particularly hepatic and osseous metastases, in patients with breast cancer.⁹³ However, PET/MRI scanners are currently less widely available compared to PET/CT which limits accessibility.^94,95 Continued research using targeted radiopharmaceuticals, radiomics, and methods to decrease radiation dose by reducing the injected activity of radiopharmaceutical will help expand the use of PET/MRI for breast cancer imaging in routine clinical settings.

Search Strategy and Selection Criteria

An electronic literature search was performed using PubMed to identify potential studies published in English until June 2021. The search terms “positron emission tomography”, “magnetic resonance imaging”, and “breast cancer” were used. Relevant studies were retrieved, and their references were reviewed to identify any additional studies. Articles relevant to the scope of this Review were included.