Automated Tools to Analyze FES PET: Translational Potential to Guide Therapy in Estrogen Receptor Positive Breast Cancer

See also article by Miller et al.

Austin R. Pantel, MD, MSTR, is an assistant professor of radiology at the University of Pennsylvania. Dr Pantel’s research focuses on the translation of emerging PET imaging probes and instrumentation. He recently received a Radiological Society of North America Research Scholar Grant and is an associate editor for Radiology: Imaging Cancer.

Sophia R. O’Brien, MD, MSEd, is an assistant professor of clinical radiology at the University of Pennsylvania. She serves as an attending radiologist in both the Division of Nuclear Medicine Imaging and Therapy and the Division of Breast Imaging, having completed fellowships in both subspecialties. Dr O’Brien’s research focuses on fluoroestradiol PET/CT, breast imaging, and radiology education. She completed her master’s of medical education at Penn in 2024 and is an associate program director for the Radiology Residency Program at Penn.

The recent introduction of the radiolabeled PET analogue fluorine 18 fluorestradiol (FES) into nuclear medicine has transformed imaging in patients with metastatic estrogen receptor (ER) positive breast cancer. The cell surface receptor profile of breast cancer—comprising ER, progesterone receptor, and human epidermal growth factor receptor 2—is routinely assessed at histology during standard diagnostic workup of patients with breast cancer, with ER positive breast cancers accounting for approximately 80% of all new diagnoses. Expression of these markers guides treatment decisions, enabling use of targeted therapies when expressed, or chemotherapy in their absence (eg, triple-negative breast cancer). Indeed, ER-targeted therapy is associated with improved survival in patients with ER positive breast cancer. As a PET imaging agent that binds to the ER, FES enables noninvasive assessment of ER expression and binding across all disease sites, providing an additional tool to inform diagnosis and treatment planning in patients with metastatic ER positive breast cancer (1).

The Food and Drug Administration approval of FES in the United States in 2020 marked the culmination of decades of research validating its use as a noninvasive marker of ER expression, with many clinical applications. An image of an FES-avid breast cancer and an ipsilateral chest wall metastasis won Image of the Year at the Society of Nuclear Medicine Annual Meeting in 1987, marking a formative moment in molecular imaging as the first PET image of a cancer receptor (2). The following year, the seminal manuscript followed that demonstrated receptor-specific uptake of FES correlated with ER levels as determined by radioligand binding assay (3). More recent studies established a correlation of FES uptake with ER expression measured by immunohistochemistry, the current reference standard, further establishing FES as an in vivo marker of ER expression (1).

Building on these foundational studies, additional work has demonstrated the clinical utility of FES PET/CT in select patient populations. In a study of 47 patients with recurrent or metastatic ER positive breast cancer, none of the 15 patients with FES standardized uptake values (SUVs) below 1.5 responded to salvage endocrine therapy, suggesting no benefit from ER-targeted therapy when receptor expression is low or absent at imaging. If FES PET/CT had been used to guide treatment selection in this study sample (ie, withholding ER-targeted therapy in patients lacking FES uptake), the response rate to endocrine therapy would have increased from 23% to 34% (4). Subsequent work from other groups further supported the role of FES PET/CT as a biomarker to inform treatment decisions.

The Society of Nuclear Medicine and Molecular Imaging (SNMMI) published Appropriate Use Criteria for FES PET/CT in 2023 (5). Assessing ER status with FES PET/CT in lesions that are difficult to biopsy or when biopsy is nondiagnostic was deemed “appropriate,” aligning with the provided indication in the FES package insert, which supports its use as an adjunct to biopsy in patients with recurrence of metastatic ER positive breast cancer. Additionally, biomarker applications using FES PET/CT to assist with treatment selection for endocrine therapy at initial diagnosis or progression of metastatic ER positive disease also received an "appropriate" designation, further supported by a recent systematic review and meta-analysis (6). Use of FES PET/CT for staging invasive lobular carcinoma (ILC) and low-grade invasive ductal carcinoma received a score of 5 of 9, indicating that it “may be appropriate.” These cancers, characterized by insidious growth not well visualized at anatomic imaging (mammography, breast MRI, breast US, CT) and low glycolytic activity limiting detection at fluorodeoxyglucose (FDG) PET, have historically been challenging to stage given their underlying biology.

In an article published in Radiology: Imaging Cancer, Miller et al (7) evaluated two automated tools for analyzing FES PET imaging. The first tool utilizes a deep learning model with a Retina U-Net architecture to detect FES-avid lesions. These artificial intelligence (AI)–identified lesions are used along with SUV_max thresholds to classify patients as having, likely having, or not having FES-positive disease. The second tool assesses concordance between lesions identified by a nuclear medicine physician at FES PET and lesions identified at either CT or FDG PET/CT. These two tools were tested as part of a secondary analysis on FES PET scans obtained from a prior prospective clinical trial comparing FES PET/CT to standard-of-care FDG PET/CT or CT in two cohorts: women with newly diagnosed stage 2 or 3 ER positive breast cancer undergoing imaging to evaluate for metastatic disease (arm 1), and previously treated patients with breast cancer undergoing imaging for suspected recurrence (arm 2). The AI tool demonstrated a sensitivity of 90% for classifying patients as having FES-avid metastases (SUV_max > 1.5) and a specificity of 55%. In a subset of 25 participants selected for concordance analysis using the second automated tool, disease heterogeneity, defined as the presence of at least one lesion identified with one modality but not the other, was observed in 12 of 13 participants when comparing FES PET/CT and FDG PET/CT and in 11 of 13 participants when comparing FES PET/CT and diagnostic CT. One participant had no metastases detected at FES PET/CT. Notably, three of the four patients with ILC had the majority of their disease burden (>70%) visualized only at FES PET/CT.

From a technical perspective, these tools perform as intended, successfully identifying sites of FES-avid disease. The next challenge lies in translating these tools into clinical practice to better inform treatment decisions—and ultimately improve outcomes for patients with ER positive breast cancer. As nuclear medicine physicians and radiologists are well aware, PET scans often reveal extensive disease burden; however, only representative lesions are typically described in clinical reports, constrained by both time pressures of modern practice, and, more importantly, limited clinical utility of describing every finding. Formalized response criteria, such as RECIST (Response Evaluation Criteria in Solid Tumors) and PERCIST (PET Response Criteria in Solid Tumors), standardize reporting by specifying the number of target lesions needed to assess treatment response. This raises an important question: How might comprehensive identification of FES-avid disease add clinical value? One promising application, as demonstrated in this study, is leveraging these tools to assess concordance or discordance between FES PET/CT and FDG PET/CT.

The potential of combined FES and FDG PET/CT to serve as a biomarker for predicting response to endocrine-targeted therapy in ER positive breast cancer has long been recognized. While FES PET measures ER expression, FDG PET measures glycolytic activity, a surrogate marker of tumor aggressiveness. In a study of women with ER positive breast cancer, combining data from both imaging modalities stratified patients into three groups with distinct responses to endocrine therapy. Patients with low FDG uptake, indicating an indolent tumor phenotype, had a relatively long progression-free survival (PFS) of 26.1 months. Among patients with FDG-avid tumors, those with high FES uptake had longer PFS compared with those with low FES avidity (7.9 vs 3.3 months, respectively). In this analysis, up to three lesions per patient were measured on the FES PET scan (8). A more recent study further explored this concept, analyzing the entire disease burden on both FES and FDG PET scans in patients treated with a novel selective ER downregulator rintodestrant. A total of 1051 lesions were analyzed across 16 patients to create a heterogeneity score as a metric of imaging discordance. Patients with low heterogeneity had a relatively prolonged PFS (>5 months), while those with 100% heterogeneity—defined as FDG uptake without FES uptake—had a short PFS of less than 2 months (9). Such data can be used to tailor therapy treatment—including foregoing hormonal therapy if the ER receptor does not bind FES or escalating treatment in the setting of glycolytic, FDG-avid disease. The automated tools evaluated by Miller et al may prove valuable in this setting. An AI-driven tool that can reliably assess and quantify the entire disease burden could facilitate heterogeneity analyses, reducing the need for tedious, error-prone human analysis. Combined with a growing body of evidence supporting dual-tracer approaches, such tools provide further impetus for developing methods to enable imaging two fluorinated radiotracers in a single imaging session.

Although not a focus of the secondary analysis presented here, it is noteworthy that the primary analysis of the original prospective clinical trial (10) suggested improved disease detection with FES PET/CT over standard of care imaging with CT, bone scan, or FDG PET in patients with ILC. However, the ILC cohort was small, and this finding did not reach statistical significance. Consistent with this finding, in three of the four participants with ILC included in the concordance analysis, the majority of metastatic lesions were observed only at FES PET/CT. Reflecting this emerging evidence, the SNMMI Appropriate Use Criteria currently deem that FES PET/CT "may be appropriate" for staging ILC. Future studies are warranted to clarify and validate the role of FES PET/CT in this challenging breast cancer subtype.

This study by Miller et al applies a cutting-edge AI-driven analysis tool to FES PET, which itself is a relatively recent addition to the imaging armamentarium. As use of FES PET continues to expand, such automated tools may enhance its clinical potential.