Imaging approaches to assess the therapeutic response of gastroenteropancreatic neuroendocrine tumors (GEP-NETs): current perspectives and future trends of an exciting field in development

Morphological and functional imaging techniques are crucial for characterizing and managing GEP-NETs in clinical practice. Various different imaging modalities are used for screening at-risk populations, detecting primary lesions, assessing the extent of the disease, and evaluating the patient’s response to treatment [5, 9‐11]. The choice of which techniques to use depends on clinical presentation and specific tumor features, including location, somatostatin receptor expression, functionality, and proliferation rate [10]. Currently, GEP-NETs are morphologically assessed by computed tomography (CT) and magnetic resonance imaging (MRI). Additionally, various ultrasound (US) approaches (transabdominal, endoscopic, and intraoperative) can be useful for detecting certain primary tumors (e.g., pancreatic) or liver metastasis (Table 1).

CT is the most commonly used technique for initial tumor localization, staging, therapeutic monitoring, and follow-up of patients with GEP-NETs. Its rapid acquisition process and ability to generate multiplanar reconstructions provide high temporal and spatial resolution (20–200 μm; pixel/voxel dimension <1 mm), which improve the probability of detection (Fig. 1). Curved reformats, three-dimensional volume rendering techniques and maximum intensity projection provide important anatomic details for surgical planning (e.g., tumor vascular encasement). CT scan also effectively detects nodal and metastatic disease. The majority of GEP-NETs are visible as enhanced or hypervascular lesions and are typically more conspicuous in the late arterial acquisition phases [12]. On precontrast images, these lesions are typically hypoattenuating, although some primary lesions may be small and have the same density as normal parenchyma and thus may be difficult to detect. It is therefore crucial to perform multiphasic contrast-enhanced CT, including arterial and portal venous acquisition phases. Indeed, the differences in the time elapsed after contrast administration may substantially affect image acquisition, potentially leading to both false-positive or false-negative findings that can result in an erroneous detection or response assessment. Particular caution is recommended regarding clinical decisions based on the emergence or fading of a single small lesion. The sensitivity of multidetector CT for primary pancreatic GEP-NETs ranges from 57 to 63 % and may be as high as 94 % if CT slices are reformatted in thin sections (1–1.5 mm). This detection rate substantially decreases, however, for primary extrapancreatic tumors, particularly those located in the small bowel. Some of these tumors may be only visible when mesentery involvement induces surrounding fibrosis (desmoplastic reaction) with tissue retraction. More recently, a novel modality, dual-energy CT (DECT), is an emerging technique based on imaging at two distinctly different energy levels (e.g., 80 and 140 kVp) to differentiate materials. DECT can provide multiple parameters, including monochromatic CT images, iodine-based material decomposition images and spectral HU curves, and virtual unenhanced images, which may add useful tissue information for detecting GEP-NETs [13, 14].

MRI has also high spatial resolution (10–500 μm; pixel/voxel dimension >1 mm) and offers several advantages over CT scans, including the lack of ionizing radiation, superior soft tissue discrimination, and greater interobserver agreement. In addition, MRI offers the advantage of combining anatomical and functional or molecular imaging techniques, including diffusion-weighted imaging (DWI), dynamic contrast-enhanced MRI (DCE-MRI), and magnetic resonance spectroscopic imaging (MRSI). NETs are typically best observed in fat-suppressed T1-weighted sequences, such as low signal intensity lesions, and typically become hyperintense on T2-weighted sequences. Intravenous contrast enhancement is also essential for adequate characterization. Improved lesion detection and characterization have been documented with MRI for certain anatomic locations, such as the liver or pancreas. MRI is the best conventional imaging technique to detect hepatic metastases in GEP-NETs (sensitivity of 95 %) and shows a good performance for the detection of primary pancreatic NETs (sensitivity 74–94 %, specificity 78–100 %), although it is less useful than CT for the detection of primary small bowel lesions [15, 16]. However, MRI is still more expensive and time-consuming than CT and requires greater patient cooperation. These factors limit the widespread use of MRI, which is generally used as a problem-solving tool in patients with negative or equivocal findings resulting from other diagnostic procedures [9‐11].

Transabdominal US is an inexpensive technique that may be used to screen solid organs in the abdomen or to direct needle biopsy for histological diagnosis (spatial resolution 50–100 μm). GEP-NETs are typically visualized as hypoechoic, well-defined masses commonly surrounded by a hyperechoic halo. The sensitivity of US for detecting GEP-NETs is, nevertheless, limited, ranging from 15 to 80 % depending on tumor size and anatomical localization [5, 9‐11]. Its utility is further limited in patients with abundant abdominal gas or in very large/fat patients, as sound waves are attenuated as they pass deeper into the body. Improved image resolution can be achieved using a high-frequency endoscopic US (EUS) probe (7.5–12 MHz) that enables the transducer to maintain close proximity to target lesions. The sensitivity of this probe is substantially higher than the standard US probe (3–5 MHz) and is especially helpful for detecting small lesions in the pancreatic head or duodenal wall [17]. In fact, EUS sensitivity is higher than that of the CT scan in this context (92 vs. 63 %), particularly for detecting small insulinomas (84 vs. 32 %) [18]. EUS is, nonetheless, invasive, technically difficult, operator-dependent, and not widely available [5, 9‐11]. Finally, intraoperative US (IOUS) may be an important aid in the surgical exploration of the pancreas and liver.

Finally, multimodal imaging is able to provide a combination of anatomical, molecular, and functional imaging quantitative parameters of tumor phenotype. In modern scanners, positron emission tomography (PET) and single photon emission computed tomography (SPECT) are combined with an anatomical cross-sectional counterpart such as CT (PET/CT and SPECT/CT) and also MRI (PET/MRI) [19]. These hybrid techniques have demonstrated to be useful of GEP-NETs for tumor detection, staging, and tumor response evaluation [20].

The first widely adopted criteria for evaluating cancer therapies were developed by the World Health Organization (WHO) in 1979 and were primarily intended for use in clinical trials that had a tumor response as the primary endpoint [21]. The WHO criteria were the first set of rules to introduce the concept of overall assessment of tumor burden based on the sum of the bidimensional products of tumor lesions. In addition, they defined the response to therapy based on the percentual change from baseline. This standardization facilitated a common universal language for reporting the results of cancer therapy in a consistent manner to enable cross-trial comparisons. Numerous modifications of the WHO criteria were subsequently made to clarify uncertain issues in the original document and to accommodate emerging new technologies such as CT and MRI. These revisions led, paradoxically, to the response criteria being no longer comparable among research organizations. An International Working Party was thereby constituted to simplify and standardize again the evaluation criteria that resulted in the Response Evaluation Criteria In Solid Tumors (RECIST, version 1.0) [3]. Key features of these updated criteria included the use of one-dimensional measurements for assessing tumor burden, the definition of the minimum size of measurable lesions to be selected as targets, instructions on how many lesions to follow and how, and a model for the overall response assessment combining both target and nontarget lesions [3, 4]. In addition, time to progression and PFS, which can be assessed in all patients, as opposed to duration of response that can only be quantified in responders, were briefly discussed as alternative endpoints in certain circumstances (i.e., investigation of noncytoreductive anticancer agents) [7, 8].

An updated version of RECIST (version 1.1) was published in 2009 [4] to address several questions and pending issues, including how to apply RECIST in trials in which progression, rather than response, is the primary endpoint, or in trials of targeted noncytotoxic drugs. Whether and how to use newer imaging technologies, such as ¹⁸F-fluorodeoxyglucose positron emission tomography (¹⁸F-FDG-PET) and MRI, was also addressed. A summary of the evolution from WHO response criteria to RECIST 1.1 is presented in Table 2.

The RECIST criteria have been widely adopted by academic institutions, cooperative groups, and industry as a standard method for reporting clinical trials. The primary strengths of these criteria are simplicity, reproducibility, and universal acceptance, which allow outcomes across different trials to be compared. Nevertheless, a number of limitations remain, and several unique features of tumor biology and specific cancer therapies make their applicability particularly challenging in the field of GEP-NETs [22, 23].

RECIST dichotomize patients into responders versus nonresponders to quantify drug efficacy, the latter category including both stabilization of disease (which in certain scenarios may be actually indicative of a drug-induced antiproliferative effect) and disease progression. Even the latest version of RECIST primarily focuses on the use of objective response endpoints for phase II trials [24]. In the context of GEP-NETs, however, this may be only applicable to poorly differentiated, highly proliferative tumors that are treated with cytotoxic therapy. As the majority of GEP-NETs are well-differentiated, slow-growing tumors, and novel targeted agents are increasingly being used to treat this disease, alternative definitions of tumor response are being actively explored and greater emphasis is being placed on progression-based endpoints [23].

Other means for assessing treatment effects, such as a minor response (tumor shrinkage <30 %), disease control rates (a combination of objective response and stable disease), or the proportion of patients that are progression-free at landmark time points, could be considered as alternative metrics for providing early indications of an agent deserving additional clinical development. A major limitation to this approach in the field of GEP-NETs is to accurately estimate the expected disease stabilization rate, in the absence of a treatment effect, in this generally indolent disease. For this reason, a randomized controlled design is preferred, when feasible, for phase II screening trials in this context.

Several recent studies have illustrated how alternative metrics may more accurately reflect the therapeutic effect of medications in well-differentiated GEP-NETS. Indeed, two somatostatin analogs (octreotide and lanreotide) and two targeted agents (everolimus and sunitinib) significantly improve PFS of patients with different subtypes of GEP-NETs with little or no effect on tumor volume (objective response rate <10 % by RECIST) [5‐8]. Waterfall plots of pivotal randomized trials, such as RADIANT-3 to evaluate everolimus in pancreatic NETs, showed that the proportion of patients achieving any degree of tumor shrinkage was substantially greater in everolimus-treated patients (64 %) than in patients in the placebo control arm (21 %), although response rates were low in both treatment arms (5 vs. 2 %) [8]. The current thresholds used to define tumor progression (≥20 %), which have become more conservative over time, have also been questioned in the context of slow-growing GEP-NETs. However, caution should be advised in lowering this threshold, as validation would be required to ensure that the measurement errors are actually inferior to the cutoff values, and this is not an easy task due to the technical difficulties frequently encountered in the precise definition of lesions in NETs.

Other clinical settings that question the validity of RECIST to address treatment failure include clinical or biochemical progression in the absence of radiological progression, focal progression that is amenable to local therapy (e.g., small bowel obstruction with stable liver metastasis), or indolent asymptomatic progression. Some technical difficulties inherent to identifying and monitoring metastases in certain patients with GEP-NETs may also limit the application of RECIST. These include patients with small volume metastatic disease or, conversely, extensive liver involvement with either multiple small or large confluent liver metastases, which can form conglomerate masses that may be difficult to individualize and monitor. In addition, certain targeted agents (e.g., angiogenesis inhibitors) may induce necrosis or cystic changes in the tumor that are not only not associated with tumor shrinkage but may even render preexisting lesions more visible, which may be misleading and erroneously interpreted as progressive disease rather than as a positive therapeutic effect. Finally, residual masses may not be adequately differentiated from fibrosis, with no viable tumor, by means of standard imaging modalities, and would never be considered as complete responses by RECIST [23].

Beyond size, additional radiological parameters can also provide very valuable information in terms of antitumor efficacy and are increasingly being considered when evaluating response in certain tumor types treated with specific targeted agents. Indeed, necrosis, hemorrhage, and myxoid degeneration may reflect pathologic tumor response in the absence of significant tumor shrinkage. One of the most representative examples of this effect was soon documented in gastrointestinal stromal tumors (GIST) after the introduction of imatinib. The dramatic changes in tumor density induced by this drug, as determined by measuring CT attenuation coefficient in Hounsfield units (HU), led Choi to propose in 2007 a new set of criteria that combined changes in both size and density for tumor response assessment in this setting [25] (Table 3). These criteria and their variants are now widely applied in GIST and are being prospectively evaluated in other neoplastic diseases. Preliminary data in patients with pancreatic NETs suggest that Choi criteria may help to early discriminate patients who might benefit from sunitinib or everolimus therapy [26].

Other alternative methods to evaluate tumor response (Table 3) have been described in detail elsewhere [27‐30] and are beyond the scope of this review. Briefly, these methods take into consideration additional parameters such as arterial enhancement (e.g., the European Association for Study of the Liver (EASL) criteria in hepatocellular carcinoma) [27], treatment-induced tumor necrosis, or other structural changes (e.g., the Morphology, Attenuation, Size and Structure (MASS) criteria for renal cancer treated with antiangiogenic agents, or the Lee criteria for nonsmall cell lung cancer treated with EGFR inhibitors) [31, 32]. Specific criteria have also been developed for novel therapeutic strategies (e.g., immune-related response criteria) [33] or imaging modalities such as PET (e.g., PET response criteria in solid tumors or PERCIST) [30, 34, 35].

¹⁸FDG-PET assesses tumor glucose uptake, which broadly correlates with cancer cell viability. ¹⁸FDG uptake may therefore provide an early and sensitive pharmacodynamic marker for monitoring response to antiproliferative agents [36]. Changes in ¹⁸FDG uptake measured by PET have been correlated with a pathological tumor response and overall survival in certain neoplastic diseases. Caution should be advised, nevertheless, as ¹⁸FDG-PET is a sensitive but nonspecific method for detecting malignancy sites. Areas of active inflammation or infection are common sources of false-positive findings [37]. A period of at least 10 days (or preferably up to 3 weeks) is advised before a PET scan is performed following chemotherapy administration, to bypass transient fluctuations of ¹⁸FDG uptake, either stunning or flare of tumor uptake, which may occur soon after treatment. False-negative findings may also arise when evaluating small lesions. PET imaging is progressively being incorporated for staging and response assessment in a number of malignancies (e.g., lymphoma) [38]. In the context of GEP-NETs, however, it would be only applicable to highly proliferative or poorly differentiated tumors, as guidelines have been specifically developed for the ¹⁸FDG radionuclide. Whether these criteria may be applied to other radiotracers (e.g., ⁶⁸Gadolinium-tetraazacyclododecane tetraacetic acid (⁶⁸Ga-DOTA) peptides) remains to be elucidated. Nevertheless, prospective trials for validating these response criteria are warranted.

Functional imaging refers to the visualization, characterization, and quantification of biological processes at the cellular or molecular level. The unique features of GEP-NETs provide distinct targets for cancer-specific functional imaging, such as somatostatin receptors (SSTR), or catecholamine transporter and synthesis pathways. Consequently, an increasing number of radiopharmaceuticals are becoming available to detect and quantify different aspects of the heterogeneous biology of GEP-NETs (Table 1). Several techniques that are increasingly being used in clinical practice will be discussed below.

As treatment options for GEP-NETs continue to expand, evaluation of therapeutic response solely on the basis of size clearly has substantial limitations, particularly in the era of targeted therapy. Indeed, new therapeutic modalities, such as angiogenesis inhibitors, may significantly increase patients’ survival by inducing a cytostatic effect that does not necessarily translate into tumor shrinkage (e.g., tumor necrosis or cavitation with no size reduction). Functional imaging techniques that monitor specific physiological and cellular processes within the tumors in response to antineoplastic agents with novel mechanisms of action are therefore gaining momentum [22, 70‐74]. In fact, these alternative response assessment techniques are increasingly being implemented at earlier stages of drug development. The main features of these imaging modalities are summarized in Table 1.

DCE imaging methods may be applied to US, CT, or MRI with contrast dyes specific to each modality. These imaging techniques assess tumors based on the appreciable differences between the heterogeneous, chaotic, and leaky neoplastic vascular network and the normal physiological vasculature of healthy tissues [22, 74]. A rapid sequence of images is acquired through a volume of interest before, during, and after the intravenous administration of contrast material. These data are then fitted to mathematical models to analyze a number of physiological processes and to obtain quantitative perfusion parameters that reflect the vascular characteristics of the examined tissue, including blood volume (BV), blood flow (BF), mean transit time, and permeability area. Although experience is limited, DCE-US, DCE-CT, and DCE-MRI are likely to become useful tools for characterizing GEP-NETs [75, 76] and to evaluate their biologic aggressiveness [77], therapeutic response [22, 74] (Fig. 4), and prognosis [78, 79].

DCE-US enables the quantification of perfusion parameters by injecting ultrasonic microbubble-based contrast dye, and international guidelines have been produced to standardize this technique [80]. Among these parameters, the area under the perfusion curve (AUC)—a criterion linked to BV—has been identified as a reliable early predictor of response to antiangiogenic therapy using the RECIST criteria [81]. This finding was further validated in a multi-institutional cohort of 539 patients with various tumor types, including GEP-NETs, treated with different antiangiogenic agents [82]. In this group, early decreases in perfusion parameters were significantly associated with improved time to disease progression and overall survival. DCE-US has also been tested in a small study of patients with GEP-NETs who were treated with transarterial embolization (n = 10) or chemoembolization (n = 7). The authors proposed a new composite parameter combining functional and morphological data, named tumor vitality index, which may warrant additional exploration [83]. DCE-US allows early identification of tumor structural changes in response to PRRT, as decreased perfusion parameters are documented as soon as 6 weeks after therapy, whereas morphological changes may require a minimum of 6 months follow-up to be detectable [84]. DCE-US yielded comparable results to DCE-CT in the assessment of perfusion and morphological changes of liver metastases, particularly lesions located in the caudal and ventral parts of the liver, although adequate visualization of subdiaphragmatic structures remains an important limitation of this technique.

DCE-CT, also called perfusion CT, similarly enables the quantification of perfusion parameters to characterize tumor vascular features without some of the limitations of DCE-US (Fig. 4). The main drawbacks of this approach are the lack of standardization for data interpretation and the high radiation dose required. A significant correlation between tumor BF assessed by DCE-CT and histological assessment of intratumor microvessel density has been reported in pancreatic NETs [78]. Of note, a higher BF was observed in tumors with lower proliferation indexes, less aggressive histological features, and improved prognosis, as opposed to what may be found in other malignancies [78, 85]. DCE-CT has successfully been used to assess tumor perfusion changes in response to therapy in a small randomized phase II study comparing two antiangiogenic therapies, bevacizumab versus pegylated-interferon alpha 2b, in patients with advanced carcinoid tumors [86, 87]. A significant decrease in BF and BV was early observed (day 2 perfusion CT) compared with baseline data in bevacizumab-treated patients, but not in patients treated with interferon. Decrease in tumor BF following bevacizumab therapy was proportional to baseline BF suggesting bevacizumab decreased BF by a fixed percentage. Bevacizumab also induced objective responses by RECIST (18 vs. 0 % in bevacizumab vs. interferon arm) and was associated with longer PFS than the control arm. A subsequent study by the same group randomized 39 patients with low- to intermediate-grade NETs to receive bevacizumab or everolimus for one 21-day cycle, adding the alternate agent on cycle 2 (bevacizumab + everolimus). Serial functional CT assessments were mandatory. Bevacizumab significantly decreased tumor BF (44 %, p < 0.0001), and the addition of everolimus to bevacizumab was associated with further decrease in BF (29 %, p = 0.02). Everolimus alone was associated with 13 % increase in mean transit time (p = 0.02). Notably, several perfusion parameters (pretreatment tumor permeability surface, posttreatment mean transit time, percent reduction in BF, and percent reduction in blood volume) were significantly associated with best percent reduction in tumor diameters [88]. Taken together, these data suggest that perfusion CT parameters could potentially become useful surrogate markers for early response assessment to novel angiogenesis inhibitors in NETs. Figure 5 illustrates an example of tumor perfusion changes assessed by DCE-CT induced by antiangiogenic therapy. DECT provides a new parameter, the iodine uptake, which is assumed to reflect vital tumor burden by measuring the iodine uptake of active tumor. Iodinated contrast medium in lesions is mainly related to blood perfusion of viable tumor. This biological-related information could be a promising tool for evaluating tumor response [89] and would be of particular interest to assess response to antiangiogenic agents. However, to our knowledge, there are no published data concerning the use of this technique in GEP-NETs.

DCE-MRI allows the calculation of quantitative perfusion parameters by using a rapid series of T1-weighted images to observe the passage of contrast media intravascularly and as it leaks into the extravascular space. These pharmacokinetic parameters include the volume of extravascular space, the transfer constant (K ^trans), and the constant of return (k _ep), some of which are of difficult physiological interpretation. K ^trans, for example, is a parameter that depends on the equilibrium between BF and the vascular network permeability. Low k _ep and K ^trans values can indicate low perfusion, low permeability, and/or a small blood vessel surface area. Thus, any observed reductions in these parameters would reflect decreased blood flow and permeability in tumor lesions, which would be an indicator of therapeutic success. Miyazaki et al. [79] found that a lower pretreatment distribution volume and a high arterial flow fraction on baseline DCE-MRI were associated with a better response to treatment with radiolabeled octreotide therapy in patients with GEP-NET liver metastasis. Moreover, tumor and whole-liver distribution volume significantly increased after treatment in responding patients, suggesting that DCE-MRI could be a useful tool for predicting and monitoring response to PRRT in these patients. By contrast, baseline radiological patterns assessed by conventional imaging procedures (CT or MRI) did not predict response to yttrium-90 radioembolization in patients with GEP-NET liver metastasis [90]. However, DCE-MRI has poor intrapatient reproducibility, particularly in liver metastases [91]. Finally, one major limitation common to all DCE-based imaging modalities is that they only assess one or a few selected targets, which may not adequately represent the global tumor behavior, particularly in a setting of frequent tumor heterogeneity.

Another functional imaging technique increasingly being used is diffusion-weighted MRI (DW-MRI), which is based on the microscopic mobility of water molecules owing to thermal agitation. Water diffusion is basically restricted by interactions with cell membranes and macromolecules, and there is an inverse correlation between the degree of water motion and tissue cellularity and cell membrane integrity [70]. Thus, DW-MRI provides insight into cellular architecture at the millimeter scale, through a quantitative measurement of water diffusivity, called the apparent diffusion coefficient (ADC). The biological premise is that malignant tissues generally demonstrate higher cellularity, tissue disorganization, and increased extracellular space tortuosity, all of which contribute to a reduced motion of water, resulting in lower ADC values in malignant tumors compared with normal tissues [70]. The diagnostic superiority of DW-MRI over morphological techniques in a wide range of malignancies has led to the implementation of this fast sequence in all MRI exams in routine clinical practice.

DW-MRI is clinically useful at all stages in patients with GEP-NETs, including detection (Fig. 6), tumor characterization, staging, and therapy response assessment. The fusion of DW images with high b value (a factor that reflects the strength and timing of the gradients used to generate DW images) and T2-weighted MRI images improves the identification of pancreatic NETs [92, 93], especially in patients with small isointense lesions observed on conventional MRI sequences [94]. Moreover, DW-MRI and ADC maps can provide information that is useful for differentiating typical and atypical hemangiomas from other hypervascular liver lesions, including GEP-NET metastases [95]. An emerging clinical application of DW-MRI is the whole-body diffusion (WBD) technique for evaluating the extent of disease. A comparative study of ⁶⁸Ga-PET/CT versus WBD showed the overall superiority of ⁶⁸Ga-PET/CT for patient staging, particularly for detecting lymph node and lung tumor deposits, whereas WBD was more accurate in detecting liver and bone metastases [96]. DW-MRI is also a valuable tool for assessing tumor aggressiveness [97, 98]. As an example, Wang et al. [97] found a significant inverse correlation between ADC values and tumor cellularity or Ki-67 proliferative index, and this may thus help to predict the growth rate of endocrine tumors.

Regarding the potential role of DW-MRI for predicting or monitoring response to therapy, it is important to note that changes observed in functional imaging appearance may substantially differ depending on the specific mechanism of action of each treatment modality and the effect it induces in tumors [71]. Treatments that cause apoptosis (e.g., chemotherapy) result in transitory increased ADC values because of cell swelling, tumor lysis, and necrosis, although ADC values may eventually decrease as a result of tissue dehydration and fibrosis following cell death. By contrast, antiangiogenic therapies induce an initial and transient decrease in ADC values probably owing to reductions in tumor perfusion and in the extravascular-extracellular space resulting from vascular normalization and decreased vascular permeability [99]. Nevertheless, the opposite effect (ADC increase) may be observed if significant tumor necrosis is induced by vascular-targeted therapies. The complexity of interpreting DW-MRI changes induced by treatment is illustrated in a study conducted by Wulfert et al. [100]. In this work, DW-MRI and ⁶⁸Ga-DOTA-TOC-PET/CT images were acquired before and 3 months after one to two cycles of intra-arterial ⁹⁰Y-/¹⁷⁷Lu-DOTA-TOC therapy in 14 patients with hepatic metastases from GEP-NET. High baseline SUV_max values in ⁶⁸Ga-DOTA-TOC-PET and ADC mean values in DW-MRI were both associated with improved response to PRRT. A decrease in SUV_max in ⁶⁸Ga-PET was observed only in responding lesions after one to two cycles of therapy. In contrast, mean ADC values significantly increased after treatment in both responding and nonresponding lesions. Nevertheless, nonresponding lesions with increased ADC values on first follow-up assessment were more likely to achieve a decrease in size with longer follow-up.

Locoregional therapies, such as transarterial chemoembolization, may also cause an early reduction in ADC values after therapy (within the first few hours), after which consistent increases in ADC values usually occur, coinciding with the development of cystic and necrotic changes [99, 101]. In the field of GEP-NETs, Gowdra Halappa et al. [102] showed a significant increase in mean volumetric ADC (27 %, p < 0.0001) in all patients with neuroendocrine liver metastases receiving intra-arterial therapy (chemo- or radioembolization), 3–4 weeks after therapy. In this study, a significant response assessed by DW-MRI (defined as an increase of at least 15 % in volumetric ADC) or by DCE-MRI (defined as a decrease of at least 25 % in volumetric enhancement in the arterial phase or a decrease of at least 50 % in volumetric enhancement in the venous phase) correlated with improved survival. By contrast, the response assessed using RECIST, mRECIST, and EASL criteria did not correlate with survival. Patients who presented a response only by DW-MRI (ADC) or DCE-MRI (volumetric tumor enhancement) had a similar outcome in terms of survival than those who showed a response using both imaging modalities [102].

Functional imaging is consequently a promising tool for evaluating GEP-NETs. Most studies performed to date, however, have been retrospective and have involved small heterogeneous patient populations. Important efforts for improving technical qualification and standardization are certainly warranted before any of the parameters evaluated by functional imaging can be accepted as surrogate biomarkers for response assessment in standard practice. Although these techniques can be performed on standard clinical systems, they require strict protocols, careful acquisition, accurate contrast-agent dosing and injection rate, image timing, and image analysis for quantification. The standardization of these issues is essential to guarantee reproducibility [103]. The time course and the type of tumor changes induced must be further defined for the entire range of individual therapies or treatment strategies, as well as the magnitude of treatment-induced effects that actually result in patient benefit. Large prospective validation studies must also be conducted before these techniques can assist clinicians in decision-making in current clinical practice.