Introduction

Gastric cancer (GC) is one of the most common malignancies worldwide [1]. As already discussed in the first part (Part 1) of this critical review [2], this disease is managed through a standardised multidisciplinary approach where radiology plays a crucial role in the detection, staging, treatment planning and follow-up [3, 4].

The most useful techniques are endoscopic ultrasound, computed tomography (CT), magnetic resonance imaging (MRI) and 18F-fluorodeoxyglucose positron emission tomography (18F-FDG PET)/CT. At this regard, the PLASTIC trial [5] is an ongoing study that will evaluate the impact and cost-effectiveness of PET and staging laparoscopy in addition to initial staging in patients with locally advanced GC.

Different image-derived quantitative parameters from these techniques could be considered promising tools in the management of GC [6, 7], as they reflect a variety of biological processes (normal or pathological) both at baseline and after therapeutic interventions.

Quantitative imaging has the potential to improve the value of diagnostic testing and enhance clinical productivity and is increasingly important in preclinical studies, clinical research, and clinical practice [7]. Oncological imaging represents an ideal setting for the collection of new image-derived quantitative parameters from different techniques that can be potentially included in the clinical scenario [6]. The Radiological Society of North America underlined their importance as non-invasive tools with different applications in oncology and has promoted their use in clinical trials [7].

In the second part, we will provide a critical review on the state of the art of dynamic contrast-enhanced (DCE) MRI and 18F-FDG PET/CT findings.

Evidence acquisition

We searched MEDLINE/PubMed for manuscripts published from inception to 17 August 2018 (Fig. 1).

Fig. 1
figure 1

Flow diagrams showing the outcome of the initial searches resulting in the full studies included in the review for dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) (a) and 18F-fluorodeoxyglucose positron emission tomography/computed tomography (18F-FDG PET/CT) (b)

Full size image

DCE-MRI and image-derived quantitative parameters

DCE-MRI is a functional imaging technique in which multiphase images are acquired over a few minutes at baseline, during and after rapid intravenous injection of a contrast agent and a saline flush. Changes in signal intensity (reflecting tissue vascularity) can be observed and parametric maps of specific microvascular image-derived quantitative parameters can be derived [8, 9]. Basic recommendations include an adequate spatial/temporal resolution and knowledge of the inherent characteristics of the contrast agent. Semi-quantitative and quantitative analysis can be performed on specific regions of interest (ROIs) or on a pixel-by-pixel basis.

DCE-MRI requires high temporal resolution (usually 4–6 s/phase) and can be degraded by motion artefacts (e.g. respiratory or bowel peristalsis) [10]. Therefore, an injection of intravenous/intramuscular anti-peristaltic agent is advised to reduce the mobility of the gastric walls.

DCE-MRI reflects tumour angiogenesis (i.e. the creation of new blood vessels) and is directly associated with tumour growth and inversely correlated with prognosis [11,12,13].

Different quantitative parameters can be extrapolated from DCE-MRI maps (Tofts model) [14] such as:

  • Ktrans (min−1): volume transfer constant of gadolinium from blood plasma to the extravascular extracellular space (EES)

  • Ve (0 to 100%): volume of the EES per unit volume of tissue (i.e. the amount of “space” available within the interstitium for accumulating gadolinium)

  • Kep (min−1): rate constant gadolinium reflux from the EES back into the vascular system (i.e. it is the ratio: Ktrans/Ve)

  • AUC (mmol/s): area under the gadolinium concentration curve during a certain period of time.

The application of DCE-MRI in GC has been increasingly growing over the last few years thanks to the technical developments (e.g. the shortening of temporal resolution) and the advantage of free-from-radiation damage compared with CT.

Although certainly interesting in a research context, this technique has been mainly applied for neuro-oncological imaging so far. However, DCE-MRI in organ systems outside the central nervous system for oncological applications remains an active area of research, especially for breast, liver and prostate cancer. Other applications of DCE-MRI have been investigated, but as yet are not routinely used in clinical practice for GC. A possible explanation is that tumours are biologically complex structures and, differently from other organs such as the brain, the DCE-MRI protocols for GC are flawed by the presence of several artefacts (especially due to peristalsis) that can easily undermine the quality of the scan and the interpretation of quantitative data from the regions of interest analysed.

DCE-MRI in the detection and diagnosis of gastric cancer

Table 1 summarises the main studies analysing the role of DCE-MRI in GC.

Table 1 Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and gastric cancer
Full size table

The first study by Kang and colleagues dates back to 2000 [15] and reports the usefulness of dynamic and delayed MRI for T staging. The thickness and enhancement pattern of normal and pathological gastric walls were compared in 46 patients through a dynamic protocol including precontrast images and additional acquisitions of 30, 60, 90 and 240–300 s after injection of gadolinium. The pathological outer layers (mucosa and submucosa) showed earlier enhancement (i.e. between 30 and 90 s) than the normal gastric wall in 43/46 patients (93%) and the peak enhancement of the normal gastric wall was > 90 s in 17/46 patients (37%). A reasonable high consistency between MR staging and pathological staging for all T stages was reported (accuracy for T stage, 83%). Such results, although not related to any specific quantitative parameter, show that dynamic MR imaging was already a promising technique for predicting T staging in GC at that time.

Joo and colleagues [16] correlated DCE-MRI parameters with prognostic factors such as pathological T staging and epidermal growth factor receptor (EGFR) expression. Ve and iAUC were significantly higher for GC (0.133 and 5.533 mmol/s, respectively) when compared with normal gastric wall (0.063 and 3.894, respectively) (all p < 0.05). Additionally, Ve was positively correlated with T staging (ρ = 0.483, p = 0.023) and Ktrans was significantly correlated with EGFR expression (ρ = 0.460, p = 0.031). These findings suggest that DCE-MRI reflects tumour biology, providing prognostic information in patients with GC.

Ma and colleagues [17] compared DCE-MRI parameters in different histological subtypes of GC and investigated their correlation with vascular endothelial growth factor (VEGF) expression levels in 32 patients treated with surgical resection. Differently from the other studies, the ROIs were placed only on the lesions and the size was constant for each patient (10 mm). Mucinous adenocarcinomas showed higher Ve (0.491) and lower Ktrans (0.077 min−1) values than non-mucinous tumours (0.288 and 0.274 min−1, respectively) (p < 0.01). Differences were also observed for the Lauren classification, as the diffuse type showed higher Ve and Ktrans (0.466 and 0.249 min−1, respectively) values than the intestinal type (0.253 and 0.183 min−1, respectively) (p < 0.001). Additionally, Ktrans showed a significant correlation with the level of VEGF expression (ρ = 0.762, p < 0.001). Ktrans and VEGF are both related to the endothelial and microvascular permeability, which are in turn related to the neo-angiogenesis that is seen in tumours: in other words, a higher Ktrans is related to a higher level of VEGF, which is strictly related to a greater degree of angiogenesis. Together with the previous study [16], these findings suggest that angiogenesis increases the extravasation of gadolinium from the intravascular to the interstitial space, supporting the role of DCE-MRI as a potential tool to differentiate GC according to different histopathological features.

Li and colleagues [18] compared the performance of conventional breath-hold to free-breathing DCE-MRI using volume-interpolated breath-hold examination sequences. DCE-MRI parameters of normal gastric wall and GC were collected and perfusion parameters for both normal and pathological gastric walls were obtained. Kep was lower (0.750 vs 1.081 min−1; p < 0.05) while Ve was higher in GC (0.228 vs 0.162; p < 0.05). No significant differences for Ktrans and iAUC values between normal and pathological gastric walls were observed (p > 0.05).

Some examples of DCE-MRI in GC are shown in Figs. 2, 3 and 4.

Fig. 2
figure 2

DCE-MRI showing a tumour of the gastric antrum (a) in a 73-year-old male. The Ktrans (b) was 0.279 min−1, the Kep (c) was 0.605 min−1 and the Ve (d) was 0.482. Final pathology (e): diffuse type (Lauren classification), staged as pT4aN3. DCE-MRI of a tumour of the gastro-oesophageal junction (Siewert III) (f) in a 68-year-old male. The Ktrans (g) was 0.117 min−1, the Kep (h) was 0.461 min−1 and the Ve (i) was 0.253. Final pathology (j): mixed type (Lauren classification), staged as pT3N1. DCE-MRI of a tumour of the gastric antrum (k) in a 49-year-old male. The Ktrans (l) was 0.016 min−1, the Kep (m) was 0.575 min−1 and the Ve (n) was 0.029. Final pathology (o): intestinal type (Lauren classification), staged as pT4aN2

Full size image
Fig. 3
figure 3

DCE-MRI showing a tumour of the gastric antrum (a) in a 66-year-old female. In the pretreatment scan, the Ktrans (b) was 0.078 min−1, the Kep (c) was 0.237 min−1 and the Ve (d) was 0.347. The tumour was confirmed at biopsy (e). In the posttreatment scan, there was a reduction in tumour size (f), and the Ktrans (g) was 0.070 min−1, the Kep (h) was 0.295 min−1 and the Ve (i) was 0.263. Final pathology (j): intestinal type (Lauren classification), staged as ypT1bN0 (tumour regression grade 1)

Full size image
Fig. 4
figure 4

DCE-MRI of a tumour of the gastric antrum (a) in a 61-year-old female. In the pretreatment scan, the Ktrans (b) was 0.085 min−1, the Kep (c) was 0.176 min−1 and the Ve (d) was 0.539. The tumour was confirmed at biopsy (e). In the posttreatment scan, the tumour is still visible (f), and the Ktrans (g) was 0.128 min−1, the Kep (h) was 0.297 min−1 and the Ve (i) was 0.455. Final pathology (j): diffuse type (Lauren classification), staged as ypT3N0 (tumour regression grade 3)

Full size image

18F-FDG PET/CT and image-derived quantitative parameters

18F-FDG PET/CT is recommended for patients with newly diagnosed GC if clinically indicated and if metastatic cancer is not evident, as well as in the posttreatment assessment and restaging.

The standardised uptake value (SUV) from 18F-FDG PET/CT is a dimensionless ratio used to distinguish between normal and abnormal levels of glucose uptake and can be considered an image-derived semi-quantitative parameter, defined as the ratio activity per unit volume of a ROI to the activity per unit whole-body volume (Figs. 5 and 6) [19].

Fig. 5
figure 5

18F-FDG PET/CT scan of a 72-year-old man with gastro-oesophageal junction cancer (a–d) demonstrated by an intense uptake of 18F-FDG before treatment (SUVmax = 10.7) (c). After two cycles of chemotherapy (paclitaxel + cisplatin + fluorouracil) (e–h), the SUVmax of the lesion decreased to 4.8 (g), showing good response to the therapy. Final pathology (i) ypT3N0 (tumour regression grade 1)

Full size image
Fig. 6
figure 6

18F-FDG PET/CT scan of a 48-year-old woman with gastric cancer (a–d) demonstrated by an intense uptake of 18F-FDG before treatment (SUVmax = 4.7) (c). After one cycle of chemotherapy (capecitabine + paclitaxel) (e–h), no significant changes in 18F-FDG uptake (SUVmax = 4.8) were observed (g). Final pathology (i) ypT4aN1 (tumour regression grade 3)

Full size image

18F-FDG PET/CT to assess the primary lesion in gastric cancer

Table 2 summarises the studies on the role of 18F-FDG PET/CT to assess the primary lesion in GC.

Table 2 18F-Fluorodeoxyglucose positron emission tomography (18F-FDG PET) and aggressiveness in gastric cancer
Full size table

Stahl and colleagues [20] analysed the relationship between SUVmean and different tumour features from biopsy (including intestinal vs non-intestinal) in 40 patients. PET had a sensitivity of 60% in identifying locally advanced GC and the SUVmean was higher in the intestinal than in the non-intestinal type (6.7 vs 4.8; p = 0.03). No significant differences in the survival rate of patients with or without FDG accumulation (SUVmean cut-off, 4.6; p = 0.75) were observed. A clear limitation of this study is that the reference standard was biopsy and not radical surgery.

Mochiki and colleagues [21] reported a significant association between SUVmean and the depth of invasion, tumour size and nodal metastasis. They compared 18F-FDG PET findings with CT and found that 18F-FDG PET was less accurate for nodal staging (23% vs 65%). The SUVmean was higher for T2–T4 than T1 tumours (p < 0.05). Differently from the previous study [20], they observed a significant difference in the survival rate (p < 0.05).

Chen and colleagues [22] reported a sensitivity of 94% for 18F-FDG PET/CT (SUVmean = 7) and a significant association between FDG uptake and tumour size, nodal involvement and other histological features. They were among the first showing that the combination of 18F-FDG PET and CT was more accurate for preoperative staging than either modality alone (66% vs 51%, 66% vs. 47%; p = 0.002).

Oh and colleagues [23] performed a retrospective 18F-FDG PET/CT analysis of 136 patients treated with radical surgery. They set a threshold for SUVpeak from primary tumour of 3.2 to define hypermetabolic lesions and found that this was associated with tumour depth and nodal involvement (p < 0.001). The sensitivity and specificity for nodal involvement using the aforementioned threshold were 75% and 74% respectively.

Another group [24] reported the relationship between measurable and non-measurable GC on 18F-FDG PET/CT (defined as 1.35*SUVmax of liver+2*standard deviation of liver SUV). Among different parameters, a higher proportion of measurable tumours was found in well- or moderately differentiated GC than poorly differentiated tumours (71% vs 33% p < 0.05). Differently from the previous study [24], there was no difference for primary tumour stage and nodal metastasis.

Namikawa and colleagues [25] reported a sensitivity of 79% for the detection of GC for 18F-FDG PET/CT and a significant difference for SUVmax for patients with T3/T4 vs T1/T2 (9.0 vs. 3.8; p < 0.001), with and without distant metastasis (9.5 vs. 7.7; p = 0.018), and between stage III/IV and stage I/II (9.0 vs. 4.7; p = 0.017) after radical surgery. The SUVmax of the primary tumour was correlated with tumour size (r = 0.461; p < 0.001). The sensitivity, specificity and accuracy of 18F-FDG PET/CT for nodal involvement were 64%, 86% and 71% respectively.

18F-FDG PET/CT in treatment response of gastric cancer

We found six studies reporting on 18F-FDG PET/CT and treatment response in GC (Table 3).

Table 3 Fluorodeoxyglucose positron emission tomography (18F-FDG PET) and treatment response in gastric cancer
Full size table

Stahl and colleagues [26] compared different 18F-FDG PET/CT protocols and calculations of the SUVmean (time delay after 18F-FDG administration, acquisition protocol, reconstruction algorithm, SUV normalisation) for the early prediction of treatment response at baseline and after the first cycle of chemotherapy. They did not find any significant difference in the baseline and follow-up SUVmean calculation between protocols (p > 0.05), but higher SUV changes for responders than non-responders were observed (p < 0.01). They were among the first to demonstrate the robustness of 18F-FDG PET/CT for therapeutic monitoring, supporting the comparability of studies obtained with different protocols.

Vallböhmer and colleagues [27] analysed the differences in pre- and posttreatment SUVmax between responders and non-responders using the same histological definition as Stahl [26] (i.e. < 10% viable tumour cells in the specimen) but no correlation with treatment response was observed (p = 0.733). Significant differences in SUVmax were observed for the Lauren classification (p = 0.023) and tumour location (p = 0.041).

In another study on 17 patients [28] undergoing diffusion-weighted MRI and 18F-FDG PET/CT before and after treatment, no differences in treatment response were observed for pre- or posttreatment SUVmean (and their percentage change) (p = 0.605, p = 0.524 and p = 0.480). Treatment response was based on tumour regression grade (TRG) [32] and responders were considered TRG 1, 2 and 3 (i.e. including patients with more than 10% of viable cells).

Two studies [29, 30] evaluated the relationship between SUVmax and treatment response in advanced GC (i.e. no surgical specimens were used as the reference standard). Although follow-up imaging was performed at different time points (14 days vs 6 weeks after the start of chemotherapy) and different SUV thresholds for response were applied (40% vs 50%), both studies showed that metabolic changes in 18F-FDG PET/CT are predictive markers for response disease also for advanced GC. One study [30] showed a correlation between human epidermal growth factor HER2 status positivity (i.e. more aggressive cancer) and higher SUV uptake (p = 0.002).

Schneider and colleagues [31] reported that 18F-FDG PET/CT is able to detect non-responders (sensitivity, 91%; specificity, 47%; positive predictive value, 50%; negative predictive value, 90%; accuracy, 63%) but they could not prove that 18F-FDG PET/CT after the first cycle of chemotherapy can predict overall pathological response.

Similarly to the PRIDE study in oesophageal cancer [33], there is growing interest to develop models that predict the probability of response to neoadjuvant therapy in GC based on quantitative parameters derived from MRI and 18F-FDG PET/CT. However, given the controversial results at this regard [34], further studies are needed.

18F-FDG PET/CT in the prognosis of gastric cancer

We found eight studies on 18F-FDG PET/CT and prognosis in GC (Table 4). Significant results on the relationship between SUVmax and SUVmean and overall survival were reported by seven of them [35,36,37,38, 40,41,42], even though each study used different SUVmax and SUV mean cut-offs (Table 4). The study that did not show any significant difference in SUVmax and SUVmean with regard to prognosis was performed by Grabinska and colleagues [39]. A possible explanation is that a long range of follow-up was introduced in this study (range, 6 days to 5.2 years; median, 9.5 months), as also reported by the same authors. Therefore, the survival analysis from their study should be interpreted with caution. However, there is evidence of the relationship between SUVmax and SUVmean and prognosis in GC (Table 4).

Table 4 18F-Fluorodeoxyglucose positron emission tomography (18F-FDG PET) and prognosis in gastric cancer
Full size table

18F-FDG PET/CT and radiomics in gastric cancer

There is growing evidence of the importance of radiomics in medical imaging [43] and this applies also to 18F-FDG PET/CT findings [44, 45].

A recent review has shown the promising role of radiomics obtained from different techniques—including 18F-FDG PET/CT—in gastro-oesophageal tumours [46].

Jiang and colleagues [47] have also developed a dedicated radiomic score using the features from 18F-FDG PET/CT in GC. In their study, they concluded that the radiomic signature was a powerful predictor of overall and disease-free survival and could add prognostic value to the traditional staging system.

However, as the current literature on this specific topic is still preliminary, there is a need of standardisation and different multicentre studies before including radiomics from 18F-FDG PET/CT in the clinical routine for GC.

Limitations

Quantitative imaging is becoming an increasingly common tool in modern radiology and its potential impact on patient care and on clinical outcomes is huge. However, it is broadly accepted that surrogate quantitative parameters of tumour biology assessed by imaging still require extensive standardisation and validation to proof that the surrogate represents the pathophysiological process under investigation. As reported by Rosenkrantz and colleagues [48], there are some practical aspects that should be considered when discussing the role of image-derived quantitative parameters. These are (i) accuracy (of a measurement, for example); (ii) repeatability and (iii) reproducibility (especially when quantitative imaging is performed in serial scans over time, as this allows to discriminate measurement error from biologic change) and (iv) clinical validity (i.e. impacting and improving patient’s life).

Therefore, some limitations from the papers discussed in this study should be reported. Firstly, for DCE-MRI, our review shows that the ROIs in all studies have been drawn on one selected axial section. This represents an important limitation, as these findings may be less representative of the whole tumour. Future studies should perform quantitative analysis on the whole volume obtained by contouring the tumour borders on each slice by planimetry. There is also a lack of optimised perfusion MRI protocols, dedicated postprocessing software programmes and high variability between MR scanners.

As far as 18F-FDG PET/CT imaging is concerned, a clear limitation is that the SUV is dependent on many factors including the ROI delineation, the activity injected, plasma glucose levels, and body size. There is variability between 18F-FDG PET/CT scanners, as well as in the accuracy of the image reconstruction and correction algorithms. The increased 18F-FDG uptake can be also seen in inflammatory or granulomatous processes and in sites of physiological tracer biodistribution.

Gastric distention, achieved by the consumption of water, milk or foaming agents before scanning, and a late-time-point 18F-FDG PET/CT scanning can relatively differentiate the physiological uptake from the malignant lesion.

Finally, standardised guidelines on how to interpret the quantitative results from DCE-MRI and 18F-FDG PET/CT have yet to be reported.

Conclusions

Similarly to the ADC from diffusion-weighted MRI and texture analysis from CT [2], different image-derived quantitative parameters from DCE-MRI and 18F-FDG PET/CT are promising tools in the management of GC. However, extensive standardisation and validation are still required before they can become an essential cornerstone for GC.