Introduction

Gastric cancer is one of the most common malignancies worldwide [1].

The recommended management of this disease is a standardised multidisciplinary approach, which involves surgery, neoadjuvant and adjuvant chemotherapy. Surgical pathology is the reference standard for staging gastric cancer, and the Tumour, Node, Metastasis (TNM) classification is the most important tool for an adequate therapeutic plan and for predicting treatment response and prognosis [2]. Imaging plays a crucial role in the detection, staging, treatment planning and follow-up of this disease [3, 4].

A biomarker indicates a biological process (normal or pathological) both at baseline and after therapeutic interventions [5]. Oncological imaging represents an ideal setting for the collection of new biomarkers from different techniques, as oncological patients usually undergo multiple imaging studies [6]. The European Society of Radiology published a white paper on imaging biomarkers that underlined their importance as non-invasive, quantitative tools with different applications in oncology [7]. The Radiological Society of North America has also promoted the use of imaging biomarkers in clinical trials [8].

Quantitative imaging is the extraction of quantifiable features from medical images to assess normal or pathological conditions [9, 10].

There is a wide range of imaging biomarkers from different techniques. It is possible to extrapolate multiple imaging biomarkers through quantitative analysis of data derived from different techniques, such as contrast-enhanced multidetector computed tomography (CE-MDCT), diffusion-weighted magnetic resonance imaging (DW-MRI), dynamic contrast-enhanced (DCE)-MRI and 18F-fluorodeoxyglucose positron emission tomography (18F-FDG-PET) [11,12,13,14,15,16,17,18,19].

While reviews and meta-analyses of other imaging modalities (e.g. 18 FDG-PET) have been already carried out [20,21,22,23,24,25], this is the first time that a critical review has been specifically designed to summarise the state of the art of imaging biomarkers such as ADC from DW-MRI and texture features from CE-MDCT in gastric cancer.

Currently, there are only a few defined diagnostic and prognostic biomarkers for gastric tumours but none are related to imaging techniques [26,27,28].

In an attempt to overcome such limitations, in this review we specifically provide the state of the art on the use of two imaging biomarkers: the apparent diffusion coefficient (ADC) from DW-MRI and texture features from CE-MDCT.

Evidence acquisition

We searched MEDLINE/ PubMed for manuscripts published from inception to 6 February 2018. Initially we searched for (gastric cancer OR gastric tumor) AND diffusion weighted magnetic resonance imaging. Then, our search included (gastric cancer OR gastric tumour) AND texture analysis AND computed tomography. Figure 1 shows the literature search and study selection.

Fig. 1
figure 1

Flow diagrams showing the outcome of the initial searches resulting in the full studies included in the review for diffusion-weighted magnetic resonance imaging (DW-MRI) (a) and contrast-enhanced multidetector computed tomography (CE-MDCT) (b)

Full size image

DW-MRI and imaging biomarkers

Eleven studies on DW-MRI were relevant to our query, and they were assessed in full.

Basic concepts of DW-MRI

DW-MRI measures the mobility of water molecules within tissues. Water diffusion is restricted in highly cellular tissues such as tumours, but some conditions (fibrosis/inflammation) can also show restricted diffusion [29].

Tissues with higher cellularity result in higher signal intensity on DWI. The b value is a parameter that identifies the measurement's sensitivity to water diffusion. The optimal b value should attenuate the healthy background tissue more than the lesion, with a reasonable signal-to-noise ratio at the same time. The greater the b value is, the stronger the DWI signal and the detection of pathological areas [29]. It is possible to generate a dedicated map that enables the calculation of the ADC. An area of higher signal intensity on DWI appears as a low-signal intensity on the ADC map. The region of interest (ROI) from which the ADC is calculated can be drawn as a small area or as a larger area encompassing the whole tumour on a single slice (Fig. 2). The ADC can also be obtained drawing the ROIs on all the slices where the tumour is visible (planimetry). This parameter is therefore a quantitative measurement of the mobility of water molecules within tissues and tumours, given their higher cellularity, have lower ADC values than normal areas [30].

Fig. 2
figure 2

The arrows indicate a tumour of the lesser curvature on T2-weighted imaging (a) and on diffusion-weighted magnetic resonance imaging (DW-MRI) using different b values (b–e). The apparent diffusion coefficient (ADC) is calculated on the ADC map (f) from a small (red) or large (yellow) region of interest (red) on a single slice

Full size image

DW-MRI to assess the aggressiveness of gastric cancer

MRI of abdominal organs is often influenced by macroscopic motion. Imaging acquisition should be performed with breath-hold or respiratory triggering sequences, after the administration of intravenous/intramuscular anti-peristaltic agents and oral water to distend the gastric lumen [30, 31].

Growing evidence supports the use of DW-MRI in the assessment of the aggressiveness of gastric cancer [32,33,34,35,36,37,38,39,40,41,42,43]. Table 1 summarises the recent studies that have addressed this topic by analysing the role of ADC.

Table 1 DW-MRI and aggressiveness of gastric cancer
Full size table

Cheng et al [38] calculated ADC of pathological lymph nodes (dimensional criteria ≥ 5 mm in short-axis diameter) on a pixel-by-pixel base. After extended lymphadenectomy, metastatic lymph nodes showed lower median ADC than benign lymph nodes (1.28 vs. 1.55 × 10-3 mm2/s, p < 0.001). A cut-off of 1.39 × 10-3 mm2/s could discriminate between pathological and benign lymph nodes with a sensitivity of 85.7% and specificity of 79.4%. Quantitative results were also compared with morphological factors including size, border irregularity and enhancement patterns. The authors concluded that ADC holds promise to identify metastatic lymph nodes in gastric cancer. Their results are relevant, but lymph nodes < 5 mm were not evaluated because of the limited resolution of DWI.

Liu et al [35] conducted a retrospective study where the mean and minimum ADCs of gastric cancer were calculated and compared with surgical specimens. Both parameters showed a significant difference in detecting the different degrees of local invasion and nodal involvement. In particular, the correlation (r) with postoperative T staging was -0.464 (mean ADC) and -0.476 (minimum ADC) (p < 0.001), while the correlation (r) with postoperative N staging was -0.402 (mean ADC) and -0.397 (minimum ADC) (p = 0.001 and 0.002, respectively).

The same group conducted a whole-lesion ADC histogram analysis in 80 patients comparing the results with radical specimens [43]. Significant differences for all ADC parameters with respect to T and N stages were found. In particular, the ADC10% provided the largest area under the curve (0.794) to identify nodal involvement.

Liu et al [34] investigated the role of whole-volume ADC-based entropy (a parameter describing tumour heterogeneity) in the preoperative assessment of gastric cancer. Entropy-related parameters from the ADC map significantly correlated with postoperative T, N and overall stage as well as vascular and perineural invasion.

Conversely, Hasbahceci et al [42] showed that it is not feasible to evaluate nodal involvement in gastric cancer by means of ADC. The cohort comprised 23 patients and ADC did not differentiate between benign and metastatic lymph nodes. However, the accuracy of DW-MRI in identifying metastatic lymph nodes in this study was 52.17%, 65.21% and 69.56% for group Ia, Ib and II, respectively. The overall accuracy of N-staging based on DW-MRI was 13%. The lack of motion correction manoeuvres and the small number of patients represent important limitations of this study.

Zhong and colleagues [41] showed that pathological lymph nodes have a lower ADC than benign lymph nodes (1.15 vs. 1.48 × 10-3 mm2/s, respectively; p <0.001). The same group compared the results from DWI with morphological MRI (T2-weighted imaging) and CE-MDCT; the combination of the three techniques yielded the highest area under the curve (0.893) for nodal involvement. The results are interesting, but a rigorous radiological-pathological correlation could not be assessed because of the large number and complex anatomic location of lymph nodes in gastric cancer. Another study [37] conducted on 89 patients reported that ADC is different according to local invasion, nodal involvement and the 7th TNM edition stage groups (p < 0.001). Therefore, there is evidence that ADC is a promising, non-invasive imaging biomarker for the evaluation of the aggressiveness of gastric cancer.

DW-MRI in treatment response of gastric cancer

In recent years, there has been considerable interest in the application of DW-MRI to assess treatment response in gastric cancer [44,45,46,47,48]. Table 2 reports the main studies that have addressed this topic by means of ADC. Two studies from the same centre [46, 47] showed that ADC could be considered a reliable indicator of treatment response. De Cobelli et al [46] analysed ADC before and after treatment and compared the results with tumour volume, considering tumour regression grade (TRG) as the reference standard [49]. Responders showed a significant increase in ADC compared with non-responders (85.45% increase vs. 8.21% decrease, respectively: p < 0.001). However, this study also included patients with oesophageal cancer. Similar results [47] were found comparing ADC and standardised uptake volume (SUV) values in 17 patients undergoing DW-MRI and PET, before and after therapy. Changes in ADC were correlated with TRG (r = -0.78; p < 0.001), while SUV did not yield significant results. The results from this study are promising even though the number of patients is small (n = 17).

Table 2 DW-MRI and treatment response in gastric cancer
Full size table

Lee and colleagues [44] investigated the usefulness of PET/MRI in predicting treatment response for advanced, unresectable gastric cancer in 11 patients. In addition to ADC, SUV and perfusion parameters were calculated prior to chemotherapy. Response to treatment was evaluated on the basis of the response evaluation criteria in solid tumours (RECIST) 1.1 criteria [50]. Differently from perfusion parameters, ADC and SUV were not significantly different between responders and non-responders. However, the lack of both post-treatment imaging and comparison with surgical specimens represents an important limitation of this study.

Zhong and colleagues [45] conducted a study on 106 patients with biopsy-proven gastric cancer undergoing DW-MRI before and after chemotherapy at different time points. Mean ADC was calculated for benign and metastatic lymph nodes, and response to treatment was evaluated on the basis of the RECIST 1.1 criteria [50]. ADC significantly increased after treatment in all groups (complete response, partial response and stable disease group), predicting nodal response to neoadjuvant therapy. These studies support the potential role of DW-MRI in tailoring the therapeutic plan for gastric cancer (Fig. 3).

Fig. 3
figure 3

Image of a tumour of the gastro-oesophageal junction on T2-weighted (a and c) and diffusion-weighted (b and d) imaging before and after treatment. There has been a decrease in the conspicuity on DW-MRI (d) and an increase in the apparent diffusion coefficient (ADC) value after treatment (d)

Full size image

DW-MRI in the prognosis of gastric cancer

At present, two studies [37, 51] have shown that ADC can be considered a prognostic biomarker for overall survival and risk stratification of gastric cancer, according to the 7th TNM edition [52]. The first study [37] reports that an ADC of 1.8 × 10-3 mm2/s may discriminate between stage I and stage II and that values ≤ 1.36 × 10-3 mm2/s are associated with stage III. The second paper [51] shows that an ADC ≤ 1.5 × 10-3 mm2/s is related to a negative outcome, along with T and N stages after surgery. These results confirm the growing interest that quantitative MR imaging in gastric cancer is gaining in the scientific community [53].

CE-MDCT and imaging biomarkers

Five studies on CE-MDCT were relevant, and they were assessed in full (Fig. 1).

Basic concepts of texture analysis

Texture analysis provides information on the intrinsic variation of pixels within an image. In oncology, this represents a non-invasive method to assess the heterogeneity within a tumour and could reflect its intrinsic aggressive biology or treatment resistance [54, 55]. There are different levels of textural features; these range from first-order statistics (that evaluate the grey-level distribution from pixels in a given area, such as energy, entropy, skewness) to higher-order statistics that investigate the relationship between different voxels [56].

This is a hot topic for quantitative imaging in oncology in order to evaluate lesion characterisation, response to therapy and prognosis [56]. However, there is still marked variability in methods and post-processing techniques in addition to the need to identify the key parameters among hundreds of potential imaging biomarkers from CE-MDCT [56].

A standardised CE-MDCT protocol for adequate texture analysis in gastric cancer

An adequate CE-MDCT scan is crucial to obtain a reliable and reproducible texture analysis for research purposes. As a rule of thumb, the CE-MDCT scan parameters usually applied for gastric cancer are: 120 kVp, 250 mAS, collimation 64 × 0.625 mm, 3 mm helical thickness and 1 mm reconstruction thickness [57]. The first step for an adequate set of images is to ensure a correct distension of the gastric cavity. This can be reached by using negative endoluminal contrast agents (500–750 ml water or gas produced by effervescent granules).

Air filling is more accurate than water filling for tumours of the antrum; this is very important when drawing the regions of interest (ROIs) for texture analysis, as shown in Fig. 4.

Fig. 4
figure 4

Tumour of the antrum (arrows) using water (a; axial view) and air filling (b; coronal view) on contrast-enhanced multidetector computed tomography. Air filling is more accurate when drawing the regions of interest (ROIs) for texture analysis (blue: primary tumour; orange: pathological lymph node)

Full size image

The second step is the use of an intravenous/intramuscular anti-peristaltic agent (e.g. scopolamine-N-butyl bromide or glucagone) similarly to DW-MRI. This reduces the mobility of the gastric walls during the scan acquisition so that the ROIs can be drawn on static images and easily compared for texture analysis (Fig. 5). In this regard, the 8th TNM Staging Manual mentions the important role of gastric distention for the staging of gastric cancer [58]. Figure 6 is a set of different CE-MDCT images of gastric cancer acquired using the aforementioned recommendations, allowing accurate ROI delineations for texture analysis.

Fig. 5
figure 5

Comparison between a protocol without and with administration of anti-peristaltic agent on multidetector computed tomography. The image shows how the region of interest (ROI) delineation is more reproducible at different time points in the second scenario. Unenhanced (a and d), arterial (b and e) and portal (c and f) phases

Full size image
Fig. 6
figure 6

Contrast-enhanced multidetector computed tomography images of different gastric cancers (arrows) that have been acquired using a standard protocol, thus allowing accurate regions of interest for texture analysis (a, b, e, taken as an example)

Full size image

Texture analysis and gastric cancer

Texture analysis has been receiving much attention in the management of gastric cancer. However, only a few studies have been published on this topic so far [59,60,61,62,63], and they are listed in Table 3.

Table 3 Texture analysis and gastric cancer
Full size table

First, there are no specific recommendations regarding the most accurate CE-MDCT phase for texture analysis. Some studies in Table 3 were conducted in the (late) arterial phase [61, 62], while others in the venous phase [63] or in both [59, 60].

However, the administration of intravenous contrast is important for texture analysis, as this allows the quantification of the vascularity before and after treatment and helps assessing the degree of infiltration of the surrounding fat. Texture analysis reflects some tumour features related to tissue heterogeneity at a cellular level; this means that differences at a microscopic level are also detectable macroscopically by CE-MDCT imaging [56].

A closer look at the results in Table 3 reveals that texture analysis can differentiate histological subtypes (adenocarcinoma, lymphoma and gastrointestinal stromal tumours) [59]. Similar results have been published by Liu et al [60]; the authors found that texture parameters can predict the degree of differentiation, Lauren classification and vascular invasion. There is also evidence that textural features can predict treatment response [61] when the response rate is based on TRG [49]. Like DW-MRI, quantitative imaging from texture analysis is associated with the prognosis of untreated [62] and treated [63] gastric cancer. The studies presented in Table 3 offer compelling evidence that CE-MDCT-based texture analysis holds promise in the management of gastric cancer, even though further studies are necessary before widespread application of this technique in common clinical practice.

Limitations

Results so far have been encouraging but have given rise to many questions in need of further investigation as they limit the application of imaging biomarkers in daily clinical practice. First, the large variability in the selection of imaging acquisition parameters for DW-MRI and CE-MDCT (e.g. standardised b values, arterial vs. venous phase) is a considerable challenge. Future studies should aim at standardising data-acquisition protocols as well as validating new methods to extract quantitative parameters from imaging techniques (e.g. ROI assessment).

Second, multicentre trials on large cohorts of patients have yet to be carried out. Focused groups of researchers willing to perform multicentre, international studies are needed to address the key questions related to imaging biomarkers for gastric cancer [64]. The radiological community should continue to play an active role in the academic research by providing valuable expertise to incorporate imaging biomarkers into future clinical trials.

Conclusions

As emerged from this paper, imaging biomarkers are promising tools of modern radiology [65, 66] and could become an essential cornerstone for gastric cancer therapy. Despite these limitations, this review highlights the great potential of ADC and texture analysis for gastric cancer. This is a vital issue for future guidelines, especially in the era of “personalised medicine”, where the goal is to tailor specific treatments for each single patient.