Detection and differentiation of paediatric renal tumours using diffusion-weighted imaging : an explorative retrospective study

Background: Diffusion-weighted imaging (DWI) is a promising supplemental technique in oncological magnetic resonance imaging (MRI). We investigated the diagnostic utility of DWI for detection and characterisation of paediatric renal tumours. Patients and Methods: Eleven consecutive patients (median 4 years, range 4 days to 15 years, females n=9) with histologically proven renal tumours (nephroblastoma n=6, nephroblastomatosis n=2, connatal nephroblastic nephroma n=1, renal cell carcinoma n=1, local recurrence of nephroblastoma n=1) underwent routine clinical MRI at 1.5 Tesla using a standardised oncological scan including free-breathing DWI. We retrospectively analysed lesion detectability and conspicuity of tumour vs. adjacent tissue on DWI and ceT1w. Results: All tumour manifestations were detectable on DWI by high signal at high b-values. Mean ADC values ranged between 0.56 to 0.95 with a wide overlap between malignant lesions and nephroblastomatosis. Mean SI ratios were significantly higher on DWI, compared to ce-T1w (4.5±3.8 vs. 1.6±0.7, p<0.001). Six small foci of nephroblastomatosis were occult on ce-T1w imaging, but clearly delineated on DWI. One small bone metastasis was seen on DWI only. In two patients with stable unilateral manifestations of nephroblastomatosis, follow-up MRI showed mean ADC values of 1.0 ± 0.3 over three years. Conclusion: According to our preliminary experience, DWI reliably detects paediatric renal tumours and metastases. Apparently, DWI cannot distinguish between malignant and non-malignant paediatric renal tumour entities based on mean ADC, but yields superior lesion conspicuity. DWI in free-breathing technique without the need of i.v. contrast application deserves further evaluation as stand-alone imaging, especially for follow-up in young patients.


Statistical analysis
Normally distributed data is presented as mean ± standard deviation. The comparison of signal intensity ratios between ce-T1 and DWI was performed with a two-sided Wilcoxon signed ranks test. α < 0.05 was considered as indicating statistical significance. Spreadsheet analyses were performed with Microsoft Excel 2007 for Windows. Statistical tests were calculated with IBM SPSS 21 for Windows.

Results
The eleven11 patients in our study group were diagnosed with nephroblastoma (WT) n=6, nephroblastomatosis (NBS) n=2, connatal mesoblastic nephroma (CMN) n=1, juvenile renal cell carcinoma (RCC) n=1 and local recurrence of nephroblastoma n=1. All MRI studies yielded diagnostic image quality, and all tumour lesions identified on standard MRI sequences were detected on DWI, as well. Data on imaging characteristics and quantitative measurements are summarised in Table 1

Patients with primary diagnosis of Wilms tumour
All six patients with newly diagnosed WT underwent MR scanning before and after neoadjuvant standard chemotherapy (SIOP 2001 [24]). In two of these patients, WT was the only tumour focus on initial staging. The third WT patient had multipleextensive retroperitoneal and hepaticliver metastaseis aton initialprimary diagnosis ( Figure 21). The fourthAnother patient had a three ipsilateral foci of NBS measuring 15 to 26 mm and three small contralateral NBS lesions of 3 to 5 mm in diameter ( Figure 32 and 34). Of these, the threeAll these small NBS foci were virtually occult on contrast-enhanced T1w imaging, but showed high signal intensity on DWI, whereas the ipsilateral larger NBS foci were detectable on both DWI and ce-T1w. The fifth patient had bifocal WT in one kidney and additional multiple small bilateral foci of nephroblastomatosis. Of the latter, five lesions measuring 5 to 10 mm in diameter were visible both on DWI and ce-T1w, while several smaller foci would have been missed on ce-T1w alone.
Finally, one WT patient presented with a 5 mm nephrogenic rest in the contralateral kidney visible on both contrast-enhanced and diffusion-weighted sequences. Involvement of the renal vein was not observed. Mean signal intensity ratio of WT was significantly higher on DWI, compared to ce-T1w (p=0.043).
All patients underwent neoadjuvant chemotherapy and pre-operative MRI re-staging, which confirmed a marked reduction in tumour size, a decrease in signal intensity ratios for both ce-T1w and DWI and an increase in mean ADC as indicators of response to treatment in all patients ( Table 1). After successful treatment with surgical tumour resection and complete remission, five patients remained in complete remission at a median follow-up of 29 months, while one patient suffered progressive disease with new pleural metastaticses spread at nine months after initial diagnosis.  All NBS manifestations showed slightly hypointense homogeneous signal on ce-T1w, compared to the renal cortex. Subjective lesion conspicuity was superior on DWI for all NBS lesions, and SI ratios on DWI were higher, compared to ce-T1w (p<0.05, Wilcoxon signed rank test) (Table1). Of the 17 NBS lesions quantified for our study, 11 (65%) were detectable with both ce-T1w and DWI, all of these measuring 5 mm or above in diameter, while six NBS foci would have been missed without DWI. These small The small NBS lesions were hardly visible or even virtually occult inon ce-T1w in the presence of overlying motion artefacts, but all showed very high signal on DWI at high b-values. The large NBS foci had mean ADC values ranging betweenof 0.3 -and 0.5, while the small lesions with 3 to 5 mm in diameter were measured with higher mean ADC of 0.7 to-1.2. Mean ADC of NBS was 0.7 ± 0.3 (Table 1).

Discussion
Our explorative single-centre study with retrospective data analysis demonstrates the usefulness of DWI for detection of renal tumours and their metastases in paediatric patients.
All tumour lesions detectable on standard MRI were also seen on DWI, but were displayed with significantly higher signal on diffusion-weighted images and thus clearly stood out from The preliminary experience from our study data, however, suggests that DWI cannot differentiate between malignant renal tumours and NBS/nephrogenic rests, based on mean ADC values.
Renal tumours account for a small, but significant proportion of malignancies in children [1].
Wilms tumour (WT), the most prevalent entity by far, is highly responsive to modern multimodal therapy, and a majority of WT patients enjoys excellent long-term survival rates [2025].
According to European guidelines, therapeutic standard in WT patients is initial diagnostic imaging for tumour staging, neoadjuvant chemotherapy, re-staging and tumour nephrectomy, possibly followed by post-operative chemotherapy and irradiation depending on tumour stage.
Pre-therapeutic tumour biopsy is not usually performed in young children presenting with solid renal tumours and imaging findings consistent with WT [5]. Standard MR imaging, however, cannot reliably differentiate between the various renal tumour entities, nor confidently predict malignancy or the biological potential of a renal mass [216]. Furthermore, considering the presence of bilateral WT in about 10% of the patients [272], the frequent association of WT and nephrogenic rests or nephroblastomatosis (NBS) [6][7] and the low rate of malignant transformation of NBS lesions [6][7][8], therapy planning and options for nephron-sparing surgery would certainly benefit from more accurate image-based characterisation of these renal lesions. In patients with bilateral WT undergoing nephron-sparing surgery [11], accurate detection and description of additional NBS foci may help to optimize surgical resection and may result in higher proportions of preserved renal tissue.
DWI is a MR imaging technique that visualises the degree of restriction experienced by water molecules in the extracellular space of biological tissues. The DWI signal is closely related to tissue cellularity [15], among other factors, and previous studies found evidence for improved tissue characterisation with quantitative measures of diffusivity. In paediatric oncology, embryonal malignant tumours typically present with high cellularity, characterized by densely packed cells and consecutively narrowed extra-cellular space with many barriers to the diffusion of extracellular water. In contrast, benign tumours or lesions rich in interstitial stroma are characterized by a low cell count, wider extracellular space and therefore less impeded diffusion [15]. While DWI features of renal tumours in adults have been a focus of several recent studies and a meta-analysis [18][19][20], there is a paucity of such data from paediatric cohorts. Tumour characterisation, based on ADC, has been reported in other paediatric neoplasiestumour entities. In a study on neuroblastic tumours in children, DWI was able to differentiate between neuroblastoma and ganglioneuroblastoma/ganglioneuroma based on mean ADC with little overlap between the two entities [1721], which otherwise show very similar imaging findings on standard MRI. At our own institution, we found significant differences in mean ADC and between malignant and non-malignant osseous and soft-tissue lesions [1923]. and established a cCut-off ADC values for mean ADC around 1 × 10 −3 mm 2 /s with lower ADC being predictive of malignancy were found in these studies [21,23]. as a sensitive and specific predictor of malignancy. This cut-off value compares well to the data reported by Gahr et al. [17]  without overlap to musculoskeletal malignancies [23].
In oThe data from our present study, free-breathing DWI produced high-quality scans of our paediatric patients ( Figure 2) and showedconfirms the high sensitivity of DWI for malignant diseaserenal tumour foci. In our small study group, we found evidence that allows us to and eextents the concept of DWI-based tumour detection to paediatric this concept to malignant renal masses. Mean signal intensity of tumour foci, as compared to adjacent tissue, was higher on DW images than on contrast-enhanced images, which translates into higher lesion contrast.
We decided to use signal intensity ratios as the quantitative metric in our study, rather than contrast-to-noise ratio (CNR). Using modern scanner hardware and software, image noise is no longer evenly distributed in MR images acquired with multi-channel receiver coils and built-in image filtering. The traditional straight-forward approach of measuring image noise with ROIs on the images may be affected in an unpredictable manner and extent under such circumstances. Quantification of image noise and evaluation of noise distribution by the means of field maps was not performed in our clinical setting.
Although DWI allowed ready tumour detection in all our patients, we could not, however, differentiate malignant from non-malignant renal tumours based on ADC. Mean ADC values lower than 1.0 weasre seen in all malignant lesions, but also in NBS, with the small ROI technique. In heterogenousheterogeneous tumours containing both solid and cystic tissue, it is our practical approach to target portions of the tumour exhibiting restricted diffusion and to , however, non-malignant lesions in our study, namely the nephrogenic rests and NBS foci, also showed markedly restricted diffusivity and mean ADC less than 1 × 10 −3 mm 2 /s.
Reference ADC values measured in the renal cortex and in the erector spinae muscle showed little variation between patients or between MRI scanners and fellshowed no substantial across-study variation in comparisioncomparison to within the previously reported datarange [1923,3027]. The association of relatively high ADC in small NBS lesions and low ADC in larger NBS lesions is very likely the result of partial volume effects, as the size of these small NBS foci were likewise seen with a marked increase in ADC and a substantial reduction in tumour size on the pre-operative re-staging (Table 1), while the two non-responsive lesions, that is the juvenile NCC and the local WT recurrence, had persistinglypersistingent low mean ADC. Based on our preliminary experience, DWIADC quantification therefore seems to present amay therefore be a useful supplemental technique for evaluating response to therapy in children undergoing neoadjuvant treatment for Wilms tumour.
For presentation of diffusion-weighted images in print and in clinical rounds, we found it helpful to use images with an inverted greay scale, as seen in Figures 1 through 5. From our experience, this mode of presentation carries a smaller risk of degraded image reproduction in print. Furthermore, our clinicians are used to looking at scintigraphic studies and PET images.
As DWI with inverted greay scale somewhat resembles bone scintigraphy or PET scans, we often use inverted DWI when communicating our findings to the referring clinicians. To date, there is no data to support the hypothesis that greay scale inversion for diffusion-weighted images may be helpful for routine readings, as it was reported for the detection of pulmonary nodules on chest films [37].

Limitations
The retrospective design of our explorative study, the limited number of patients available for analysis and the different scanner hardware used in data acquisition all need to be taken into account when drawing conclusions from the presented study data. Apart from Wilms tumour and nephroblastomatosis, other paediatric renal tumour entities are generally very rare with frequencies below 5%, so that only one case of NCC and CMN each were among our study patients. In fact, only rudimentary statistical testing was possible considering the small sample size. A prospective multi-centre approach with standardiszed examination protocols will be necessary to recruit larger patient cohorts and to arrive at more definite conclusions. MRI examinations in our study were performed at 1.5 Tesla on three different MR scanners from the same manufacturer with very similar scanning parameters. The scanning protocols were designed with a particular emphasis on comparability and our results in previous studies did not show any substantial inconsistencies. Overall image quality wasis certainly superior for examinations performed on the Avanto and Aera scanner, compared to the older Magnetom Symphony. At any rate, a potential impact of hardware and software configuration on our study results cannot completely be ruled out. AAnother technical issue of concern is the still limited spatial resolution of DWI, relative to small-sized NBS lesions, which renders quantitative ROI analysis of ADC susceptible to partial volume effects, as discussed above.
Another limitation arises from the technical specifications chosen for the DWI sequence. We acquired DW images at two b-values only, being aware that bi-exponential modelling based on three or more b-values produces more accurate ADC values [13]. While ADC values computed from a bi-exponential model may differ to some extent from the ADC values reported in our study, the differences would be small, in our experience, compared to variability arising from intra-and inter-observer variability of ADC measurements in a clinical setting. Time is a crucial factor in paediatric MRI and measuring additional b-values considerably adds to total scanning time. We argue that the potential benefit of measuring additional b-values or higher b-values, requiring more averages because of less signal [13], would not justify the extra-time and the additional discomfort to the patients, some seriously ill and in bad clinical condition and some undergoing MRI in sedation.
As it is, the study presents proof-of-concept work. Future studies should include inter-observer variability analyses and collect observations at different levels of experience to identify learning curves with DWI and standard MRI sequences.
In conclusion, our study is a first systematic attempt to evaluate the diagnostic performance of diffusion-weighted imaging in paediatric patients with renal tumours. DWI evidently is a useful and promising imaging modality for detecting and monitoring paediatric renal masses. While DWI based on ADC with mono-exponential modelling may add little diagnostic utility to standard MRI for discriminating between various renal tumour entities, lesion conspicuity on DWI was foundis superior to contrast-enhanced T1w imaging because of a higher inherent lesion-to-background signal and less interference of motion artefacts. Advanced DWI techniques, such as kurtosis imaging or IVIM, should be evaluated for imaging paediatric renal neoplasiestumours lesions. DWI is particularly well-suited for paediatricyoung patients for its merits of ultra-fast echo-planar imaging (EPI) acquisition and fast scanning in free-breathing technique without the need of breath-holding, or respiratory triggering and withoutor i.v. contrast application. DWI beingas a "gentle"e and truly non-invasive scanning technique   Figure 43. Same patient as in Figure 32. An additional coronal DWI scan (left image: coronal DWI b=800, inverted greay scale, maximum intensity projection (MIP) -reconstruction) was performed to improve visualisation of the multiple renal lesions. The patient was eventually diagnosed with Wilms tumor (lesion 2) and three large foci of NBS (lesions 1, 3 and 4) in the left kidney, which were all histologically confirmed after neoadjuvant chemotherapy and leftsided nephrectomy. Three small foci (3-5 mm) in the left kidney (lesions 5 and 6, lesion 7 not visible on MIP reconstruction) assumed to present additional manifestations of NBS were no longer visible after neoadjuvant treatment and did not relaps during two years of follow-up.
Overlay of colourised DWI signal on standard sequenzes (right image: colourised coronal DWI b=800 overlay on coronal T2w HASTE, 3D FUSION, Siemens Medical) facilitates co-registration of DWI signal and anatomical background and is useful in communicating MRI findings to referring physicians.