Imaging for the diagnosis and response assessment of renal tumours

Symptomatic patients typically present with more advanced RCC than patients with incidentally detected tumours [19, 20]. The prevalence of RCC in patients presenting with visible haematuria is 0.9–2.0%; 0.3–1.0% in patients with microscopic haematuria [21, 22]. Computed tomography (CT) urography is recommended as the most appropriate first-line test for patients presenting with unexplained visible haematuria by the AUA and the CUA [9, 10]. However, there is a lack of consensus regarding the optimal investigation of asymptomatic microscopic haematuria in different guidelines worldwide in terms of imaging modality (ultrasound vs. CT urography) and age threshold to prompt investigation [23]. An ultrasound (US) of the renal tract may be considered initially in low-risk young patients with non-visible haematuria as a cost-effective non-ionising technique for assessing the kidneys and bladder [24, 25]. However, US sensitivity is low for renal lesions < 1 cm (26%) [26] and it does not evaluate the collecting systems adequately; hence, CT urography (pre- and post-contrast) is recommended as a first-line tool for patients presenting with non-visible haematuria by a number of American associations [9, 11, 23, 27]. CT urography following contrast administration assesses the entire urinary tract [28] and has a better diagnostic yield for renal cancer than intravenous urography (IVU), with a sensitivity of 100%, specificity of 97.4% and accuracy of 98.3% [29]. Magnetic resonance (MR) urography has the advantage of higher soft tissue contrast compared to CT and it has no radiation burden, but is prone to motion artefacts and limited by MR contraindications and scanner availability [30]. MR urography may be reserved for problem-solving, in pregnancy or when patients have an iodinated contrast allergy or renal failure.

Up to 50% of new cases of renal cancer will be detected incidentally in asymptomatic patients undergoing imaging for an unrelated indication [31]. In fact, over 40% of Medicare insurance beneficiaries in the USA undergo CT of the chest and abdomen over a 5-year period [32]. In a study of 3001 adults undergoing CT colonography, 14.4% of patients had at least one renal mass > 1 cm [33]. However, characterisation of incidentally detected small renal tumours remains a major challenge for imaging.

The Bosniak classification is a well-established system to classify renal cysts based on CT findings [34]. Risk of malignancy is predicted based on the appearance of the cyst wall, number and thickness of septations, calcification and enhancement (Fig. 1) [35]. Simple cysts do not enhance following contrast administration; however CT pseudo-enhancement may occur, where simple cysts appear to demonstrate enhancement < 20 Hounsfield units (HU) due to the CT reconstruction algorithm [36]. Two recent meta-analyses have demonstrated malignancy rates < 6% in Bosniak II to IIF cysts, with rates of over 50% in Bosniak III and approximately 90% in Bosniak IV lesions [35, 37]. EAU guidelines recommend surveillance for Bosniak IIF cysts and operative management for patients with Bosniak III–IV cysts, and therefore differentiating category IIF and III lesions is crucial [5]. However, considerable inter-observer disagreement has been noted in differentiating IIF and III cysts [37]. The effectiveness of the Bosniak classification system for category III cysts has been shown to be suboptimal, leading to considerable operative overtreatment in patients who are found to have benign disease (Fig. 2) [37]. Pitra et al. suggest that MRI may be used for reclassification of Bosniak IIF and III cysts, leading to significant changes in operative management in over a third of cases [38]. Magnetic resonance imaging has superior soft tissue contrast resolution compared to CT, potentially resulting in exaggerated septal thickness and more prominent septal enhancement, and may also upgrade the Bosniak classification in small (< 2.5 cm) cysts [39]. MRI also allows improved visualisation of haemorrhagic cysts, typically hyperintense on T1-weighted sequences, compared to CT [39]. Contrast-enhanced ultrasound (CEUS), involving encapsulated microbubbles of gas injected intravenously as contrast agent has shown to be as effective as CT for renal cyst classification with the Bosniak system [40], and even superior to CT for the detection of malignancy in complex renal cysts [41]. Additionally, CEUS has also been found to have a higher sensitivity and specificity compared to MRI for the differentiation of complex renal cysts and malignancies [42].

Few imaging features are discriminatory in the diagnosis of solid renal masses and approximately 20% of lesions removed at surgery will be benign as a consequence [43]. There is an increasing likelihood of malignancy and higher grade with increasing lesion size [34, 44, 45]. Of benign SRM that are surgically excised, 10–38% are found to be angiomyolipomas and 34–58% are found to be oncocytomas [46‐48]. Ultrasound, contrast-enhanced CT and MRI may aid diagnosis (Table 1). The presence of enhancement, i.e. a change of ≥ 15 HU before and after contrast administration in CT, is considered the most important criterion for the differentiation of malignant solid SRM subtypes, with clear cell RCC enhancing much more compared to chromophobe and papillary RCC [5, 49]. However, reliable differentiation between RCC and oncocytoma and fat-poor angiomyolipoma remains a challenge.

Angiomyolipoma (AML), the most common benign renal tumour, is in most cases characterised by the presence of macroscopic fat (Fig. 3). Therefore, AMLs appear markedly hyperechoic on ultrasound and typically there is low attenuation (− 10 to − 100 HU) on unenhanced CT. On MRI, signal drop on fat-suppressed and opposed-phase T1 sequences is thought to be typical [50, 51]. Some RCCs may contain fat; however, and the presence of calcifications may point towards a diagnosis of malignancy [52]. AMLs may contain minimal fat in approximately 5% of cases and therefore may be difficult to differentiate from RCC. Fat-poor AMLs demonstrate marginally higher attenuation on unenhanced CT than RCC [52]. The combination of homogeneous enhancement and prolonged enhancement pattern on CT resulted in a positive predictive value of 91% and negative predictive value of 87% for the differentiation of fat-poor AML and RCC [50]. A meta-analysis has demonstrated a pooled sensitivity of 0.67 (95% CI 0.48–0.81) and specificity of 0.97 (95% CI 0.89–0.99) for the ability of CT to diagnose fat-poor AML [53].

Differentiating oncocytomas, the second most common benign tumour type, from RCC remains a diagnostic challenge both on imaging and renal biopsy. As a result, oncocytoma is found on 3–4% of nephrectomy pathology specimens [54]. In fact, whilst the positive predictive value of diagnosing malignancy on renal biopsy is > 99%, one in four renal biopsies reported as oncocytoma are found to be RCC following surgical excision [55]. On ultrasound, oncocytomas classically appear as well-circumscribed, homogeneous masses which may be isoechoic or hypoechoic. Oncocytomas are well vascularised; therefore, Doppler ultrasound may show increased vascularity in the periphery of the mass. On CT, these benign tumours often display homogeneous enhancement, and in the absence of calcification, necrosis and haemorrhage [56]. Oncocytomas may display a characteristic central stellate scar in up to one-third of cases, although this may be less evident in smaller lesions [57]. Chromophobe RCCs may contain a central scar, whilst other RCC subtypes that contain central necrosis may be confused for a stellate scar [58]. Segmental enhancement inversion, a tumour segment with lower signal intensity in the arterial phase and higher intensity at the early excretory phase, is considered suggestive of oncocytomas: with a high specificity (87–100%), but discrepancies in reported sensitivity [57‐59].

In one study of biopsy-confirmed renal tumours, CT growth pattern, interface with the parenchyma, presence of a scar, segmental inversion of enhancement, unenhanced CT histogram and pattern of enhancement on triphasic MDCT were studied [49]. This study found that only gradual enhancement was suggestive of a benign pathology; however, it is well known that benign lesions also demonstrate early rapid enhancement, confounding interpretation [60]. For example, benign oncocytomas and malignant clear cell renal cancer share enhancement characteristics due to the dense vascularisation of oncocytomas.

Multiparametric MRI has the potential to improve the characterisation of solid renal tumours, but its diagnostic accuracy and cost-effectiveness are yet to be investigated prospectively [61, 62]. Retrospective studies have used it to differentiate small renal masses that demonstrate a high signal intensity central area on T2-weighted imaging, which may represent either oncocytoma with a central scar or RCC with central necrosis. Complete late enhancement of the central area on gadolinium-enhanced T1-weighted images, typical of fibrosis, was suggestive of oncocytoma, whereas absence of central enhancement (typical of necrosis) or presence of a signal drop on chemical shift imaging (in keeping with the presence of intracellular fat) were used to rule out oncocytoma, as it does not contain either [58]. Galmiche et al. suggest that multiparametric MRI may distinguish between oncocytoma and chromophobe RCC based on enhancement and diffusion characteristics [63]. In a study by Taouli et al., contrast-enhanced MRI performed better than diffusion MRI in differentiating between solid RCC and benign tumours (when excluding fat-rich AML), with a sensitivity of 100% and specificity of 89%; contrast-enhanced MRI combined with diffusion MRI achieved 96% specificity [61]. An important consideration is that since 2007, administration of gadolinium has been avoided in end-stage renal disease or dialysis patients due to increased risk of nephrogenic systemic fibrosis, limiting use [64].

CEUS has been increasingly applied to differentiate small renal masses, due to its ability to detect slow and low flow in the microcirculation [5, 65, 66]. In a meta-analysis of 567 histologically confirmed RCC and 313 benign masses, the pooled sensitivity of CEUS was 88% (85–90 95% CI) and specificity was 80% (75–85% 95% CI) [65]. A recent feasibility study has demonstrated the usefulness of magnetic resonance elastography (MRE), as part of a multiparametric MR imaging protocol, to characterise 21 indeterminate SRM. MRE viscoelastic profile may discriminate between oncocytomas and clear cell RCC, and further prospective studies are warranted [67]. The emerging concept that biomedical images contain ‘hidden’ information about the underlying tissue biology that can be revealed via quantitative image analysis, an approach known as radiomics, has prompted the application of automated image analysis tools in renal tumours [68]. Early retrospective data have suggested that texture analysis of CT images combined with machine learning could distinguish RCC subtypes with an AUC > 0.90, warranting prospective investigation [69]. Molecular imaging with ^99mtechnetium–sestamibi single photon emission computed tomography/computed tomography (^99mTc-MIBI SPECT/CT) has also been used to differentiate oncocytomas and indolent hybrid oncocytic/chromophobe tumours from other more aggressive malignant small renal tumours. The former contain numerous and dense cellular mitochondria and therefore demonstrate “hot” radiotracer uptake relative to the ipsilateral renal parenchyma, whereas the latter appear “cold” [70, 71]. Sheikhbahaei et al. evaluated the diagnostic accuracy of standard pre-operative CT, MRI and ^99mTc-MIBI SPECT/CT in 48 patients with small renal masses undergoing partial or radical nephrectomy. The area under the receiver operator characteristic curve for the differentiation of benign from malignant masses was 0.60 for CT and MRI alone, increasing to 0.85 with the addition of ^99mTc-MIBI SPECT/CT [71]. Quantitative techniques for image analysis may enable increased accuracy and use of ^99mTc-MIBI SPECT/CT in future [72].

Similarly, challenges remain in the imaging-based response assessment of patients with metastatic RCC. An accurate assessment of response to therapy is crucial to guide clinical decisions regarding continuation of treatment. Furthermore, imaging-based progression is routinely used as a surrogate marker for survival in clinical trials; therefore, the development of standardised and validated criteria is key. Reduction in tumour size remains at the core of therapy assessment with contrast-enhanced CT. For example, response assessment is performed by measuring the serial change in tumour size in up to five target lesions for Response Evaluation Criteria in Solid Tumors (RECIST) 1.1 (Fig. 6) [12]. However, in advanced RCC, response assessment based on size change has its limitations for both anti-angiogenic and immunotherapies. Anti-angiogenic therapies have a cytostatic rather than cytotoxic mechanism of action: tumour stabilisation occurs in the majority of cases and size reduction is often less pronounced and late occurring; therefore, RECIST criteria tend to underestimate response. Some tumours may even demonstrate an early increase in size due to necrosis. As a result, a number of alternative response criteria have been proposed to assess radiologic evidence of a reduction in tumour vascularity (Table 2). Such surrogate markers include: a reduction in attenuation, evidence of necrosis (new areas of non-enhancing soft tissue) and changes in the degree and pattern of enhancement [14‐16, 94, 95].

Additionally, a number of emerging techniques have been investigated. CT texture analysis using an image processing algorithm to assess heterogeneity in tumour morphology has been proposed as a marker of response [96]. Functional imaging has been investigated, including dynamic contrast-enhanced (DCE) CT, DCE-MRI, DCE-ultrasound and PET [97]. DCE imaging follows the bio-distribution of a contrast agent injected intravenously and then absorbed into the tumour microcirculation, providing information on the tumour microenvironment and vascularity pre- and post-anti-angiogenic therapy [94]. A radiomic approach synthesising results from a combination of imaging modalities may play a role in future [97]. Recently, Crusz et al. noted that in a group of 27 patients with metastatic RCC receiving tyrosine kinase inhibitors, over 50% of individuals demonstrated heterogeneous responses to therapy (defined as lesions in at least two of the following three response categories: responding, progressing and stable), potentially reflecting molecular intra-tumoral heterogeneity [98]. This may have important implications for clinical decision making, such as the choice to continue therapy in individuals with heterogeneous responses.

Due to their targeted mechanism of action, eliciting an immune response, immunotherapies have been known to lead to “pseudoprogression” or “tumour flare,” with consequent underestimation of the overall survival benefits using the traditional RECIST 1.1 criteria [18]. Partly because the pathophysiological mechanism underlying pseudoprogression is incompletely understood, differentiating this phenomenon from true progression represents an imaging diagnostic challenge with important clinical implications [99]. Four patterns of response to immunotherapy have been described: (1) an initial transient increase in target lesion size followed by durable response; (2) the development of new lesions followed by durable response in target lesions; (3) reduction in target lesion size from the outset; and (4) initially no change in target lesion size followed by a slow reduction in size [17]. To capture this response heterogeneity and variability, the ‘Immune-Related Response Criteria’ (irRC) were first developed, based on a bi-dimensional assessment of target lesion size (Table 3). In contrast to RECIST 1.1, the development of new lesions in irRC is not automatically considered as evidence of progression; rather, the size of the new lesion is added to the total tumour burden. Following irRC, a modified version of RECIST 1.1 for immune-based therapeutics (iRECIST) was proposed, returning to unidimensional assessment of lesions. The main novelty in iRECIST criteria is the differentiation between unconfirmed progressive disease (iUPD) and confirmed progressive disease (iCPD), to reflect the fact that iUPD may indeed represent pseudoprogression [100]. Most recently in 2018, the Immune-modified RECIST (imRECIST) criteria were published, once again utilising unidimensional measurements to improve reproducibility (by reducing measurement variability and error inherent in acquiring two dimensions per lesion). All these sets of criteria require prospective validation. Tumour response patterns may even vary between specific immunotherapy agents, and between tumour types: further response criteria modifications may be inevitable in the future [18].