Kidney tumors are increasingly detected by accident during imaging performed for unrelated conditions. While up to 70% of small renal masses are ultimately diagnosed as renal cell carcinoma, a significant proportion are benign or very slow-growing and do not require immediate treatment.
The challenge is that conventional imaging such as CT scans and ultrasound cannot reliably distinguish between cancer and benign tumors. As a result, many patients undergo surgery to remove tumors that would never have harmed them, with approximately 25% of removed small tumors found to be benign on pathology.
Currently, renal mass biopsy is the most reliable pre-surgery test for determining tumor type, but it is invasive and carries a 10-15% rate of inconclusive results due to sampling limitations. Tumor heterogeneity (variation within the same tumor) further complicates biopsy interpretation.
There is a strong need for better, non-invasive ways to characterize kidney tumors before deciding on treatment. The goal is a so-called virtual biopsy: imaging tools that can non-invasively determine a tumor's type, grade, and behavior with enough certainty to guide clinical decisions.
This collaborative review examined three main areas of innovation in kidney tumor imaging: multiparametric MRI (mpMRI), which combines multiple MRI sequences to probe tumor biology; molecular imaging using targeted radiopharmaceutical agents that bind to specific tumor proteins; and radiomics plus artificial intelligence (AI), which extract and analyze detailed imaging features beyond what the human eye can perceive.
The review systematically searched the medical literature published over the past 10 years, screening 149 full articles and summarizing the highest-quality evidence on each technology. Evidence was evaluated with particular attention to diagnostic accuracy, clinical feasibility, and proximity to routine clinical use.
Of all the technologies reviewed, two molecular imaging tests emerged as closest to widespread clinical adoption: sestamibi SPECT/CT for identifying benign oncocytomas, and PET/CT with radiolabeled girentuximab for identifying clear cell renal cell carcinoma.
While each technology has its own strengths and limitations, the authors envision that a combination of imaging approaches might eventually provide a comprehensive tumor profile, guiding treatment decisions much as a tissue biopsy does today, but without the needle.
Multiparametric MRI (mpMRI) combines multiple imaging sequences including T1 and T2 weighted images, chemical shift imaging to detect fat, diffusion-weighted imaging (DWI) to assess how densely packed tumor cells are, and dynamic contrast enhancement to evaluate blood vessel patterns within the tumor.
The clear cell likelihood score (ccLS), developed by multiple research groups, uses mpMRI findings to assign a score from 1 to 5 indicating how likely a tumor is to be clear cell renal carcinoma. Studies found that scores of 4 or 5 predicted clear cell cancer with 91% positive predictive value, while scores of 1 or 2 had a 94% negative predictive value.
Using a clinical decision approach where low ccLS tumors are observed, intermediate scores lead to biopsy, and high scores proceed to surgery, researchers estimated that the biopsy rate could be reduced to 20% while keeping the rate of unnecessary removal of benign tumors very low (below 5%).
MRI also shows promise for distinguishing aggressive from non-aggressive tumors. Diffusion-weighted imaging can identify high-grade versus low-grade clear cell carcinoma with moderate accuracy, potentially guiding decisions about urgency of treatment without requiring tissue sampling.
Technetium-99m sestamibi (99mTc-sestamibi) is a radiotracer already widely approved for heart and parathyroid imaging. It accumulates in cells with high numbers of mitochondria and low expression of drug-resistance pumps. Benign kidney tumors called oncocytomas have very high mitochondrial content, making them ideal targets for sestamibi imaging.
Clear cell renal carcinoma, by contrast, has very few mitochondria and high drug-resistance pump expression, so it appears as a dark (photopenic) area on sestamibi imaging while oncocytomas appear as bright (uptake-positive) areas. Using SPECT/CT (single photon emission CT combined with anatomic CT), this contrast allows differentiation of benign from malignant tumors.
A prospective study of 50 patients found sensitivity of 87.5% and specificity of 95.2% for identifying oncocytomas and hybrid oncocytic/chromophobe tumors. External validation studies confirmed these findings. Sestamibi SPECT/CT can be performed just 75 minutes after injection of the radiotracer, making it logistically straightforward.
A cost-effectiveness analysis suggested that using sestamibi SPECT/CT as an initial test, followed by biopsy only to confirm benign findings, would be cost-effective compared to biopsy alone or immediate surgery, while minimizing both unnecessary treatments and missed cancers.
Girentuximab is a monoclonal antibody that binds specifically to carbonic anhydrase IX (CAIX), a protein overexpressed in approximately 95% of clear cell renal cell carcinomas. When radiolabeled and injected, it accumulates in clear cell tumors and can be detected by PET imaging.
The landmark REDECT trial, a prospective multicenter study of 195 patients, found that 124I-labeled girentuximab PET/CT identified clear cell kidney cancers with 86.2% sensitivity and 85.9% specificity, substantially outperforming conventional CT (75.5% sensitivity, 46.8% specificity).
A follow-up study, the ZIRCON trial, is testing a version of girentuximab labeled with zirconium-89 (89Zr), which becomes trapped inside tumor cells after the antibody is internalized, producing a stronger and more durable signal. This trial is underway at 30 sites worldwide.
A limitation of girentuximab imaging is the 5-7 day wait between injection and imaging, because antibodies remain in the bloodstream for a long time. Researchers are exploring smaller molecules targeting the same CAIX protein that could clear the bloodstream faster and allow same-day or next-day imaging.
Radiomics involves extracting hundreds or thousands of quantitative features from medical images, including shape, texture, and signal intensity patterns that are invisible to the human eye. These features can be analyzed individually or fed into machine learning algorithms to build predictive models.
AI models based on deep learning have shown the ability to distinguish benign from malignant kidney tumors, differentiate tumor subtypes, and predict nuclear grade. In one example, a deep learning model combining clinical data with T1 and T2 MRI sequences achieved 70% accuracy in distinguishing benign from malignant tumors, outperforming expert radiologists.
The field of radiogenomics connects imaging features to underlying gene mutations and molecular signatures. Studies have found associations between CT and MRI texture patterns and specific gene mutations (VHL, BAP1, PBRM1) in clear cell carcinoma, potentially allowing non-invasive prediction of molecular tumor characteristics.
Despite impressive results in research settings, widespread adoption of radiomics faces barriers including lack of reproducibility across different scanners and institutions, need for large validated datasets, and the challenge that deep learning models function as black boxes whose predictions may not be explainable in clinical terms.
Multiple promising technologies are converging toward the goal of a comprehensive virtual biopsy: a non-invasive test that can characterize kidney tumors with enough detail to guide clinical decisions without tissue sampling. This could spare many patients from unnecessary biopsies and surgeries.
Among the technologies reviewed, sestamibi SPECT/CT and girentuximab PET/CT are the closest to routine clinical use. Both have prospective trial data supporting their accuracy and are based on approved or near-approved agents with established safety profiles.
AI-powered analysis will likely be necessary to maximize the value of any imaging approach, extracting more information from images than any human reader could. Combining molecular imaging data with radiomics features and clinical information into integrated multimodal models represents the most ambitious and potentially most powerful direction for this field.
For patients and families, these advances mean that the evaluation of a newly found kidney tumor may soon involve sophisticated imaging tests that provide detailed biological information, allowing doctors to confidently recommend active surveillance for benign or indolent tumors while expediting treatment for true cancers.