The 2023 AUA Annual Meeting featured a wide array of abstracts and presentations reporting on recent technological advances, particularly in the area of imaging. Thirty-two abstracts were presented at 2 imaging-specific moderated poster and podium sessions, and dozens of others were included across sessions among the various urological subspecialties and societies. Several trends emerged at the conference, and although we regretfully are unable to discuss them all, there are several we’d like to highlight.

Novel Molecular Imaging

Results from the phase 3 ZIRCON Study were presented by Dr Brian Shuch at the late-breaking abstract session on behalf of an international group of collaborators. This open-label, multicenter study examined the use of positron emission tomography (PET) imaging with radiolabeled girentuximab, a monoclonal antibody targeting carbonic anhydrase IX, an enzyme highly expressed in clear cell (cc) renal carcinoma (RCC). The investigators aimed to use ⁸⁹Zr-DFO-girentuximab PET/CT as a means of accurately and noninvasively delineating the pathology of indeterminate renal masses with the goal of reducing the need for unnecessary surgical intervention among those with benign or indolent masses. The coprimary end points were the sensitivity and specificity of PET imaging to diagnose ccRCC with the gold standard comparator being histopathology from planned surgery. Among almost 300 patients, sensitivity and specificity were both greater than 85% in detecting ccRCC. Among those who were PET positive but ccRCC negative on final pathology, most patients had papillary RCC, with 1 patient each found to have sarcoma and oncocytoma. These promising findings highlight the potential application of ⁸⁹Zr-DFO-girentuximab PET/CT to better characterize indeterminate renal masses prior to surgical intervention.

Several other presentations this year introduced novel molecular imaging techniques, including Dr Escobar and colleagues (PD09-12), who reported their early experience with 10 patients who underwent gallium-FAP (fibroblast activation protein) PET imaging for localized bladder cancer treated with curative intent. FAP-PET demonstrated 100% concordance with true pathological disease status. Pretreatment imaging findings changed management in 2 patients: 1 with metastatic disease that was not detected by conventional imaging but was discovered with FAP-PET, and 1 whose conventional imaging was concerning for regional lymph node involvement, but FAP-PET scan, later confirmed by lymph node biopsy, was conversely negative for metastases.

Radiomics

Radiomics, the use of a computerized approach to evaluate medical images, has been shown to aid human-based diagnostic imaging, with the potential to uncover tumor patterns and characteristics that fail to be appreciated by the naked eye. As such, key applications for radiomics may include improvements to standard diagnostic and risk-stratification techniques when combined with clinical and histopathological data. Such utility was demonstrated in the realm of prostate cancer by Saunders et al (PD22-03), who trained a deep learning classification model to enhance differentiation between tumor and benign cofounders on MRI. Their approach eliminated more than 70% of false-positive lesions while maintaining 91% sensitivity, with the ultimate clinical aim of reducing unnecessary biopsies. Investigators at the National Cancer Institute (MP09-02) reported the disease-specific radiomics features of cribriform architecture on prostate MRI with concordance at the histological level, highlighting a role for radiomic evaluation to assist in risk stratification for men with this potentially aggressive form of prostate cancer. Furthermore, a technique known as δ radiomics (PD33-04) was employed in muscle-invasive bladder cancer to analyze feature variation at multiple acquisition time points on MRI, improving predictability of complete response to trimodal therapy. While these studies are preliminary, they are proof of concept that radiomics may serve as an important adjunct to human-read images and could be used in the near future to impact clinical decision-making.

Artificial Intelligence and Machine Learning

Artificial intelligence (AI) and machine learning (ML) have shown promise in improving workflow efficiency by assisting radiologists and clinicians with the more labor-intensive aspects of image interpretation. With the ability to streamline processing of large data sets, these methods may improve diagnostics by standardizing approaches and limiting subjectivity. Uhlig et al (PD22-07) demonstrated that with AI-based diagnostics, a trained neural network significantly reduced the need for human-based manual segmentation of renal tumors. Similarly, a group at Cleveland Clinic (MP47-15) proposed a scalable AI-based R.E.N.A.L. (for radius, exophytic/endophytic, nearness of tumor to collecting system, anterior/posterior, location relative to polar line) nephrometry scoring system. This approach aims to standardize the nephrometry assessment of small renal masses used to predict meaningful oncologic and surgical outcomes. Khanna et al (PD08-04) developed a radiomics and ML model that reached 90% accuracy in distinguishing benign oncocytoma from malignant neoplasm based on CT images alone, positing the implementation of a virtual renal mass biopsy to avoid procedural risks. Two separate AI models (PD08-06, PD43-06) were shown to accurately predict postoperative glomerular filtration rate in partial nephrectomies that can aid in preoperative patient counseling and surgical planning. Moreover, Dr Cacciamani and colleagues (MP55-18) reported the application of an ML framework to designate an accurate PI-RADS (Prostate Imaging Reporting & Data System) score to patients with a single MRI lesion comparable to human radiologists. With fair concordance and discrimination ability reported, the authors demonstrated that this standardized algorithm could eliminate existing interreader variability and improve overall quality of prostate MRI readings.

Innovations in Surgical Planning and Radiation Safety

Immersive virtual reality (iVR) was also introduced as a possible adjunctive tool at AUA2023. Cumpanas and colleagues (MP09-12) studied the impact of a preoperative iVR before percutaneous nephrolithotomy, with use of the iVR model improving stone-free rates and decreasing postoperative complications. Interestingly, 1 in 3 surgeons changed their calyceal access site with use of the model. Additionally, Clark et al reported on radiation exposure in pregnant surgeons during percutaneous nephrolithotomy, with the utilization of low-dose and pulsed fluoroscopy in combination with wearing appropriate lead protection increasing the number of cases needed to reach unsafe fetal radiation exposure levels from 12 to 11,700.

Conclusions

As the field of urology continues to progress, further studies into the dissemination and implementation of new technologies across a range of clinical settings must be performed. While developments in imaging modalities have vast potential to improve patient counseling and surgical planning, understanding the influence these diagnostics have on stage migration, treatment appropriateness, and overall prognosis remains paramount.