Dr Li on Progression Risk in Low-Grade NMIBC and the Emerging Role of AI-Based Pathology Assessment

“Typically, when we talk about progression, it's really progressing to muscle-invasive disease and metastatic disease. There are very few patients [who’ have progressive events from low grade—I would say probably less than 5%—and when you're looking at large cohorts, usually out of several hundred patients, there are only 1 or 2 patients who have progressive events to muscle-invasive and metastatic disease. However, I do think that there is a lot more events of progression to high-grade, invasive disease.”

Roger Li, MD, a genitourinary oncologist at Moffitt Cancer Center, discussed the risk of progression among patients with low-grade non–muscle-invasive bladder cancer (NMIBC) and the potential role of artificial intelligence (AI) in improving risk stratification.

He emphasized that although low-grade NMIBC is generally associated with favorable oncologic outcomes, progression does occur along a spectrum, and its clinical implications vary depending on whether progression is to high-grade disease, muscle-invasive disease, or metastatic disease.

Li noted that true progression from low-grade NMIBC to muscle-invasive or metastatic bladder cancer is rare, typically occurring in fewer than 5% of patients. In large cohorts, he explained, only a few patients may experience progression to muscle-invasive disease. However, this rarity should not obscure the more frequent occurrence of progression from low-grade to high-grade disease, which carries distinct management implications. He estimated that 10% to 20% of patients with low-grade NMIBC may progress to high-grade disease, although he acknowledged variability in grading due to subjective interpretation among pathologists.

Because grading discrepancies and subtle histologic features can complicate risk assessment, Li highlighted the growing interest in AI-based pathology tools. He explained that AI models trained on digitized hematoxylin and eosin (H&E) slides offer a practical and scalable approach, requiring only standard pathology images that are already available in routine clinical practice. Unlike genomic or molecular assays, these tools do not depend on specialized sequencing platforms, making them potentially accessible across diverse clinical settings, including community urology practices.

Li underscored that AI platforms can interrogate nuclear and cellular features at a scale that exceeds human capability. By analyzing thousands of morphologic parameters that may not be readily discernible to pathologists, AI could identify patterns associated with clinically meaningful end points, such as the likelihood of progression to high-grade disease. This level of granularity, he stated, allows AI to capture biologic signals that could refine prognostication and improve early risk identification.

If validated prospectively, AI-driven pathology assessment could support more personalized surveillance and treatment strategies in low-grade NMIBC. Patients identified as having higher-risk morphologic signatures might benefit from intensified monitoring or earlier therapeutic intervention, while those with lower-risk profiles could avoid unnecessary procedures or overtreatment.

Li concluded that AI-enabled pathology has the potential to enhance clinical decision-making by offering objective, reproducible, and widely deployable risk stratification tools for patients with NMIBC.