AI Shows Promise in Detecting Bladder Tumors During Cystoscopy

A urologist threading a thin camera into a patient's bladder is looking for tumors—small growths that can hide in the folds and shadows of the organ's interior. It's painstaking work. The human eye, even trained and experienced, can miss lesions. Some slip past during the initial examination. Some are caught only after symptoms worsen. Now researchers including Salvador Jaime-Casas, Amir Khan, and Benjamin Chung have demonstrated that artificial intelligence can serve as a second set of eyes during these procedures, flagging suspicious tissue in real time and potentially catching cancers that might otherwise go undetected.

Bladder cancer kills thousands of Americans each year, and early detection is one of the few reliable levers for improving survival. The standard diagnostic tool is cystoscopy—a procedure in which a physician inserts a flexible endoscope through the urethra to visualize the bladder's interior directly. The doctor looks for abnormal growths, takes biopsies, and sometimes removes small tumors on the spot. But the procedure's effectiveness depends entirely on what the physician sees. Fatigue, lighting conditions, the angle of approach, the sheer difficulty of scanning every millimeter of a hollow organ—all of these can lead to missed lesions. Studies suggest that a meaningful fraction of early-stage tumors are overlooked on first examination.

The research team trained machine learning models to recognize the visual signatures of bladder tumors in cystoscopy video feeds. The AI system learned to distinguish between normal urothelium—the tissue lining the bladder—and malignant or pre-malignant growths. During testing, the algorithm demonstrated the ability to identify tumors that might escape notice during routine examination. The implications are straightforward: if an AI system can reliably flag suspicious areas in real time, the physician can investigate further, take additional biopsies, or remove tissue that might otherwise have been left behind.

What makes this work significant is not that it replaces the urologist—it doesn't, and it shouldn't. Rather, it augments human judgment at the moment when judgment matters most. A surgeon performing a cystoscopy is managing multiple cognitive tasks simultaneously: navigating the instrument, interpreting visual information, deciding whether a lesion warrants biopsy or removal, and managing the patient's comfort. An AI system that reliably highlights areas of concern reduces the cognitive load and increases the likelihood that subtle abnormalities will be caught.

The research by Jaime-Casas, Khan, and Chung points toward a future in which AI-assisted cystoscopy becomes standard practice in urology clinics and hospitals. Wider adoption could reshape how bladder cancer is detected and treated. Fewer missed diagnoses mean earlier intervention, which typically translates to better outcomes and less aggressive treatment. For patients, it could mean the difference between catching a cancer at a stage when it can be managed with transurethral resection versus discovering it only after it has invaded the muscle wall and requires more extensive surgery or systemic therapy.

The work is still in the research phase, and questions remain about how the technology will perform across different patient populations, different cystoscopes, and different clinical settings. But the promise is real. As these systems mature and become integrated into clinical workflows, they may become as routine as the endoscope itself—a tool that helps physicians see what they might otherwise miss.