Novel liquid immunocytochemistry with machine learning analysis for bladder cancer detection

Bladder cancer diagnosis is challenged by invasive monitoring and workflow inefficiencies impacting diagnostic reliability. This prospective study enrolled 150 patients (January 2020-December 2022) and evaluated a novel liquid-based immunocytochemistry platform, coupled with integrated machine learning, for detecting urothelial carcinoma in voided urine. All 150 cytology slides met the adequacy threshold of ≥2,644 urothelial cells and showed preserved cytomorphology. Eight cases of papillary urothelial neoplasm of low malignant potential (PUNLMP) were set aside a-priori, yielding an analytic cohort of 142 patients (115 urothelial-carcinoma, 27 benign) for performance analysis. hTERT (sensitivity 92.2%, specificity 66.7%), GATA-3 (67.0%, 88.9%), and CK17 (89.6%, 66.7%). In multi-marker analysis, sensitivity reached 100% (95% CI 96.8-100) when any marker was positive, whereas specificity reached 100% (95% CI 87.3-100) when all three markers were positive. The workflow-optimized platform standardizes specimen preparation and multi-marker interpretation, offering a robust foundation for urine-based bladder-cancer diagnostics. Larger, multi-center validation studies are warranted to refine specificity estimates and facilitate laboratory integration. This study demonstrates that addressing fundamental workflow challenges in bladder cancer diagnostics before implementing artificial intelligence creates more effective diagnostic tools. By prioritizing specimen integrity and standardization through a novel liquid immunocytochemistry platform, exceptional diagnostic performance was achieved with 100% sensitivity and specificity under defined marker parameters across various cancer stages. This workflow-first approach to integrating machine learning with advanced biomarker analysis offers a model for developing clinically practical diagnostic innovations that can reduce reliance on invasive monitoring procedures while improving detection accuracy.