Bladder cancer is one of the most common urologic malignancies worldwide, with an estimated 614,000 new cases and 220,000 deaths reported in 2022 (Sung et al., 2021). Despite advances in surgery, intravesical therapies, chemotherapy, and immunotherapy, clinical outcomes remain suboptimal, particularly in muscle-invasive and metastatic disease. Delays in detection, limited sensitivity of current diagnostic tools, and challenges in predicting recurrence and treatment response continue to complicate patient care.
Artificial intelligence (AI) is rapidly transforming oncology, and AI in bladder cancer has emerged as a promising tool to enhance detection, diagnosis, prediction, and treatment. By leveraging large-scale datasets—including cystoscopy images, radiology, pathology, genomics, and electronic health records—AI provides clinicians with data-driven insights that improve accuracy, efficiency, and personalization of care.
This article explores how AI is applied in bladder cancer across four key domains: detection, diagnosis, predictions, and treatment.
Read About AI in Oncology on OncoDaily
AI in Bladder Cancer Detection
Early detection of bladder cancer is critical to improving survival and reducing recurrence. The gold standard diagnostic approach remains cystoscopy, supplemented by urine cytology. However, cystoscopy is invasive, operator-dependent, and costly, while cytology suffers from low sensitivity, particularly for low-grade tumors.
AI-enhanced cystoscopy has shown promise in addressing these limitations. Deep learning models applied to cystoscopic video sequences can automatically detect suspicious lesions, highlight them in real time, and assist urologists in identifying flat lesions such as carcinoma in situ (Shkolyar et al., 2019). This not only improves sensitivity but also reduces interobserver variability.
Similarly, AI applied to urine-based diagnostics is advancing non-invasive detection. Machine learning algorithms analyzing molecular biomarkers (DNA methylation, exosomal RNA, or protein signatures) can distinguish malignant from benign samples with higher sensitivity than cytology. For example, support vector machine (SVM)-based models have demonstrated strong performance in predicting bladder cancer from urinary biomarkers (Ward et al., 2016).
AI-driven radiomics is also improving early detection by extracting quantitative features from CT urography or MRI, enabling the differentiation of bladder tumors from benign conditions and characterizing tumor aggressiveness.
AI in Bladder Cancer Diagnosis
Accurate diagnosis and staging are essential for treatment planning. Conventional histopathology, while effective, is subject to variability, and assessing molecular features requires additional resources.
AI in pathology has been transformative. Deep learning algorithms trained on digital histology slides can identify bladder cancer cells, grade tumors, and quantify tumor-infiltrating lymphocytes (Cheng et al., 2020). These models can standardize pathology reporting and support pathologists in busy clinical settings.
Moreover, AI is increasingly used for molecular profiling. Radiogenomics links imaging features with underlying genetic alterations, allowing prediction of actionable mutations non-invasively. For instance, radiomic models have been developed to predict FGFR3 mutation status in bladder cancer, a critical marker for targeted therapy selection.
Urinary cytology, traditionally plagued by limited sensitivity, also benefits from AI integration. Computer-aided cytology platforms have shown higher accuracy in detecting urothelial carcinoma compared to human observers alone, particularly for high-grade disease.
Read About AI in Cancer Drug Discovery on OncoDaily
AI in Bladder Cancer Predictions
Bladder cancer is characterized by a high recurrence rate, with non-muscle-invasive bladder cancer (NMIBC) patients facing up to a 70% chance of recurrence within five years (Babjuk et al., 2019). Predicting which patients are at higher risk of recurrence, progression, or treatment resistance is vital for tailoring surveillance and therapy.
Machine learning models trained on large clinical datasets can predict recurrence and progression more accurately than traditional risk calculators such as EORTC or CUETO. These models incorporate diverse variables, including demographics, tumor characteristics, treatment history, and molecular biomarkers.
For example, ML-based prognostic tools have been developed to predict BCG (bacillus Calmette-Guérin) therapy response. By integrating clinical features and genomic profiles, AI models can identify patients unlikely to benefit from BCG, allowing earlier transition to alternative treatments (Liu et al., 2021).
AI also predicts survival outcomes in advanced disease. Deep learning applied to CT or MRI scans can extract tumor features correlated with progression-free and overall survival, enabling risk stratification in metastatic bladder cancer patients.
AI in Bladder Cancer Treatment
Treatment of bladder cancer ranges from surgery and intravesical therapy to systemic chemotherapy, immunotherapy, and antibody-drug conjugates. AI plays an expanding role in optimizing therapeutic decisions, improving outcomes, and reducing toxicity.
Surgical management
AI supports surgical planning by analyzing preoperative imaging to assess tumor depth, location, and resectability. In robotic-assisted radical cystectomy, AI-guided systems are being developed to provide intraoperative navigation, improving precision and reducing complications.
Intravesical therapy optimization
For NMIBC, BCG remains the standard of care, but response varies. AI algorithms can analyze patient-specific features to personalize intravesical therapy schedules, reducing recurrence while minimizing overtreatment.
Systemic therapy and immunotherapy
Checkpoint inhibitors and targeted therapies have reshaped the metastatic bladder cancer landscape. AI models predict response to immune checkpoint inhibitors by integrating PD-L1 expression, tumor mutational burden, and radiomic features. Studies suggest that radiomics-based biomarkers may outperform PD-L1 alone in predicting immunotherapy benefit (Sun et al., 2018).
In precision oncology, AI-driven molecular profiling tools identify targetable alterations such as FGFR3 mutations, ERBB2 amplifications, or DNA repair deficiencies. These insights guide the selection of targeted therapies or enrollment in clinical trials.
Radiotherapy personalization
AI assists in radiotherapy planning by automating tumor contouring, optimizing dose distributions, and predicting toxicities. Machine learning models trained on historical radiotherapy data can personalize treatment while minimizing adverse effects on surrounding tissues.
Drug discovery in bladder cancer
AI platforms accelerate drug discovery by designing molecules targeting bladder cancer-specific pathways. Insilico Medicine and other biotech firms are applying deep learning for de novo drug design, potentially shortening timelines for developing novel therapies.
Challenges and Limitations
Despite its promise, AI in bladder cancer faces significant challenges.
- Data limitations: Most models are trained on small, single-institution datasets, limiting generalizability.
- Interpretability: Many AI models operate as “black boxes,” raising concerns among clinicians and regulators.
- Integration into clinical workflows: Adoption requires infrastructure, training, and regulatory approval.
- Bias and equity: Underrepresentation of certain populations in training datasets risks perpetuating disparities in bladder cancer care.
Robust validation through multicenter trials and explainable AI approaches will be crucial for wider adoption.
Read About Can AI Cure Cancer on OncoDaily
The Future of AI in Bladder Cancer
The future of AI in bladder cancer lies in multimodal integration—combining imaging, pathology, genomics, and clinical data into unified predictive models. Digital twins, virtual representations of patients, may allow simulation of therapeutic strategies before real-world application.
Collaborations between academic institutions, industry, and regulatory agencies will accelerate clinical translation. Federated learning approaches, which allow AI training across institutions without sharing raw patient data, may overcome privacy barriers and expand dataset diversity.
Ultimately, AI will not replace clinicians but will augment their decision-making, enabling earlier detection, more accurate diagnosis, personalized predictions, and optimized treatments.
Conclusion
Artificial intelligence is poised to transform bladder cancer management across the spectrum of care. From improving detection through AI-enhanced cystoscopy and cytology, to enhancing diagnosis with digital pathology and radiogenomics, predicting recurrence and survival, and guiding treatment personalization, AI holds the potential to improve outcomes and efficiency in bladder cancer care.
As validation studies progress and integration challenges are addressed, AI in bladder cancer will become a cornerstone of precision oncology, offering patients more timely, accurate, and effective interventions.
You Can Watch More on OncoDaily Youtube TV
Written by Armen Gevorgyan, MD