Artificial Intelligence in Bladder Cancer: Current Trends and Future Possibilities

Bladder cancer is the most common malignant tumor of the urinary system in the Asian population, with a particularly high prevalence in China. Each year, approximately 80,000 new cases are reported in China, and the incidence continues to rise. Among these cases, 75% are classified as non-muscle invasive bladder cancer (NMIBC), while the remaining 25% are muscle invasive bladder cancer (MIBC). For both NMIBC and MIBC patients, radical cystectomy is often considered a primary treatment approach due to the higher risk of tumor progression. However, the diagnosis and staging of bladder cancer rely heavily on cystoscopy and traditional transurethral resection, which are associated with significant risks, including complications and high peri-operative mortality. Additionally, cystoscopy has a critical limitation: its inability to efficiently differentiate between malignant tumors and healthy urothelium, particularly in cases of multifocal disease or inconspicuous but significant lesions such as carcinoma in situ.

Artificial intelligence (AI) has emerged as a promising tool to address these challenges in bladder cancer diagnosis and treatment. By leveraging AI, particularly machine learning (ML) and deep learning (DL), clinicians can achieve more accurate and noninvasive methods for identifying and managing bladder cancer. One of the most significant advancements is the use of 3D image-based features derived from computed tomography (CT) or magnetic resonance imaging (MRI) to identify the distribution of heterogeneous tumors. Machine learning algorithms, when integrated with 3D image-based MRI, can differentiate between low-grade and high-grade tumors, enabling physicians to perform less invasive surgical procedures. This approach has been shown to reduce intraoperative blood loss, shorten hospital stays, accelerate gastrointestinal function recovery, and reduce overall complications.

In the context of MIBC, machine learning has been implemented to quantify tumor buds in immunofluorescence-labeled slides. Tumor budding is a critical factor correlated with the tumor-node-metastasis (TNM) staging system. Based on this, MIBC patients have been classified into three novel staging standards, which are determined by their disease-specific survival rates. This classification provides a more nuanced understanding of disease progression and prognosis. Furthermore, quantitative analysis of tumor buds using automated slide analysis offers an alternative staging model with significant prognostic value for MIBC patients.

Machine learning algorithms have also been employed to build predictive models for recurrence rates and survival times based on imaging and surgical data. These models have demonstrated high sensitivity and specificity, with both metrics exceeding 70% for predicting recurrence and survival rates at 1, 3, and 5 years post-cystectomy. By optimizing surgical data acquisition, these predictive models can assist physicians in developing personalized follow-up plans, improving treatment strategies, and enhancing overall patient care. Additionally, integrating the genome-wide atlas of frozen NMIBC specimens into genetic programming algorithms can generate mathematical model classifiers for outcome prediction, further refining prognostic accuracy.

Despite these advancements, the adoption of machine learning and deep learning models in clinical practice remains limited. Several challenges hinder their widespread implementation. One of the primary obstacles is the need for standardized parameters and the adjustment of device differences to ensure the universality of these models. Additionally, collecting data from multiple institutions is essential to enhance the robustness and generalizability of AI models. Furthermore, the formulation of diagnosis and treatment plans is often based on evidence-based scientific papers and physicians’ clinical experience. However, the development of guidelines for diagnosing and treating bladder cancer in many countries is hindered by poor data research, particularly in terms of population-specific references. Differences in ethnicity and other factors further reduce the reference value of these studies.

To overcome these challenges, machine learning and deep learning predictions should be further incorporated into data and model retraining processes, particularly in the context of personalized patient care. By resolving these issues, AI models can be trained using comprehensive bladder cancer datasets, including pre-operative, intra-operative, and post-operative data, to accurately predict each patient’s prognosis. Extensive patient datasets and electronic medical records can be semi-automated to provide real-time predictive analysis, enhancing the understanding of various disease processes. However, the accuracy of these predictions depends heavily on effective data integration from different sources. While these models will not replace shared decision-making, they can complement the information obtained from traditional methods, providing additional insights for clinicians and patients.

Healthcare organizations and technology companies must collaborate to encourage the collection of data and the execution of AI-based research. This collaboration can enable physicians to provide more reliable clinical and evidence-based treatment for their patients. The application of AI in the healthcare system is still in its initial phase, but it holds tremendous potential in treatment measurements and the diagnosis of bladder cancer. Similar to all medical interventions, the application of machine learning for cancer diagnosis has both advantages and disadvantages. While AI can improve the speed and consistency of diagnosis, it may also exacerbate overdiagnosis. Machine learning cannot solve the gold standard problem, but it can further expose the predicament of intermediate categories between “cancer” and “noncancer.” Ultimately, what matters to patients and physicians is whether the diagnosis of bladder cancer is related to the length and quality of life.

Before AI technology is widely adopted in clinical practice, serious consideration should be given to training machine learning algorithms to identify intermediate categories between “cancer” and “noncancer.” The application of AI in the clinical field will continue to develop and modify the diagnosis and decision-making process. Physicians’ experience and evidence-based scientific papers will remain vital in ensuring that AI systems operate as expected and provide accurate and reliable outcomes. The integration of AI into bladder cancer diagnosis and treatment represents a significant step forward, offering the potential to improve patient outcomes and enhance the overall quality of care.

In conclusion, artificial intelligence has the potential to revolutionize the diagnosis and treatment of bladder cancer. By leveraging machine learning and deep learning algorithms, clinicians can achieve more accurate and noninvasive methods for identifying and managing the disease. However, several challenges must be addressed before AI can be widely adopted in clinical practice. These include standardizing parameters, adjusting device differences, and collecting data from multiple institutions. With continued research and collaboration between healthcare organizations and technology companies, AI can become an invaluable tool in the fight against bladder cancer, ultimately improving patient outcomes and enhancing the quality of care.

doi.org/10.1097/CM9.0000000000001830

Was this helpful?

YesNo

0 / 0