AI predicting recurrence in non-muscle-invasive bladder cancer: systematic review with study strengths and weaknesses

Saram Abbas ¹, Rishad Shafik ¹, Naeem Soomro ²

⁰School of Engineering, Newcastle University, Newcastle Upon Tyne, United Kingdom.

Frontiers in Oncology +

|

January 22, 2025

View abstract on PubMed

Summary

Machine learning (ML) and artificial intelligence (AI) show promise for predicting non-muscle-invasive bladder cancer (NMIBC) recurrence, outperforming traditional methods. Further research is needed to improve model generalisability and clinical integration for better patient management.

Area Of Science

Oncological Urology
Medical Artificial Intelligence
Precision Oncology

Background

Non-muscle-invasive bladder cancer (NMIBC) has a high recurrence rate (70-80%), leading to significant patient burden and high management costs.
Current NMIBC recurrence prediction tools often overestimate risk and lack accuracy.
Machine learning (ML) and artificial intelligence (AI) offer enhanced predictive precision by integrating molecular and clinical data.

Purpose Of The Study

To critically review ML-based frameworks for predicting NMIBC recurrence.
To analyze the statistical robustness and algorithmic efficacy of existing ML models.
To identify strengths, weaknesses, and limitations of ML applications in NMIBC recurrence prediction.

Main Methods

Systematic literature search focusing on ML frameworks for NMIBC recurrence prediction.
Categorization of studies by data modalities (radiomics, clinical, histopathological, genomic) and ML model types (neural networks, deep learning, random forests).
Analysis of studies for statistical robustness, algorithmic efficacy, performance metrics, generalisability, interpretability, and cost-effectiveness.

Main Results

ML algorithms show significant potential, with neural networks achieving 65-97.5% accuracy, especially with multi-modal datasets.
Multi-modal data integration consistently outperforms single-modality approaches.
Challenges include limited generalisability due to small datasets and the 'black-box' nature of advanced models; explainability methods like SHAP show promise but need refinement.

Conclusions

ML and AI hold substantial promise for improving NMIBC recurrence prediction and advancing precision oncology.
Optimizing multimodal data utilization and enhancing model explainability are crucial for clinical adoption.
Further research and robust datasets are necessary to translate ML advancements into tangible patient benefits.

Keywords:

NMIBC artificial intelligence machine learning non-muscle-invasive bladder cancer prediction recurrence