Study Progress of Radiomics With Machine Learning for Precision Medicine in Bladder Cancer Management

Lingling Ge, Yuntian Chen, Chunyi Yan, Pan Zhao, Peng Zhang, A. Runa, Jiaming LiuFrontiers in Oncology 2019 · DOI

Abstract: Bladder cancer is a fatal cancer that happens in the genitourinary tract with quite high morbidity and mortality annually. The high level of recurrence rate ranging from 50 to 80% makes bladder cancer one of the most challenging and costly diseases to manage. Faced with various problems in existing methods, a recently emerging concept for the measurement of imaging biomarkers and extraction of quantitative features called "radiomics" shows great potential in the application of detection, grading, and follow-up management of bladder cancer. Furthermore, machine-learning (ML) algorithms on the basis of "big data" are fueling the powers of radiomics for bladder cancer monitoring in the era of precision medicine. Currently, the usefulness of the novel combination of radiomics and ML has been demonstrated by a large number of successful cases. It possesses outstanding strengths including non-invasiveness, low cost, and high efficiency, which may serve as a revolution to tumor assessment and emancipate workforce. However, for the extensive clinical application in the future, more efforts should be made to break down the limitations caused by technology deficiencies, inherent problems during the process of radiomic analysis, as well as the quality of present studies.

Introduction: The Challenge of Bladder Cancer Management

Bladder cancer remains one of the most difficult malignancies to manage due to its staggering recurrence rate of 50–80%. The current clinical "gold standard" for monitoring is cystoscopy, an invasive, uncomfortable, and expensive procedure. Furthermore, traditional biopsies and standard radiological assessments often struggle to capture intratumoral heterogeneity—the biological diversity within a tumor that determines its aggressiveness and response to therapy.

There is a critical clinical need for objective, non-invasive tools that can distinguish between Non-Muscle-Invasive Bladder Cancer (NMIBC) and the more aggressive Muscle-Invasive Bladder Cancer (MIBC). Radiomics, combined with machine learning (ML), offers a promising path toward this goal by treating medical images as minable databases rather than just pictures.

The Radiomics Workflow

The transition from raw medical imaging to clinical decision support involves a standardized four-step pipeline designed to extract high-dimensional biomarkers.

1. Image Acquisition and Preprocessing

Data from various modalities—including Computed Tomography Urography (CTU), Multiparametric MRI (mp-MRI), and PET scans—are standardized to ensure consistency across different scanners and protocols. This step is vital for the reproducibility of the quantitative features.

2. VOI Segmentation

The "Volume of Interest" (VOI) must be accurately delineated to ensure the data extracted belongs solely to the tumor. While manual segmentation is common, the field is shifting toward semi-automatic and automatic methods to improve efficiency and reduce inter-observer variability.

Semi-automatic VOI segmentation of a bladder tumor on a CT scan.

3. Feature Extraction

This is the core of the radiomics process. Thousands of quantitative descriptors are extracted, typically falling into four categories:

  • Shape features: Geometric properties like volume and sphericity.
  • First-order statistics: Voxel intensity distributions (e.g., mean, skewness).
  • Second-order (Texture) features: Spatial relationships between pixels that reveal heterogeneity.
  • High-order features: Data processed through mathematical filters (e.g., wavelets) to identify complex patterns.
The four categories of radiomic features: first-order, shape, second-order (texture), and high-order features.

4. Model Building and Validation

ML algorithms, such as Support Vector Machines (SVM) or Deep Learning Convolutional Neural Networks (DL-CNN), are used to select the most predictive features. These models are then validated against clinical outcomes to ensure their accuracy and generalizability.

Key Clinical Insights and Results

Recent studies have demonstrated the transformative potential of this approach:

  • Staging and Grading: One study showed that ML methods could reduce the inaccuracy rate of bladder cancer staging by up to 50%. MRI-based radiomic models have achieved an AUC of 0.9276 in preoperatively grading tumors.
  • Predicting Progression: Radiomics risk models for predicting progression-free intervals have reached an AUC of 0.956, serving as powerful independent prognostic tools.
  • Therapy Response: Deep learning models have improved the accuracy of assessing chemotherapy response compared to manual expert assessments, providing a more objective measure of treatment efficacy.
  • The "Virtual Biopsy": Unlike a needle biopsy, which only samples a small portion of the tumor, radiomics assesses the entire volume, providing a comprehensive view of the tumor’s microenvironment and genetic makeup (radiogenomics).

Limitations and Future Outlook

Despite these milestones, several bottlenecks remain before radiomics can be fully integrated into clinical practice. Most current studies are retrospective and involve small cohorts, leading to risks of overfitting. There is also a lack of universal standardization for image acquisition and feature extraction.

Future efforts must focus on large-scale, multi-center prospective trials and the development of "explainable AI" to ensure clinicians can trust and interpret the models' outputs. As these technologies mature, the integration of radiomics with genomics (radiogenomics) promises to realize the full potential of precision oncology for bladder cancer patients.

Summary updated

Abstract

Bladder cancer is a fatal cancer that happens in the genitourinary tract with quite high morbidity and mortality annually. The high level of recurrence rate ranging from 50 to 80% makes bladder cancer one of the most challenging and costly diseases to manage. Faced with various problems in existing methods, a recently emerging concept for the measurement of imaging biomarkers and extraction of quantitative features called "radiomics" shows great potential in the application of detection, grading, and follow-up management of bladder cancer. Furthermore, machine-learning (ML) algorithms on the basis of "big data" are fueling the powers of radiomics for bladder cancer monitoring in the era of precision medicine. Currently, the usefulness of the novel combination of radiomics and ML has been demonstrated by a large number of successful cases. It possesses outstanding strengths including non-invasiveness, low cost, and high efficiency, which may serve as a revolution to tumor assessment and emancipate workforce. However, for the extensive clinical application in the future, more efforts should be made to break down the limitations caused by technology deficiencies, inherent problems during the process of radiomic analysis, as well as the quality of present studies.
BibTeX
@inproceedings{ge2019study,
  author = {Lingling Ge and Yuntian Chen and Chunyi Yan and Pan Zhao and Peng Zhang and A. Runa and Jiaming Liu},
  title = {{Study Progress of Radiomics With Machine Learning for Precision Medicine in Bladder Cancer Management}},
  booktitle = {Frontiers in Oncology},
  year = {2019},
  doi = {10.3389/fonc.2019.01296},
}
Figure