Volume 33, Issue 1, January 2023, Pages 70-75Author links open overlay panel

Machine learning (ML) and artificial intelligence (AI) have demonstrated potential to improve the care of radiation oncology patients. Here we review recent advances applicable to the care of bladder cancer, with an eye towards studies that may suggest next steps in clinical implementation. Algorithms have been applied to clinical records, pathology, and radiology data to generate accurate predictive models for prognosis and clinical outcomes. AI has also shown increasing utility for auto-contouring and efficient creation of workflows involving multiple treatment plans. As technologies progress towards routine clinical use for bladder cancer patients, we also discuss emerging methods to improve interpretability and reliability of algorithms.

Introduction

Radiation therapy (RT) is a mainstay in the treatment of bladder cancer. Recent developments in machine learning (ML) and artificial intelligence (AI) have shown the potential to improve many aspects of the care of bladder cancer, with applications extending to diagnosis, prognosis, pathology, imaging, contouring, and radiation treatment planning.1,2 Even as challenges remain in validating and implementing AI-enabled solutions for the clinic, new methods are emerging to improve the accuracy and interpretability of algorithms applied to bladder cancer, using a variety of data sources including clinical records, pathology slides, imaging studies, molecular markers, and treatment plans. Furthermore, recent studies of other cancer types may also provide a view into the future of bladder cancer. Within the inherent limitations of retrospective datasets, oftentimes with few samples, cooperative research approaches continue with the aim of pooling data sources and establishing common data models, ontologies, and interoperability standards.3 Here we review recent literature on AI applied to bladder cancer with the aim of forecasting what advances may come next (Fig. 1).

Section snippetsDiagnostic, Prognostic, and Outcome Prediction

AI has been applied to improve estimates of incidence, staging, risk stratification, and outcome prediction for bladder cancer based on clinical data and pathology. Accurate diagnosis and categorization determine the treatment paradigm, for instance non-muscle-invasive (NMIBC) vs muscle-invasive bladder cancer (MIBC) and Ta vs T1.

Across radiation oncology, ML models have allowed for outcome prediction based on both structured and unstructured datasets. In a review of methods applied towards

Radiomics

Extraction of phenotypic characteristics from imaging has led to AI models based on radiomic features.16,17 In bladder cancer, radiomics has been applied to improve the accuracy of staging, grading, classification of NMIBC vs MIBC, and response assessment.18 In one study, pre-operative MRI from 70 patients was analyzed to predict tumor grade, resulting in a multiparametric MRI model with AUC 0.92 that was superior to models based on single modality MRI.19 Such findings may guide choice of

Molecular Biomarkers

ML may also identify biomarkers predictive of outcomes from bladder cancer radiation therapy, including toxicity. One genome-wide association study of late genitourinary toxicity applied preconditioned random forest regression to combinations of single nucleotide polymorphisms (SNPs) from 324 prostate cancer patients, predicting risk of weak stream at 2 years (AUC 0.70) and identifying SNPs associated with biologically plausible pathways.22 Although this study was performed in prostate cancer

Auto-contouring

Automated image segmentation has practical value for radiation oncology. Auto-contouring technologies continue to improve, incorporating techniques such as deep learning, decision forests, and generative adversarial neural networks.3 The development of the U-Net enabled auto-segmentation across multiple imaging modalities by more efficiently using available training data, and reporting localization information of features.24 Early problems noted in atlas-based auto-contouring for prostate

Dosimetry, Treatment Planning, and Quality Assurance

AI models have been developed to improve volumetric dose predictions for genitourinary treatment planning, for instance DoseNet applied to prostate SBRT, which demonstrated that CNNs offer improvements over fully connected methods.29 ML-based radiation treatment planning has been implemented and evaluated in clinical practice for prostate cancer patients.30 Using an extension of random forest-based approaches applied to manually contoured images, the authors showed that ML-based treatment

External Validation and Standardization

Clinical implementation of AI algorithms will require external validation across multiple cohorts, as well as standardization of models and benchmarks. Many AI studies performed to date including some cited here have been conducted on small cohorts featuring a single dataset, and do not reflect the patient and systems variability of the world at large. For instance, the field is currently dominated by studies using data from hospitals located in 3 states - California, Massachusetts, and New

Conclusion

AI and ML are advancing the management of bladder cancer across diagnostic, prognostic, predictive, and treatment planning domains that will continue to shape radiation oncology practice. For clinicians, technological progress will come with gain in some skills, including a working understanding of the strengths and weaknesses of various AI tools, with implications for training.3,50 Much preclinical and clinical research remains to validate early studies, which can be accelerated by

References (50)M.Z. Joel et al.Using adversarial images to assess the robustness of deep learning models trained on diagnostic images in oncology

JCO Clin Cancer Inform

(2022)

A.-M. Noone et al.Machine learning methods to identify missed cases of bladder cancer in population-based registries

JCO Clin Cancer Inform

(2021)

Q. Song et al.A machine learning approach for long-term prognosis of bladder cancer based on clinical and molecular features

AMIA Jt Summits Transl Sci Proc

(2020)

J.C. Hong et al.Predicting emergency visits and hospital admissions during radiation and chemoradiation: An internally validated pretreatment machine learning algorithm

JCO Clin Cancer Inform

(2018)

R. Gupta et al.The emergence of pathomics

Curr Pathobiol Rep

(2019)

S. Chen et al.Clinical use of machine learning-based pathomics signature for diagnosis and survival prediction of bladder cancer

Cancer Sci

(2021)

P.-N. Yin et al.Histopathological distinction of non-invasive and invasive bladder cancers using machine learning approaches

BMC Med Inform Decis Mak

(2020)

N. Brieu et al.Automated tumour budding quantification by machine learning augments TNM staging in muscle-invasive bladder cancer prognosis

Sci Rep

(2019)

N. Tokuyama et al.Prediction of non-muscle invasive bladder cancer recurrence using machine learning of quantitative nuclear features

Mod Pathol

(2021)

E. Chang et al.Comparison of radiomic feature aggregation methods for patients with multiple tumors

Sci Rep

(2021)

L. Ge et al.Study progress of radiomics with machine learning for precision medicine in bladder cancer management

Front Oncol

(2019)

H. Wang et al.Radiomics analysis of multiparametric MRI for the preoperative evaluation of pathological grade in bladder cancer tumors

Eur Radiol

(2019)

S.S. Garapati et al.Urinary bladder cancer staging in CT urography using machine learning

Med Phys

(2017)

Y. Zhang et al.Prognostic value of transfer learning based features in resectable pancreatic ductal adenocarcinoma

Front Artif Intell

(2020)

V.L. Kouznetsova et al.Recognition of early and late stages of bladder cancer using metabolites and machine learning

Metabolomics

(2019)

View full text

© 2022 Elsevier Inc. All rights reserved.