Radiomics-based machine learning approach for the prediction of grade and stage in upper urinary tract urothelial carcinoma: a step towards virtual biopsy

Abdulsalam Alqahtani ^1,2, Sourav Bhattacharjee ³, Abdulrahman Almopti ¹

⁰School of Medicine, Centre for Medical Engineering and Technology.

International Journal of Surgery (london, England) +

|

May 5, 2024

View abstract on PubMed

Summary

Machine learning models using radiomics data accurately predict upper tract urothelial carcinoma (UTUC) grading and staging. This approach offers a promising, non-invasive alternative to traditional histopathology for cancer diagnostics.

Area Of Science

Medical imaging analysis
Machine learning in oncology
Urologic pathology

Background

Upper tract urothelial carcinoma (UTUC) is aggressive, necessitating early detection.
Accurate tumor grading and staging are crucial for effective UTUC management.
Current diagnostic methods may have limitations in precision and invasiveness.

Purpose Of The Study

To develop machine learning models for predicting UTUC tumor grade and stage.
To utilize computed tomographic urogram (CTU) data and radiomics features.
To compare machine learning predictions against histopathological diagnoses.

Main Methods

CTU data from 106 UTUC patients were analyzed.
3D visualization and digital tumor segmentation were performed.
Radiomics features were extracted and used to train 11 predictive machine learning models.

Main Results

Machine learning models demonstrated satisfactory performance in predicting UTUC grades and stages.
A multilayer panel achieved 84% sensitivity and 93% specificity for grading.
Logistic Regression showed 83% sensitivity and 76% specificity for staging.

Conclusions

Radiomics-based machine learning can effectively model UTUC tumor grading and staging.
These AI tools show potential for improving cancer diagnostics, offering a 'virtual biopsy'.
The findings have implications for clinical practice and advancing diagnostic paradigms.