Decision tree models combining Bi-parametric vesical imaging reporting and data system and apparent diffusion...

Area of Science:

• Radiology
• Oncology
• Medical Imaging

Background:

• Bladder cancer muscle invasion is critical for treatment decisions.
• Accurate assessment of muscle invasion impacts patient outcomes.
• MRI plays a vital role in non-invasive staging of bladder cancer.

Purpose of the Study:

• To compare whole-lesion apparent diffusion coefficient (ADC) histogram analysis with representative ADC values for assessing muscle invasion in bladder cancer.
• To develop a decision tree model integrating bi-parametric Vesical Imaging Reporting and Data System (bp VI-RADS) with ADC values for improved diagnostic accuracy.

Main Methods:

• Retrospective analysis of 82 bladder cancer patients undergoing 3T MRI.
• Scoring of bp VI-RADS using T2-weighted, diffusion-weighted imaging, and ADC maps.
• Whole-lesion ADC histogram analysis and calculation of minimum mean ADC from representative regions of interest.
• Comparison of ADC parameters between muscle-invasive (MIBC) and non-muscle-invasive bladder cancer (NMIBC).
• Diagnostic performance evaluated using ROC curve analysis; combination models built with logistic regression and decision tree analysis.

Main Results:

• The 25th percentile ADC and minimum mean ADC demonstrated similar diagnostic performance, with strong correlation between them.
• Single parameters (bp VI-RADS, 25th percentile ADC, minimum mean ADC) showed comparable accuracy (0.74-0.76).
• Logistic regression models combining bp VI-RADS with ADC parameters achieved higher accuracy (0.87-0.88).
• Decision tree models integrating bp VI-RADS and ADC achieved accuracies of 0.80-0.82, particularly stratifying bp VI-RADS 4 lesions.

Conclusions:

• Both 25th percentile ADC and minimum mean ADC offer similar diagnostic utility for predicting muscle invasion in bladder cancer.
• A decision tree model combining bp VI-RADS with ADC measurements provides an interpretable and clinically applicable tool.
• This combined approach is especially beneficial for refining the assessment of bp VI-RADS 4 lesions.