A General Bayesian Functional Spatial Partitioning Method for Multiple Region Discovery Applied to Prostate Cancer...

Area of Science:

• Medical Imaging
• Computational Statistics
• Oncology

Background:

• Prostate cancer lesion detection using multiparametric magnetic resonance imaging (mpMRI) is subjective and relies heavily on reader expertise.
• Current automated methods lack the ability to accurately estimate the number, size, and location of cancerous lesions.
• Existing spatial partitioning and frequentist methods have limitations in handling complex lesion shapes and quantifying uncertainty.

Purpose of the Study:

• To propose a novel Bayesian functional spatial partitioning method for detecting multiple prostate cancer lesions with an unknown number of lesions.
• To develop a method that models smooth boundary curves for each lesion and quantifies uncertainty in detection.
• To improve the accuracy and reliability of prostate cancer lesion detection from mpMRI data.

Main Methods:

• Utilized a Bayesian functional spatial partitioning approach to model lesion boundaries.
• Employed a Reversible Jump Markov Chain Monte Carlo (RJ-MCMC) framework with novel jump steps.
• Jointly estimated the number of lesions, their boundaries, and spatial parameters, quantifying uncertainty.

Main Results:

• The proposed Bayesian method demonstrated robustness to varying lesion shapes and numbers.
• The method effectively detected multiple prostate cancer lesions using MRI data.
• Quantified uncertainty in the estimation of lesion number, boundaries, and spatial parameters.

Conclusions:

• The novel Bayesian functional spatial partitioning method offers an accurate and reliable approach for multiple prostate cancer lesion detection.
• This method addresses the limitations of current techniques by providing objective estimates and quantifying uncertainty.
• The findings have significant implications for improving prostate cancer diagnosis and treatment planning through advanced MRI analysis.