Application of an unsupervised multi-characteristic framework for intermediate-high risk prostate cancer localization...

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Prostate cancer diagnosis relies on imaging, but accurate tumor localization remains challenging.
• Diffusion-weighted magnetic resonance imaging (DW-MRI) offers valuable insights into tissue microstructure.

Purpose of the Study:

• To propose an unsupervised framework for tumor localization by combining multiple DW-MRI parameters.
• To evaluate the framework's efficacy in identifying prostate cancer lesions.

Main Methods:

• A voxel-by-voxel analysis of DW-MRI was performed using Intravoxel Incoherent Motion (IVIM) and kurtosis models.
• Parametric maps for D*, D, f, and K were generated for ten patients with moderate or high-risk prostate cancer.
• A multi-parameter combination strategy was employed in a "positive-if-all-positive" manner.

Main Results:

• The framework detected 24 lesions, with 14 true positives, achieving a 58% tumor detection rate and 100% sensitivity on a lesion basis.
• Compared to multiparametric MRI (mpMRI) with PIRADSv2, the framework showed higher sensitivity (100% vs. 86%) and comparable positive predictive value.
• Index lesions were consistently visible on framework-derived maps, identified as most suspicious in 90% of patients.

Conclusions:

• The proposed unsupervised framework demonstrates high patient-based sensitivity for detecting aggressive prostate cancer index lesions.
• This approach shows promise for improving tumor localization accuracy in prostate cancer diagnostics.