Related Experiment Video
Updated: Jun 24, 2025
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
Risk score model to automatically detect prostate cancer patients by integrating diagnostic parameters
Rossana Castaldo1, Valentina Brancato1, Carlo Cavaliere1
1Bioinformatics and Biostatistics Lab, IRCCS SYNLAB SDN, Naples, Italy.
Introduction:
Prostate cancer (PCa) is one of the prevailing forms of cancer among men. At present, multiparametric MRI is the imaging method for localizing tumors and staging cancer. Radiomics plays a key role and hold potential for PCa detection, reducing the need for unnecessary biopsies, characterizing tumor aggression, and overseeing PCa recurrence post-treatment.
Methods:
Furthermore, the integration of radiomics data with clinical and histopathological data can further enhance the understanding and management of PCa and decrease unnecessary transfers to specialized care for expensive and intrusive biopsies. Therefore, the aim of this study is to develop a risk model score to automatically detect PCa patients by integrating non-invasive diagnostic parameters (radiomics and Prostate-Specific Antigen levels) along with patient's age.
Results:
The proposed approach was evaluated using a dataset of 189 PCa patients who underwent bi-parametric MRI from two centers. Elastic-Net Regularized Generalized Linear Model achieved 91% AUC to automatically detect PCa patients. The model risk score was also used to assess doubt cases of PCa at biopsy and then compared to bi-parametric PI-RADS v2.
Discussion:
This study explored the relative utility of a well-developed risk model by combining radiomics, Prostate-Specific Antigen levels and age for objective and accurate PCa risk stratification and supporting the process of making clinical decisions during follow up.
