Improving Prostate Lesion Classification and Diagnostic Accuracy Using Machine Learning: A Comprehensive Evaluation and Development of a PI-RADS 3 Classifier
Overview
- Phase
- Not Applicable
- Intervention
- Not specified
- Conditions
- Prostate Cancer
- Sponsor
- Paracelsus Medical University
- Enrollment
- 173
- Locations
- 1
- Primary Endpoint
- Quantitative Signal - Intensity - Measurements with Region of Interest in specific in high b-value (800, 1500, 4000) axial MRI Images
- Status
- Completed
- Last Updated
- 2 years ago
Overview
Brief Summary
The investigators propose an AI methodology combining machine learning, histological results and expert image interpretation for the development of a PI-RADS 3 classifier.
Detailed Description
Prostate cancer is the most common carcinoma in male patients in Western industrialized countries. Multiparametric prostate MRI (mpMRI) can select patients who may be potential candidates for biopsy. In this study, the investigators present a comprehensive methodology that evaluates a multitude of AI algorithms and assesses their performance on a large and high-quality dataset, aiming to generate an efficient model and develop a PI-RADS 3 classifier. By combining the power of machine learning with the information provided by mpMRI, histopathological results as well as expert image interpretation, the investigators attempt to improve the diagnostic accuracy, which in the future my lead to more informed clinical decisions and reduce unnecessary biopsies.
Investigators
Dr. Panagiota Manava
Dr. med. Panagiota Manava, MD, senior physician
Paracelsus Medical University
Eligibility Criteria
Inclusion Criteria
- •Only patients with a clinical indication for mp prostate MRI will be included in this prospective study.
- •No allergies to GBCA
Exclusion Criteria
- •Contraindications for MRI
Outcomes
Primary Outcomes
Quantitative Signal - Intensity - Measurements with Region of Interest in specific in high b-value (800, 1500, 4000) axial MRI Images
Time Frame: through study completion, an average of 3 years
Regions of interest for signal intensity measurements will be drawn in various prostate lesions, the size of the region of interest will depend on the target structure. Signal intensity will be measured and normalized in mm2/s
Quantitative Signal - Intensity - Measurements with Region of Interest in specific in Apparent diffusion coefficient (ADC) axial MRI Images
Time Frame: through study completion, an average of 3 years
Regions of interest for signal intensity measurements will be drawn in various prostate lesions, the size of the region of interest will depend on the target structure. Signal intensity will be measured and normalized in mm2/s
Normalized Quantitative Signal - Intensity - Measurements with Region of Interest drawn in specific T2-weighted axial MRI Images
Time Frame: through study completion, an average of 3 years
Regions of interest for quantitative signal intensity measurements will be drawn in various prostate lesions, the size of the region of interest will depend on the target structure. Image analysis will be performed on a PACS workstation. Signal intensity will be measured and normalized, therefore no units needed.
Signal - Intensity - Measurements with Region of Interest in specific dynamic contrast enhanced (DCE) MRI Images
Time Frame: through study completion, an average of 3 years
Regions of interest for signal intensity measurements will be drawn in various prostate lesions, the size of the region of interest will depend on the target structure. Signal intensity will be measured and normalized. Image analysis will be performed on a PACS workstation. The original Time inteisity curves are transformed in relative enhancement curves. Thus, they are normalized with respect to first point in time and represent the percentage increase compared to the time before contrast arrival, no units needed.