Improving workflow in prostate MRI: AI-based decision-making on biparametric or multiparametric MRI

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Dynamic contrast-enhanced (DCE) sequences are crucial in prostate MRI but increase scan time and cost.
• Determining the necessity of DCE requires expert interpretation, which can be time-consuming.
• Optimizing MRI protocols can improve efficiency and patient throughput.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI) algorithm for selecting dynamic contrast-enhanced (DCE) sequences in prostate MRI.
• To assess the algorithm's performance in identifying patients who would benefit from DCE.
• To compare the AI algorithm's accuracy against human readers and a radiology technician.

Main Methods:

• A convolutional neural network (CNN) was trained on 300 prostate MRI examinations.
• Expert reader consensus served as the reference standard for DCE necessity.
• The CNN was validated on separate cohorts (100 and 31 examinations) from the same and different vendors, respectively.

Main Results:

• The CNN achieved a sensitivity of 94.4% and specificity of 68.8% (AUC: 0.88) for DCE necessity.
• It correctly assigned 78% of patients to biparametric or multiparametric protocols, with only 2% requiring re-examination.
• The AI demonstrated improved sensitivity (30.5% increase) compared to a radiology technician.

Conclusions:

• The developed AI algorithm accurately identifies the need for DCE sequences in prostate MRI.
• Integrating this AI can streamline MRI protocols, potentially making on-table monitoring obsolete.
• This approach optimizes contrast-enhanced MRI use, improving efficiency and resource allocation.