Automated Coregistered Segmentation for Volumetric Analysis of Multiparametric Renal MRI

Area of Science:

• Medical imaging analysis
• Artificial intelligence in radiology
• Renal function assessment

Background:

• Multiparametric renal MRI (mpMRI) is crucial for kidney assessment.
• Current postprocessing is often manual, time-consuming, and prone to variability.
• Automated solutions are needed to improve efficiency and accuracy in renal MRI analysis.

Purpose of the Study:

• To develop and evaluate an automated deep learning pipeline for multiparametric renal MRI.
• To enable accurate kidney alignment, segmentation, and quantitative feature extraction.
• To create an efficient, single-workflow solution for renal MRI postprocessing.

Main Methods:

• A deep learning segmentation network delineates renal structures.
• A deep learning registration algorithm aligns multiparametric MRI series.
• Quantitative parameters are extracted from aligned and segmented data.
• Validation performed using five-fold cross-validation against manual analysis in 34 subjects.

Main Results:

• The automated pipeline demonstrated high agreement with expert kidney segmentation.
• Significant alignment improvements were achieved through registration.
• Volumetric analysis showed strong correlation (r > 0.9) with manual results.
• Feature extraction yielded high intraclass correlation coefficients with minimal bias.
• The complete pipeline processed scans in approximately 15 seconds.

Conclusions:

• A reliable automated pipeline for renal multiparametric MRI postprocessing has been established.
• The pipeline offers high accuracy and efficiency for clinical use.
• This technology can enhance diagnosis and treatment planning for kidney diseases.