Clinical Implementation of an Artificial Intelligence Algorithm for Magnetic Resonance-Derived Measurement of Total...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Nephrology

Background:

• Autosomal dominant polycystic kidney disease (ADPKD) is a genetic disorder characterized by kidney cyst development.
• Accurate measurement of total kidney volume (TKV) is crucial for monitoring ADPKD progression and guiding treatment.
• Magnetic resonance (MR) imaging is a key modality for TKV assessment, but manual segmentation can be time-consuming.

Purpose of the Study:

• To evaluate the real-world clinical performance of an internally developed artificial intelligence (AI) algorithm for MR-derived TKV measurement in ADPKD patients.
• To assess the accuracy and reliability of the AI algorithm compared to manual segmentation and interobserver variability.
• To determine the impact of AI-driven TKV measurements on ADPKD disease classification.

Main Methods:

• Retrospective analysis of 170 MR imaging examinations from 161 adult ADPKD patients undergoing routine clinical care.
• Implementation and tracking of an AI algorithm for automated TKV calculation over one year.
• Comparison of AI-derived TKV with manually edited segmentations by radiologists and technologists, assessing agreement and volume differences.
• Analysis of ADPKD disease severity class assignment based on AI-derived versus manually edited TKV.

Main Results:

• Successful clinical implementation of the AI algorithm for TKV measurement.
• High agreement between AI-based and manually edited TKV segmentations, with a small mean volume difference of -3.3% in edited cases.
• AI-derived TKV showed non-inferior performance compared to interobserver variability and other established methods.
• High concordance in ADPKD disease class assignment between AI-based and manually edited TKV values.

Conclusions:

• The AI algorithm for MR-derived TKV in ADPKD demonstrated robust performance in a real-world clinical setting.
• Careful development, validation, and implementation in environments matching development conditions are key to preserving AI algorithm performance.
• The AI tool shows potential for efficient and accurate TKV assessment in clinical practice, supporting disease management in ADPKD.