To systematically study the accuracy of lower limb segmentation and quantification as a function of artifact severity and specific data augmentation strategies.
Key Findings:
Motion artifacts affect up to a third of clinical MRI sequences, necessitating repeat scans in approximately 20% of cases, highlighting the need for effective mitigation strategies.
Data augmentation strategies can enhance the robustness of diagnostic AI models against motion artifacts, potentially reducing the need for repeat scans.
The performance of the AI model decreases with increasing artifact severity but can be stabilized through domain-specific data augmentation, suggesting a practical approach for clinical implementation.
Interpretation:
The study suggests that implementing data augmentation techniques can significantly mitigate the negative impact of motion artifacts on AI model performance in MRI imaging, thereby improving diagnostic accuracy.
Limitations:
The study was limited to a specific patient population and imaging protocol, which may not represent broader clinical scenarios.
Generalizability of findings may be restricted due to the controlled study conditions, necessitating further research in diverse settings.
Conclusion:
Data augmentation strategies are essential for improving the robustness of AI models in the presence of motion artifacts, potentially enhancing diagnostic accuracy in clinical settings.
by Simon D. Westfechtel, Kristoffer Kußmann, Cederic Aßmann, Marc S. Huppertz, Robert M. Siepmann, Teresa Lemainque, Vera R. Winter, Alexandra Barabasch, Christiane K. Kuhl, Daniel Truhn, Sven Nebelung
