Reconstructing Reality with Deep Learning – From 2D to 3D in Musculoskeletal MRI-Reconstructing Reality with Deep Learning

Video Transcription

Video Summary

The webinar featured Dr. Judith Hermann discussing how deep learning is transforming musculoskeletal MRI reconstruction from 2D to 3D imaging, especially in wrist MRI. She explained the longstanding MRI trade-off between speed and image quality, and showed how AI can reduce scan times while improving image clarity and diagnostic confidence.<br /><br />Examples from knee, shoulder, spine, and wrist imaging demonstrated that deep learning reconstructions can significantly reduce noise and acquisition time without losing important pathology. In wrist MRI, 3D isotropic sequences enhanced visualization of small structures like the TFCC and ligaments, while allowing multiplanar reconstruction and faster protocols.<br /><br />Dr. Hermann also highlighted the broader benefits of AI, including improved workflow, increased scanner availability, and potential energy savings. She cited data showing major reductions in scan time, energy use, CO2 emissions, and operating costs with AI-assisted protocols. However, she also emphasized the need to consider the energy and environmental costs of training and running AI systems.<br /><br />Her overall message: AI is not just hype, but a practical tool that can improve radiology if thoughtfully integrated into clinical practice.

Keywords

deep learning

musculoskeletal MRI

3D imaging

wrist MRI

AI reconstruction

scan time reduction

image quality