Case Study: GE Healthcare achieves faster, more consistent MRI slice placement with TensorFlow

Intelligent Scanning Using Deep Learning for MRI

GE Healthcare, a leading provider of medical imaging equipment, faced a challenge with the manual process of planning brain MRI scans. This procedure, which relied on the skill of individual scan operators to identify anatomical landmarks from low-resolution images, was time-consuming, prone to error, and led to inconsistencies that could complicate diagnosis and even require patients to return for additional scans.

Using TensorFlow with the Keras interface, GE Healthcare developed a deep-learning framework for intelligent slice placement. This TensorFlow-based solution automatically identifies key anatomical structures to determine the optimal orientation and location for high-resolution diagnostic scans. The results included a 40% to 60% reduction in the time required to plan scans and improved accuracy, leading to shorter exams, a reduced chance of patient callbacks, and more consistent diagnostic quality.

GE Healthcare

Jason A. Polzin

GM Applications and Workflow