Weeks to Minutes: Improve Analysis Cycle Time in Drug Development With MONAI Label
, Director, Head of Bioimaging Analytics, GSK
We present a deep learning framework for automated image segmentation in drug development, where raw outputs of multiple bioimaging technologies are processed by MONAI Label in minutes, rather than the weeks needed for manual human annotation. We exemplified three business applications: we assessed mouse kidney total volumes and disease states through MRI, estimated tumor growth over time, and extracted whole-body radiomics for nonhuman primates. Utilizing MONAI Label toolbox integrated with 3D Slicer frontend and in-house high-performance computing, we demonstrate the improved efficiency of bioimaging analytics in drug development. Results indicated a strong correlation between expert-level manual analysis and automated model inference while reducing cycle time by several hundredfold. Future work involves addressing scalability and generalizability, with the inclusion of multi-modal segment-anything model applications in animal studies.
Events & Trainings:
Date:
Level:
Industry:
Topic:
NVIDIA technology: Cloud / Data Center GPU,CUDA,cuDNN,MONAI,RTX GPU