Download presentation
Presentation is loading. Please wait.
Published byAnis Owen Modified over 8 years ago
1
1 The Data Explosion How can we achieve interoperability? James Williams Siemens Corporate Research Princeton, NJ
2
2 James Williams Siemens Corporate Research What are the right questions? Size : data is growing, how do we cope? Repeatability : how do we standardize and normalize image generation? Availability : how do we get what we want, where we want, in time to make a difference? Processing : how do we make algorithms comparable and portable? Validation : what are the right mechanisms for validation? Context : how can we convert data into knowledge?
3
3 James Williams Siemens Corporate Research The challenge of size Data sizes grow relentlessly Spatial resolution: CT Temporal Resolution: US Multiple modalities, PET/CT… Follow-up studies Network bandwidth as the bottleneck: Faster networks or… Move processing near the data! When we can, compress: What is diagnostic quality? Where do we need it & when?
4
4 James Williams Siemens Corporate Research The challenge of repeatability Sources of variation in acquisition Hardware Calibration Protocol Reconstruction Automation of MR scan-rescan protocol: AutoAlign & Phoenix
5
5 James Williams Siemens Corporate Research Goals of availability The imaging workspace available anywhere reading room office home All data relevant to the patient in one click All data relevant to the protocol available and searchable (teaching cases etc..) Some steps in the right direction: web clients, hanging protocols
6
6 James Williams Siemens Corporate Research Towards the compatibility and comparability of processing Visualization, detection, segmentation, registration measured with respect to reference standards (VRD) Cross platform executables Plug-in processing, DICOM WG23 ITK, a reference implementation? Open the platforms, but we must assure: Safety Privacy Stability
7
7 James Williams Siemens Corporate Research The challenge of algorithm validation Create reference databases as open challenges Should part of the database be hidden to avoid over-training? Database must grow and change as acquisition evolves Fast-tracking through the FDA Validation & liability as a 3 rd party business?
8
8 James Williams Siemens Corporate Research Data in context Beyond images: patient record, lab data, genomic, proteomic data The universal medical record: technology is not the problem. Individual privacy, ethics and economics are the key drivers. The anonymous database of everything. The benefit is clear, how do we cover the cost? Making knowledge from data. The structure has to support exploration for anecdote, and testing of hypothesis.
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.