Use Case:
QIPCM as a sandbox for AI
Altis Labs is an early stage company that develops imaging biomarkers using deep learning to accelerate clinical development of promising cancer therapies.
In order to build, test and validate their deep learning algorithms on datasets from public sources (such as TCIA) as well as from their customers, Altis Labs is using the QIPCM platform to ingest imaging datasets, collect labels, train algorithms on GPU enabled VMs and run testing and validation on the HPC4Health compute cluster.
Only anonymized images are sent to QIPCM, after which they are archived in our safe, secure, and backed-up PACS. Access to these images are logged and made available only to permitted personnel. Permitted personnel can thus log-in to the QIPCM platform to review trial images anytime, from anywhere with internet access.
| Requirement | QIPCM Solution |
|---|---|
| Image Anonymization and Transfer from Altis Labs customers | Custom pipeline for de-identification and secure transfer from Altis Labs customers to the QIPCM repository |
| Image transfer from Public data sets | Download and curation into the QIPCM repository |
| Image QA | Image QA by QIPCM including QA of the meta data to provide summary of adherence to acquisition protocols and anomalous cases. |
| Centralized Storage with Image Access Logs | Images that pass QA are forwarded to centralized PACS |
| Remote Connection | QIPCM Virtual Desktop Infrastructure (VMs) |
| GPU for deep learning | QIPCM access to HPC4health compute cluster |
| The software of Choice for labelling and feature extraction by Multiple Users | Installation of desired software onto user VMs |
