Services & Fees
Computational Resources
In collaboration with the institutional Oncology Care & Research Information Systems (OCRIS), a department with the MD Anderson Information System (IS) Division and vendors, QIAC offers a robust computational and communication resource for imaging research.
OCRIS along with OneIS support teams have a dedicated team of software developers and system administrators to implement the QIAC application and manage high-performance LINUX servers, Oracle database servers and mass data storage, backup and archive systems to serve Diagnostic Imaging and other research entities. QIAC is allocated with 70 TB of IBM GPFS disk storage, with capacity planning review conducted regularly to determine whether to increase the storage allocation. Image analysis software (e.g., Matlab, 3D Slicer, R) is available on a high-performance computing cluster as multi-user site licenses or open source licenses. QIAC has integrated with Epic to provide its imaging analysis reports to clinical trials teams. QIAC has established a pathway to upload image analysis results to MD Anderson’s Big Data platform where cross-disciplinary research on imaging, molecular, clinical, demographic data, etc., can be performed.
Office Space
Adequate office space is available for faculty, staff and trainees. The QIAC headquarters is located in 3SCR with over 600 sq. ft. of dedicated space to accommodate up to 18 imaging specialists and other technical and management staff.
Fee Structure
DICOM Images
DICOM Image De-identification
Current workflows that send DICOM images to PACS for clinical use will remain intact. Images acquired from scanners are archived in the routine clinical PACS. QIAC will implement an alternate flow that sends DICOM images through a de-identification process.
DICOM images flowing through QIAC will be intercepted. The intercepted DICOMs will have all PHI either removed or anonymized. The original identifiers that associate DICOM images to actual patients will be preserved, encrypted, and stored in an Oracle database.
The de-identified DICOMs will then be stored in a robust, high capacity, and high performance storage system running IBM’s General Parallel File System (GPFS).
DICOM Image Quality Assurance
Once de-identification and metadata extraction are completed, QIAC Imaging Analysts will be notified (by email) and directed to a web interface (QIAC Image QA portal) where they can perform quality assurance (QA) on the de-identified DICOM images. The QA will validate image quality as well as the accuracy and completeness of the deidentification.
DICOM images that pass QA will be made available for search and retrieval via another web interface (QIAC Data Explorer). DICOM images that fail QC will trigger another process that will cause the images to be regenerated from scanner and resubmitted through the de-identification workflow.
DICOM Metadata Extraction
Once de-identified DICOMs are persisted in the GPFS file system, the QIAC system then extracts metadata (DICOM tags and values) and stores them into a robust, highly scalable and multi-structured database system (HBase) which will become the main data source for the DICOM image data query portal (QIAC Data Explorer). QIAC Access and Management Portal is the user and admin interface, can be integrated with external infrastructure such as MDA big data warehouse, MDA OEA, EPIC interface, etc.
DICOM Image Analysis
QIAC researchers will be able to access the QIAC Data Explorer to search (based on any or all available attributes) and identify images of interest. They will then be able to select a set of images and submit them for image analysis. This will send the images through a semi-automated image analysis pipeline that will be running on a high performance, high memory compute cluster.
The image analysis results will also be stored on the GPFS file system. In addition to storing the analysis results file on the file system, all or some subset of the imaging features (as specified by the QIAC science team) will be extracted from the results files and stored into HBase for downstream data exploration and integration as needed.