Imaging Standardization Proposal — April 2018

Prepared by the ADC Imaging Standardization Committee: Tammie Benzinger, Charles DeCarli, Brad Dickerson, Cliff Jack, Bill Jagust (Chair), Sterling Johnson, Bud Kukull, Beth Mormino, Gil Rabinovici, Emily Rogalski, Andrew Saykin (NIA: Nina Silverberg, Cerise Elliott, John Hsiao).

Goal

To facilitate cross-ADC scientific collaboration that uses imaging methods by developing protocols to standardize image acquisition, quality control, curation, and analysis.

Infrastructure

The overall goal would be facilitated by a steering committee that has responsibility for:

  • Defining the appropriate protocols for acquiring images (PET and MR) and review/modify as needed
  • Assisting sites with technical aspects of image acquisition, data upload and data use ("help desk"). This may include both expert queries and a knowledge based list/serve
  • Monitoring data flow, curation and analysis

These activities would be coordinated/assisted by NACC.

Sites would be incentivized to acquire standardized images through funding. This could be as a lump sum budget and/or reimbursement on a per-scan basis.

Image acquisition

Images would be acquired using ADNI standards when possible, which includes MRI sequences and most PET tracers.

We propose 2 standard protocols (short and long) for MRI, both using sequences prescribed in ADNI3:

Short:

  • 3DT1
  • 3D FLAIR
  • 3D T2 (to be confirmed)
  • Gradient Echo

Long (in addition to Short):

  • DTI
  • Resting state fMRI
  • Arterial spin-labeled

PET data acquisition would also follow ADNI standards, or other widely accepted standards in the imaging community as follows:

  • [11C]PIB: 40-60 or 50-70 min post-injection frames (10-15 mCi injection)
  • [18F]Florbetapir: 50-70 or 50-60 min post-injection frames (10 mCi injection)
  • [18F]Florbetaben: 90-110 min post injection (8 mCi injection)
  • [18F]Flortaucipir: 80-100 min post injection (10 mCi injection)
  • [18F]MK6240 70-90 min post injection (10 mCi injection) — to be verified

Other tracers pending include PI2620, GTP1, Flutemetamol

Data upload, annotation, quality control, curation

Procedures would need to be developed for performing all of the functions listed in this section. Some of these procedures such as basic data uploads should be relatively straightforward and are already being accomplished by NACC. Others, such as image de-identification, annotation and quality control, are quite complex and require high-level software capabilities specific to medical images, MR physics knowledge, interactions with multiple scanner vendors, and detailed knowledge of MR protocols across platforms.

For data uploads and annotation

  • Develop methods for assuring protocol compliance by using software to check for acquisition parameters and/or developing ADC-entered data fields to define scan parameters. This is a complex undertaking that requires high-level MR physics knowledge and information/practices of multiple scanner vendors.
  • Assure complete image de-identification while not cleaning DICOM tags needed for image processing and analysis
  • Integrate images with other NACC data linking by subject ID
  • Create a system where scan dates can easily be linked to dates of other relevant clinical, neuropsych, etc. data
  • Track and label all images as they pass or fail quality checks
  • Produce regular reports and feedback to sites for monitoring of uploads and stored data

For data quality control

  • For both PET and MRI: Inspect all images manually for artifacts (motion), overall image quality, correct transmission of the image, full coverage of the head.
  • For PET data, assess if count rate and statistics are adequate, perform averaging of frames to provide a single image format. Pre-processing of images to result in standard spatial orientation, voxel size, count rates, and resolution.
  • Assure that image acquisition complied with protocol using both image headers and affiliated uploaded meta-data. For MR this will require detailed knowledge of DICOM, an MR physics knowledge base that covers all 3 vendors and working relationships with/access to MR scientists from GE, Siemens, and Philips.
  • Contact sites for re-scans when necessary
  • Maintain a log of all scanners, track software and hardware upgrades

For data curation

  • Provide web-based search functions permitting users to define key image features to query and assemble databases
  • Permit download of original or processed scans to user laboratories/analysis laboratories.
  • Receive and manage uploads of analyzed data from data analysis laboratories in tabular form. Link analysis results to scans, permitting queries and downloads
  • Maintain a database indicating which scans have been analyzed with which methods
  • Provide a website including a knowledge base and expert query function

Data analysis

The function of data analysis would be to provide basic data results for each data type. Standard results would be uploaded to NACC for curation and a specific set of final values would be returned to the ADC that collected the scan.

For MRI

  • Regional brain volumes measured with standard software (e.g., FreeSurfer)
  • Regional volumes of white matter hyperintensities
  • Number and location of infarctions
  • Number and locations of microbleeds
  • Regional values for standard diffusion metrics, e.g., fractional anisotropy, mean diffusivity, etc.
  • Regional values for relative cerebral perfusion
  • To be determined summary statistics for task-free fMRI — e.g., network-based connectivity metrics

For PET

  • Total and regional amyloid or tau tracer uptake normalized to a standard reference region
  • Summary value for amyloid or tau
  • Dichotomous rating for positive/negative for amyloid or tau