ADNI was launched in 2003 by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, the Food and Drug Administration, and private pharmaceutical companies and non-profit organizations. ADNI is the result of efforts of many co-investigators from a broad range of academic institutions and private corporations and subjects have been recruited from over 50 sites across the US and Canada. For up to date information, see www.adni-info.org
The ADNI PET core has pioneered the development of standards for quality control and acquisition of FDG-PET and PIB-PET data that are now widely utilized in both academic and pharmaceutical studies worldwide. The key design features of this protocol include: (1) simplicity, permitting multisite application; (2) compatibility with current commercially marketed PET cameras; (3) flexible data acquisition for post-processing to standardize formats, intensity and resolution; (4) low radiation dose; (5) high patient acceptance; (6) real-time quality assurance and quality control; and (7) centralized storage and standardization of images acquired on different scanners. The protocol has been successfully implemented at all ADNI PET sites in North America and has been adopted by a number of pharmaceutical clinical trials and the Japanese ADNI. The protocol was developed collaboratively amongst ADNI investigators with input from the sites and an outside panel of experts. In addition, in the PIB “add-on” we developed a standardized protocol for the acquisition of PIB-PET data that has also been widely adopted.
During the first two years of the project, sites were approved for PET scanning by performing a pair of phantom scans on the 3-D Hoffman brain phantom following a protocol that matched the acquisition and reconstruction parameters to be used for the human phase of the ADNI project and electronically transmitting these images to the Laboratory of Neuroimaging at UCLA. Scans were passed through a quality control process that checked for statistical noise, agreement with a digital version of that phantom (the gold standard), and assessed for image resolution and image uniformity. This process also helped develop a method for assessing and correcting for differences in PET images across sites1
. There are a total of 17 different scanner models from three vendors that have been used by participating ADNI sites. Intrinsic resolution differs by more than a factor of two across scanners, from as high as 3-4mm to as low as 8mm effective resolution, thus requiring assessment of resolution and partial volume effects
Human PET studies were first acquired at the start of the second year of ADNI, September 2005. All human PET scans are run through a stringent quality control procedure to assess image quality. QC checks include the number of detected coincidence events (for statistical quality), motion assessment across temporal frames, checking for full coverage of the brain, visual checks of images to look for the most common PET artifacts (such as normalization problems or motion between attenuation and emission scans), as well as image header checks to make sure the exact ADNI protocol has been followed. Half the enrolled ADNI subjects underwent FDG-PET, and we now have 404 baseline scans, 368 6 month scans, 336 12 month scans, 154 18 month scans (MCI subjects only), 283 24 month scans, and to-date 106 36 month scans (MCI and control subjects only). Besides FDG, approximately one quarter of the PET subjects (n=103) also received PIB scans to image amyloid deposition. Repeat scans at a one-year interval have been performed on 80 of the 103 subjects, while 34 subjects have had a 3rd annual PIB scan.
The original “raw” PET image sets are uploaded to the LONI image repository. Routines read and convert PET images sets to a standard DICOM file format. The different temporal frames are co-registered, and both a dynamic image set and a single-frame averaged image set are produced in the original patient orientation and in the original intrinsic in-plane pixel size and plane spacing for that scanner. In addition, all images sets are re-oriented to a common spatial orientation and interpolated onto a uniform image grid (1.5mm3
voxels). In the final step, images are smoothed with a scanner-specific filter derived from each site's Hoffman phantom1
to provide a common isotropic resolution of 8 mm FWHM resolution. These processed PET image sets are then uploaded back to LONI in a consistent DICOM format (see for examples). To date, the QC failure rate is 11%, which has declined from the initial year of the project when it was above 15%, to less than 7% for the 18 month and 24 month scans. The majority of scan failures was due to problems with image reconstruction and could be corrected with reprocessing; less than 2% of scans require rescanning.
Figure 1 Examples of the ADNI processing stream. Top row: Baseline FDG scan, MCI subject. Bottom row: 36m FDG scan, same subject. Left: FDG-PET image in native format and orientation, averaged 6 × 5 min frames. Middle: Same as left but with standardized (more ...)