Working Group on MRI - Summary of Findings
FINAL CONSENSUS – WORKING GROUP ON MRI - NIA ALZHEIMER'S INITIATIVE
Committee Members: Cecil Charles, Charles DeCarli, Nick Fox, Michael Grundman, Harold Hampel, Clifford Jack, Jorge Jovicich, Jeffrey Kaye, Leyla Toledo-Morrell, Henry Rusinek, Norbert Schuff, Art Toga, Michael Weiner
The following represents the consensus findings of the MRI working group assembled by NIA concerning a possible NIA Alzheimer's Imaging Initiative. Following the first teleconference, several members prepared language concerning key issues. Draft documents were circulated and discussed. Final documents were prepared after a second conference call. No effort has been made to harmonize the style and language of different documents.
I. MRI Acquisition Guidelines:
1. The long term goal of the RFA is to create a generally accessible data repository so that multi-site data can be combined for hypothesis testing to determine how best to track disease progression and to assess the value of diagnostic assessments. Details of the MR acquisition protocol are up to the individual applicants. The following general principles might be considered in designing the protocol:
Choice of parameters in the MR acquisition protocol should be optimized to reflect the methods of image analyses proposed.
2. A 3D volumetric acquisition will likely be the backbone of the protocol in terms of subsequent structural MRI analyses. The 3D T1-weighted sequence should provide similar contrast-to-noise properties across sites and across various MR vendors. The sequence should provide full head coverage with anti-aliasing features enabled with isotropic or nearly isotropic voxel size close to 1 mm cubed resolution.
3. The protocol should contain a T2-like sequence that is sensitive to brain pathology (which generally has elevated T2 relaxation times relative to normal brain tissue) – for example, dual fast spin echo or FLAIR.
4. While DICOM is a standard format for image transmission, alternate file formats are acceptable. Each MRI volume should be kept in a single file accompanied by a text header with well documented fields. Code for reading the header and the image should be provided to facilitate image processing by diverse organizations. Central storage facility is desirable.
5. Intensity calibration across subjects might be considered (discussed below).
6. Total magnet time to complete the protocol should accommodate the tolerance of this patient population
The applicants may propose additional MR-based modalities (MRS, DWI, MT, fMRI) provided that they demonstrate the feasibility of performing reliable multi-center image interpretation and serial exams.
Control subjects and MCI subjects should be studied for 3 years while AD subjects can be studied for 2 years.
II. Recommendations for Phantom and Scanner Calibration
Procedures to handle variations inherent to the use of different scanners at different site and to problems with scanner instability over the course of the study must be addressed. Image contrast and gradient spatial fidelity, including gradient strength and non-linearity, are the main features that should be calibrated across the participating using a set of calibration phantoms. In addition, image contrast and gradient strength should be evaluated periodically over the course of the study to minimize measurement errors. Specific procedures to assume geometric accuracy, field homogeneity, slice position accuracy, signal /noise etc should be considered.
Procedures to evaluate phantom studies and to make scanner adjustments should be fully described, as well as criteria for inclusion/exclusion of sites and the range of acceptable instrumental variations should be established.
After calibration and tuning of the scanners, it is recommended to consider a validation phase to determine whether MRI protocols at different sites result in comparable image quality and whether variability between sites is negligible compared subject variability.
If additional MR modalities, such as MRS, DWI, MT, or fMRI are proposed for this study, procedures for quality control of data acquisitions must be addressed separately for each modality.
III. Recommendations concerning field strength
While a large majority of MR sites utilize 1.5 Tesla systems, there is a growing number of academic as well as non-academic sites installing 3 Tesla machines that have 510-K approval. For a multicenter trial it is the recommendation that the primary field strength be widely available and that the technology is stable. This would favor the use of 1.5 Tesla systems, however, substudies comparing 3T and 1.5T would be useful for a number of reasons, especially in the development of image analysis algorithms. Additionally, this issue should be considered carefully with respect to the timing of study initiation as the stability of 3T systems if rapidly improving. A potential drawback will be that high recruiting sites (which may be non-academic) may not have access to 3T systems.
IV. Screening
The aim of this initiative is to develop a naturalistic study of a large group of individuals at high risk for Alzheimer's disease by virtue of having mild cognitive impairment and a smaller group of mildly affected Alzheimer's disease patients and controls. Current health care guidelines proposed by the American Academy of Neurology recommend the use of structural brain imaging for the diagnosis of Alzheimer's disease. The purpose of these guidelines are to exclude other diseases that may cause cognitive impairment mimicking Alzheimer's disease, such as brain tumors and stroke. It is common, therefore, for clinical research to derive criteria for the inclusion and exclusion of patients based on structural brain imaging. In this initiative, the applicant is strongly encouraged to consider the following:
Criteria for patient exclusion based on structural imaging should be evidence based and reliable, particularly with regard to the extent of cerebrovascular disease (e.g. current ‘standards of practice' exclude individuals with large strokes or more than one lacune, but do not evaluate the extent of white matter abnormalities).
Criteria for subject exclusion due to new imaging findings during follow-up should also be clearly stated. The applicant should strongly consider arguments for and against exclusion of patients relating to the observation of new, but clinically asymptomatic cerebrovascular events.
IV. Between Scan Intervals
The initiative aims to conduct a longitudinal naturalistic study to determine how best to track progression in MCI and early AD. The initiative also aims to provide information on the natural history of these conditions as well the diagnostic and prognostic value of the assessments performed. Finally the initiative aims to provide a data set of imaging and biological samples for use with alternative and future methods of analysis. Therefore
the acquisition and storage of data should be appropriate for this aim. The QA methods should be appropriate for the acquisitions and particular analysis planned but should be mindful of other potential uses of the data.
The applicant should detail how their proposal will provide information for future potential therapeutic trials on:
1) which neuroimaging and biological markers (solely or in combination) would be most effective as outcome measures for the patient group being studied; 2) the particular requirements (including QA), considerations (practical and ethical) and problems associated with acquiring these markers; 3) the optimum timing and frequency of the measurements to be performed for a given patient group and given study size and length. The information gained in this initiative should be sufficient to allow robust, evidence-based sample size calculations for a range of potential future study designs. For the purposes of providing applicable information to guide future study design the applicant should consider the effect of changes in available technology and analysis methods during and following the period of the study.
Timing and Frequency of MR Acquisition: The applicant should detail the timing and frequency of MR acquisitions. The applicant should be aware of the need to compare MRI outcomes with different markers (e.g. PET, biological) for this 2 to 3 year study of MCI and early AD. For a particular outcome measure (e.g. rates of global or regional cerebral atrophy) the proposal should address the value of:
1) multiple intervals between MRI in order to allow power calculations to be performed for a range of different study lengths (with a minimum of a scan at beginning and one at the end);
2) a combination of multiple scans in order to allow a reduction in variance or measurement error in rates of change.
The applicant should estimate the time course of patient drop-outs and design the scan frequency and intervals to take account of these factors. The applicant should consider how the schedule they propose might affect these drop-outs.
It is envisaged that the MRI component of the initiative will involve several scans at different points in the study. The MRI schedule proposed should allow for rates of progression to be calculated from two scans at a variety of different intervals between scans.
The applicant may wish to consider a schedule which includes a baseline scan and then further scans at 6 monthly intervals for the duration of the study (2 for AD and/or 3 years for MCI; controls could be studied annually for 3 years). The reasons for this or any other particular schedule should be clearly argued. If the applicant wishes to have a subset of the study with an alternative scanning schedule (e.g. additional scans) then this too should be justified.
Sample sizes for disease progression studies are determined by the variance in rates of progression as measured by the particular outcome measure. That variance is due to a combination of 1) between-individual variance in rates of atrophy progression (related to disease severity, genetic factors etc); 2) within-in dividual variance (e.g. physiological fluctuations, concomitant medication, change in rate of progression over time), and 3) measurement error/variability due to scan acquisition variability and image analysis error. The applicant should address how their proposal might address these different aspects that influence disease progression study designs.
V. Analytical Strategies
The goal of this initiative is identify biological markers that are relevant measures of Alzheimer's pathology. Important to this aim is the early detection and measurement of the progression of the disease. With the identification of specific biological markers, it is hoped that new medical treatment strategies may be proposed and tested in a cost and time-efficient manner. All such measures, however, must be as precise and as valid as possible with the added demand for widespread availability to academia and industry. These measures should also be associated with and possibly predict clinical symptoms of the disease.
The relatively distinct topographic and chronological progression of Alzheimer's disease makes MRI a suitable candidate as a biological marker for the disease. While multiple studies have examined the ability to detect the presence of Alzheimer's disease, its progression and measures that are correlated with the pathology of the disease, no consistent analytical strategies have emerged. Data relating MRI measures to other biological markers are even more limited. Therefore, for this initiative, the applicant is strongly encouraged to consider the following:
Identify a priori specific brain regions to be considered for measurement. Choices should be based on evidence from pathological studies and MRI feasibility. Applicants should carefully distinguish between measures that are specific to Alzheimer's disease and those that reflect a general state of neurodegeneration. For example, while global measures of brain atrophy are ideally suited to measure progression of Alzheimer's disease, this should not be considered in lieu of more specific regional measures. A measure of intracranial volume that will reflect individual differences in head size is also strongly encouraged.
All measurements should be tested for reliability at development and on an ongoing basis. The applicant should strongly consider methods for which measurement error is generally less than 1% of the mean volume of the object measured.Applicants are encouaraged to describe methods and provide data concerning their precision and bias. Specific discussion of skill sets required for the analysts and the extent of necessary training to acquire these skills also should be included.
The applicant should strongly consider the general availability and utility of any proposed measurement technique. Since the goal of this study is to have biological markers that can be easily applied to clinical studies of Alzheimer's disease, the applicant should pursue methods that can be widely distributed to academia and industry or be prepared to freely distribute any method developed. It is also important that the applicant consider a variety of methods that will show convergent validity. Involvement of multiple sites using multiple techniques would enhance the likelihood of success in this regard.
The applicant should strongly consider methods that are robust with regard to differences in imaging quality due to patient movement or variable MRI hardware or software. Rejection of more than a small percentage of images due to technical limitations of the method would be deemed unacceptable. In this regard, the applicant should also develop methods for quality control of image acquisition. These controls, however, should not exclude large numbers of sites or individuals imaged. While general suggestions regarding imaging sequences have already been given, the applicant should be careful not to rely on special sequences for their analytical methods that might limit general utility of the method or pose an undo burden to the subjects.
Analytical methods should be able to generate measures of individual size as well as measures of change in size of all regions studied. This will be important for correlated studies aimed at validating other biological markers that may be more specific to the disease process (e.g. CSF tau levels or temporal/parietal FDG metabolism).
V. Recommendations for Measures
Consideration for choices of structural MRI measures should be guided by sensitivity to detect brain atrophy and by specificity to measure changes related to AD. Since the primary purpose of the initiative is to provide the most powerful measures of disease progression, the applicants are encouraged to consider a variety of different approaches to assess sizes of brain structures and change in sizes.
Applicants should consider global measures of atrophy, which are suited to assess progression of general brain changes with high sensitivity. In addition, applicants should consider measurements of a-priori identified regions of interest that are thought to reflect changes specifically related to AD. Furthermore, as an alternative to a-priori region of interests, applicants are encouraged to consider unbiased approaches to measure regional atrophy, for example using statistical brain mapping procedures.
As cerebral vascular disease might be a confound for brain atrophy, it is strongly recommended to consider evaluation (qualitative and quantitative) of white matter lesions. It is also strongly encouraged obtaining a measure that reflects individual differences in head size, as individual size of brain structures, in addition to changes in size over time tha might be important for staging progression.
Applicants may propose additional MR-based measures (MRS, DWI, MT, fMRI) provided that they justify that the additional measures will improve power to measure disease progression.
Page last updated Feb 16, 2008
