The Power of Normative Data (faq)  

To fully unlock the power of fMRI, Notus Neuropsychological Imaging (NNI) spent over 5 years compiling a reference range for its protocols.

This normative database acts as a statistical baseline for individual patient data to be compared with.  A standard fMRI with no “normal” references would do little to benefit a patient or a physician treating a patient because it would not allow any meaningful interpretation.  For a simple example:  In a similar way to the Body Mass Index, Blood Panel, or a Toxicology Screen, individual data would be meaningless without a quantitative characterization of the healthy population.


NNI provides working solutions to the major challenges that have been identified in many recent efforts toward making fMRI clinically viable

At present, no standardized cognitive/neuropsychologic testing packages have been certified by any duly charged governing body in terms of appropriate stimuli or tasks. Testing guidelines in terms of choice of stimuli, tasks, and performance parameters should thus presently conform in general to standard guidelines accepted for neurobehavioral and/or neuropsychologic testing.
–Hart et al., 2007
The myriad ways in which fMRI studies can vary from center to center has complicated the standardization of methods across sites. Little standardization currently exists.
Despite the potential for developing representative norms for clinical application, no such datasets exist.
–Brown, 2007

Normative database allows for detection of previously undetectable conditions

NNI’s normative database also allows for further interpretation of certain activation patterns which can reveal subtle to severe pathology undetected in standard neuropsychological testing and standard structural imaging.  This is partly because fMRI allows physicians to measure and understand cognitive performance broken down into specific functional areas of the brain.

For just one example of how this is useful, in some instances of injury one area of the brain will over-activate to compensate for another deficient area of the brain (Compensatory Activation).  This may result in a false negative “passing” grade on a neuropsychological exam, despite the fact that severe impairment could still exist.

Other examples of the unique detection capabilities of NNI testing include Hyperactivation of certain areas (more activation than normal, without compensatory activity), and instances where abnormal brain tissue still activates normally.

For several other in-depth examples that demonstrate this, see:

NNI Case Studies – Real Life Applications of fMRI      READ MORE —>

Technical FAQs About the NNI Database

Platform Stability

Q. The original NNI norms were collected on a single 1.5T scanner using an EPI sequence with uniform psd parameters , receiver values, shimming procedures, and so forth. Cross-facility variation seems critical when using these norms. This seems especially important when using a 3T scanner versus a 1.5T scanner. What assurances do I have that these norms are valid for my scanner?

A. The NNI norms began with an initial set of control subjects (ages 20-40) tested using an identical scanning environment.  Since then, the NNI protocols have been used to collect data from additional control subjects using 4 additional different scanning environments. To date, then, the norms consist of data collected from 5 different scanners—these include the following scanner and sequence types:

1. GE 1.5T EPI interleaved

2. GE 1.5T EPI sequential

2. GE 1.5T Spiral

2. GE 3T EPI sequential

4. Siemens 3T EPI interleaved

Whenever the NNI protocols are used on a new scanner, we collect data from at least 2 control subjects (more if desired). These new data are then checked against the existing NNI normative database, with the expectation that no value should fall more than 1 SD from the mean for any region on any protocol.

Putting this into context, this means that so far, we have found less variability among new controls scanned on 3T magnets or Spiral sequences, than we found among the original set of controls from the same 1.5T scanner and sequence.

Although there are theoretical reasons to expect vastly different results with different magnet strengths, empirically, we have not found this to be the case. The most likely explanation for the uniformity we consistently find when using different scanners has to do with normalization procedures applied both during statistical analysis and peak extraction processing.

Finally, the interpretation procedure itself includes an additional measure against “scanner dependence.” Specifically, functional abnormality is virtually always manifest as inconsistency in z-score levels across brain regions within a patients, rather than uniform levels of high or low  z-scores within a subject, compared to the group norms.

Nonetheless, as NNI protocols are used in more facilities, we envision complete normative databases for 1.5T versus 3T scanners in the near future.


How normal are the NNI norms?


Q. The NNI analysis presents patient activation results for each brain region in terms of z-scores, which are only valid for normal distributions. How normally distributed are the NNI control data?

A. All of the NNI protocol norms (all brain regions for each protocol) have been evaluated for normality using a few different analyses. Probably the most relevant (and stringent) test is the Anderson-Darling sample-size-adjusted A2* test. The results of this test reveal that all brain regions included in the NNI norms surpass at least a 95% confidence level for normal distribution, and most regions exceed a 99% confidence level.

It is fortunate that statistical measures of brain activation tend to yield normally distributed data across control subjects as it makes patient assessment much more straightforward, than if, for example, highly skewed or multimodal distributions were typical across controls.