it’s about time i started to put some excerpts from the neuroethics book edited by judy illes. here’s another wonderful ditty from chapter 11: a picture is worth 1000 words, but which 1000?
italics are mine.
…from ch. 11 in Neuroethics…
What constitutes a ‘significantly greater’ activation than another, is in a way, in the eye of the beholder… lowering the threshold will create more regions that are statistically significant, whereas raising the threshold will reduce the number of significant regions. The choice of the threshold is largely determined by convention among researchers, rather than absolute standards. Reporting brain activation patterns is therefore primarily a statistical interpretation of a very complex dataset, and may be interpreted differently by different researchers. (Canli and Amin 2002)
YES!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
While group averages are vital for achieving acceptable signal-to-noise ratios, individual differences, from both anatomical and functional variability may become diluted and overshadowed (Beaulieu 2001; President’s Council on Bioethics 2004). When dealing with single-subject data, as is the case for presurgical planning, it is often desirable to minimize false-negative voxels in order to avoid erroneously excising potentially healthy tissue (M. P. Kirschen et al., under review). Outside the clinical setting, we can easily extend these considerations to any analytic objective set to pinpoint activation areas for function in individuals:
…the image of an activation pattern from a poorly designed study is visually indistinguishable from one based on an exemplary study. It takes a skilled practitioner to appreciate the difference. Therefore, one great danger lies in the abuse of neuroimaging data for presentations to untrained audiences such as courtroom juries. What can be easily forgotten when looking at these images is that they represent statistical inferences, rather than absolute truths. (Canli and Amin 2002)
YES!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Lastly, the interpretability of fMRI activation maps is dependent on how the data are displayed. The colour-coded statistical maps are usually overlaid on high-resolution anatomical MR images to highlight the brain anatomy. There are several media for displaying these composite images. The most rigorous is to overlay the functional data onto single anatomical slices in any imaging plane. While this is the most comprehensive means of examining the data, it is often difficult to localize the activations to a particular region, given a particular scan plane, and researchers are limited in the number of slices they can include in a publication or lecture. Alternatively, the activation maps can be presented on a three-dimensional rendered brain. While this technique gives good visualization of prominent external brain structures, internal regions like the hippocampus or basal ganglia are not well characterized on these models. Researchers often use both of these techniques to examine data, but ultimately choose the one that best highlights the main results of the study for presentation.
Since basic research is usually done to infer characteristics bearing on populations, the extension to individual applications is challenged by a scarcity of normative data that can support, for example, conclusions of abnormal activation (Rosen and Gur 2002). There are risks that measures will vary between individuals or that the meaning of data compared with normal individuals will be difficult to establish. Abnormality and predictive validity could even be more problematic in the context of real-world behaviours, especially those that are potentially value laden or culturally determined (Illes et al., 2003).
