Is it really so that fMRI can enhance lie detection? This entirely depends upon your method of analysis, the experimental setup, and especially controlling for factors that influence the scanning results. Let’s take them one at a time
First, notice that during the brief explanation of the neuroimaging study, we can hear that they compare 5 subjects who lie to another 5 subjects that do not lie. In general, an analysis of only 10 subjects is barely enough to call a study. This alone is reason for suspicion.
Second, how you analyze your study influences what you find. For example, if you are doing a so-called univariate analysis (see this article, PDF), i.e. analyzing one brain voxel at a time. Here, statistical power is needed, and what you’d like to avoid is having too few subjects or too few repetitions of the same task. A promising method is multivariate analysis methods (see this article, PDF), which can be simplified as “pattern recognition”, i.e. the estimation and recognition of brain activation patterns. This method holds the promise to analyze single runs and identify crucial aspects of single runs, e.g. if a person is lying on one particular question.
Third, even if multivariate analysis works as it should, there are still several obstacles to be dealt with. The main problem here is the brain recording itself. The relationship between neural activity and the BOLD fMRI signal is not fully understood, and the interpretation of results with this method should still be cautious to overstate the findings. Another pertinent question is how the BOLD signal is influenced by different factors such as central stimulants, age, sleep, gender, and even the time of day. As for pathological factors, it is known that vascular changes occur in brain injury or neurodegenerative diseases such as dementia. Vascular factors are thought to play a part in the BOLD signal.
Finally, what do we really know about the “normal brain”? As of yet, we know little, if anything about the normal distribution of brain activation. It is even more complicating to see that the BOLD signal itself shows a lot of individual as well as intra-individual differences. So there is no standard score for BOLD fMRI that we can use. It is possible to make use of the percentage of signal change between e.g. two cognitive states. But without knowledge about what the normal range and variations are, let alone the relationship to the factors mentioned above, single-subject fMRI is hazardous.
As long as those factors are not settled, let’s keep fMRI out of the courtrooms. Pioneering work is necessary in order to drive the field forward. Applying it to life and death situations, or the imprisonment of a person is indeed a different story.