Читать книгу Philosophical Foundations of Neuroscience - P. M. S. Hacker - Страница 27

The genius that was manifest in all of Sherrington’ s experiments was his capacity to elicit profound observations from relatively simple techniques, including that for establishing correlations between cortical blood flow in an area of the brain and the activity of the cortex in that area. When, a century later, Seiji Ogawa led the way in the use of a much more elaborate technique, namely fMRI, he too, like Sherrington, was careful not to ask that the technique deliver more on its own than it was capable. Unfortunately, this has not been the case with many of those following in his path in using fMRI. Here a number of potentially serious technical problems need to be mentioned that should put on guard readers of papers using fMRI that make various claims concerning cortical localization of functions supporting psychological powers. These caveats may be responsible for a significant lack of reproducibility in fMRI measurements.¹⁰¹

The shapes of BOLD signals over the cortex of single subjects during a neuropsychological test

It has been traditional when analysing the fMRI signals to restrict such analyses to ‘box-car’-shaped signals, that is those which have a rectangular-like waveform that may be considered, for example, as a convolution of a gamma-variate function with a box-car function. The assumption that such signals are the only relevant ones has not been sustained by any analytical argument. It seems to have been based on spurious reasoning that those easily computed signal amplitudes seem to make sense. However, as has been emphasized by a number of recent studies, there is a wide variety of different shaped fMRI signals during simple neuropsychological tests, that is with different temporal profiles to that of the box-car like signals which occur over wide swathes of cortex (see Figure 1.1 and colour version, Plate 1, between pp. 232 and 233).¹⁰² It is clear that just measuring the ‘amplitude’ of such different temporal profiles is an inadequate way of characterizing and comparing them.

Figure 1.1 Significant activations over the whole brain revealed following averaging BOLD fMRI signals over 100 neuropsychological tests on a subject. (A) axial view of a map of the significant signal changes through the entire cortex to a simple motor response to a visual decision neuropsychological test, described in the text. (B) corresponding sagittal view of the map. (C) time course of the BOLD fMRI signals corresponding to coloured regions in the brain.

Source: Gonzalez-Castillo et al., 2012

Furthermore, there is no objective basis for accepting some shapes and rejecting others. Many of the fMRI determinations of the spatial distribution of activations of the cortex during neuropsychological tests are based on box-car temporal profiles and so assume that only these are relevant when defining the term ‘activation’. This assumption is unlikely to be correct and neglects important information regarding cortical activity accompanying a neuropsychological test. A plausible explanation for the large deviations of the fMRI signals from a box-car shape has recently been offered in terms of their originating from different extents of impulse activity in the region of interest conditional on temporal changes in the amount of inhibitory and excitatory activity there.¹⁰³ This being so, all shapes of the fMRI signal provide important information concerning the integrative activity in a given region.

The amplitude of BOLD signals over the cortex of single subjects during a neuropsychological test and on-going cortical activity

The threshold at which a particular fMRI signal is to be accepted for determining the sites over the cortex involved in a neuropsychological test depends on both the signal-to-noise ratio, as well as the amplitude of the signal taken to be of the box-car variety. However, it has been known for some time that the amplitude of the signal, taken as a measure of the transient activity during a neuropsychological test, is conditional on the activity of the part of the cortex of interest at the time of the test: the larger the background signal the smaller the transient signal and vice versa, independent of the neuropsychological test being used.¹⁰⁴ Using quantitative relations integrating BOLD responses, energetics and impulse firing, it has been shown how baseline activity can determine the amplitude of both positive and negative sustained BOLD signals.¹⁰⁵ It may therefore be necessary to measure the ongoing activity in the area of interest in order to correct for the fMRI amplitude during the neuropsychological test. In animal studies this can be done by using PET to determine the ongoing glucose metabolism, a direct measure of activity, in conjunction with the fMRI measurements.¹⁰⁶ This can be carried out using combined MRI/PET, but involves the injection of radioactive fluorodeoxyglucose as a ligand, a procedure that is invasive for human subjects. The extent that the baseline varies between different cortical areas of interest in the cortex is not clear at this time, so there are reservations concerning the interpretation of the amplitude of the BOLD signal recorded in different areas during a psychological test.¹⁰⁷

The existence of a ‘global BOLD signal’ across the cortex, and the question of the regression of the signal

In fMRI the ‘global signal’ refers to the average signal across all brain voxels in the time series of signal intensity. Given that most fMRI studies aim to specify changes in activity that are specific to particular brain areas, and that this global signal is at least in part generated by sources other than those due to such activity, such as the cardiac and respiratory cycles,¹⁰⁸ it is important to identify such contributions and eliminate them from the global signal. The easiest path to take here involves elimination of the entire global signal from measurements by simply regressing it from the time series of the signals in each voxel, using linear regression.¹⁰⁹ An important caveat to this apparently straightforward procedure is that the mathematical techniques involved can introduce spurious relations between fMRI measurements in different brain voxels. Opposite conclusions have been reached by senior brain imaging experts concerning whether to regress out the global signal or not.¹¹⁰ This has left researchers in a quandary as to how to proceed, a quandary that has been highlighted by the contending groups in joint papers.¹¹¹ Such difficulties are compounded by the possibility that the global signal might contain biological information concerning brain function, for which there is evidence.¹¹² All in all, there is a need for caution when deliberating on fMRI observations claiming to delineate different functional networks of activity in the brain.