How was fMRI performed before EPI became available?

Question

How did fMRI work before Echo-planar imaging (EPI) became available? EPI allows fast imaging and high temporal resolution, but it became commercial in the second half of the '90s. How was it possible to perform fMRI in the early '90s and to observe neural activation using slow gradient echo imaging which requires 3 minutes to acquire a whole image?

Answer 1

Short answer
Older image collection techniques like GRASS and FLASH are slower than EPI, but still fast enough to operate within the time span of the neurovascular delay.

Background
The speed of fMRI processing is an important limitation to what processes can be investigated with this technique. Figure 1 shows a number of physiological processes along with a time line of a number of key fMRI strategy developments (source: Cohen). Figure 1 shows a steady decrease in MR imaging times in the first decade of its conception. The cerebral blood flow response used in fMRI BOLD imaging typically begins within ~500 ms and peaks 3–5 s after stimulus presentation (Hilman 2014). fMRI BOLD is based on the neurovascular coupling, referring to the relationship between local neural activity and subsequent changes in cerebral blood flow (CBF). Any imaging technique that utilizes the neurovascular coupling, notably fMRI BOLD, should be fast enough to operate withing= the time window of the neurovascular delay.

Echo-planar imaging (EPI), was already known by 1977, but was implemented not until 1990 for fMRI. EPI collects image data in 40 to 150 milliseconds and hence is more than sufficient to obtain a BOLD response. Earlier image collection techniques (gradient echo methods) such as Fast Low Angle Shot (FLASH) and Gradient Recalled Acquisition in Steady State (GRASS) are much slower and need several hundreds of ms (FLASH) or several seconds (GRAS) to obtain an image. Nonetheless, these processing times are still sufficient to obtain a cerebral blood flow response.

^{Fig. 1. Duration of various physiological processes (top) and development of imaging techniques and their image collection times (bottom). The neurovascular peak delay (around 8 sec as depicted here) has a time window sufficiently long to be recorded by GRAS, TurboFLASH and EPI have processing times within that time window and can theoretically be used for fMRI. source: Cohen}

<sub>Reference
- Hilman, Annu Rev Neurosci (2014); 37: 161–81</sub>