Over the past 23 years I have seen many changes in the hardware used on MRI systems.
I summarize them below from the perspective of walking up to an MRI scanner for a scan.

The first difference you see in an MRI scanner today compared to 23 years ago is the console controlling the scanner is smaller and more modern looking. There is a mouse, a graphical user interface (GUI), and a high-resolution liquid crystal display color monitor. The scan room may have a video camera for monitoring the patient rather than a window through a faraday shield. The magnet room no longer has iron shielding to contain the fringe field from the magnet because most magnets are self-shielded to contain the fringe Bo field.

In the magnet room we see a magnet with a more stylish appearance and a smaller footprint. The predominant clinical field strength has gone from fractions of a Tesla to 1.5 T in the late 80s to 3 T by the early 2000s [1]. Owing to the higher magnetic fields, it is very rare to see a resistive or permanent magnet on an MRI system today.

In the 1980s and 90s magnets required liquid nitrogen refill every week and liquid helium refill every month. These cryogens vented to the atmosphere. Today, you will often hear the sound of a compressor in the magnet room. The compressor sound comes from a refrigeration unit found on many magnets that eliminate the need for the weekly nitrogen fill and extends the helium hold time. There are even commercial systems that nearly eliminate the need for a helium refill. In addition to horizontal-bore magnets we see geometries which appear more like horseshoe magnets, called open systems, that allow the patient more visibility and minimize the claustrophobic feeling. With the horizontal-bore magnet you will find slightly more space then there may have been in a manufacturer's system two decades earlier.

Decades ago there were predominantly two types of radio frequency (RF) coils: the body coil and the head coil. On a horizontal bore magnet there were saddle coils. The saddle coil was replaced with the birdcage coil, which is slowly being replaced by the phased array coil. Initially, RF coils were transmit and receive coils. Since the early 1980s we saw the introduction of a variety of extremity coils [2]. Some of these coils were transmit/receive coils but the majority were receive only, utilizing the body coil as the transmitter.

Unfortunately, the MRI scan will not be any quieter. Gradient coils are shielded to allow faster switching times without eddy currents. This and higher power gradient amps causes a sequence to be louder. Although this is an inconvenience for the patient, many gradient echo and echo planar imaging (EPI) sequences are only possible because of this advance. In the USA, the maximum dB/dt from a gradient pulse is now limited by the value that causes nerve stimulation in the patient.

The time to acquire a single image has decreased for several reasons. First, there are faster sequences such as EPI and fast spin echo. There are also phased-array RF coils that when used with parallel imaging sequences can reduce the acquisition time by close to the number of coils in the array.

The digitizer speed has increased from ~16 kHz to ~2MHz. This increased speed has come with digital filtering which allows the imaging of smaller field of views within a larger object without wraparound.

I mentioned many of the aesthetic aspects of the computer earlier, but we see the biggest improvement in compute power once the raw data is collected. Just as with personal computers, computer speed, memory, and software has improved on MRI systems. In the mid 80s the speed of computers was so slow that an array processor was found on many MRI systems for the fast Fourier transform (FFT) of a 256x256 pixel image. Today the FFT of volume images with a 512x512 size is performed by the cpu. In the 80s it was thought that EPI might require optical image processing to perform the FFT rapidly enough to see dynamic MRI from EPI [3]. This was not necessary as processing speed increased on the computers.

Images are rarely transferred to film. Images are transferred electronically to a picture archival and communication systems (PACS) which may be accessed by radiologists both within and outside a hospital. A courtesy copy of images is routinely distributed to patients on a CD for viewing on their computer.

In the equipment room the number of 19" racks has decreased. The RF frequency synthesizer is on a single board along with the pulse shaping hardware and RF detector.

I expect advances in hardware to continue. For example, I am sure that routine clinical magnet strength will increase as the US FDA has approved 8 Tesla field strengths for clinical studies with all humans over the age of 1 month.