A modern MRI scanner should lose almost no liquid helium, because a machine on the roof re-condenses it, and the level is telemetered to the manufacturer daily. Scanner 2 loses about four percent a month and every part of the cooling chain tests good.
I am the imaging engineer for a trust with five scanners.
Why the helium level is watched so closely.
The magnet is superconducting, which means the wire has no resistance and the current in it runs for ever, and it only does that while it is very nearly at absolute zero. The coil sits in a bath of liquid helium inside a vacuum vessel.
A cold head on the roof, driven by a compressor, re-condenses the helium that boils off, so a healthy modern system loses nearly nothing. Ours are specified at under one percent a year.
If the level falls far enough the magnet quenches: the helium boils off in a few minutes, out through the quench pipe, and the field collapses. That is a hundred thousand pounds and a scanner out for a fortnight.
So the level is on a gauge, on the console, on a telemetry feed to the manufacturer, and it is checked every day by the radiographers and every week by me.
Scanner 2 loses about four percent a month.
Not four a year. Four a month, which is a fill every two years instead of never, and about eleven thousand pounds a time.
What has been checked, and the manufacturer has had four engineers on it.
The cold head. Replaced twice. Its performance is measured directly: the temperature it achieves at the first and second stages, against a curve. Both stages are within specification and the second stage is better than specification.
The compressor. Replaced, and the helium supply lines to it flushed and recharged.
The vacuum. A magnet's insulation is a vacuum vessel, and a poor vacuum is the classic cause of high boil off, because heat gets in by conduction. The vacuum was measured at ten to the minus six millibar, which is a good vacuum, and it was measured again a year later and it is still good.
The quench pipe. A leaking quench valve would let helium out of the top and that is the other classic cause. Pressure tested. It holds.
The room temperature, the chilled water to the compressor, the power supply. All fine, all logged.
The manufacturer's fourth report says, in the summary, that all cryogenic subsystems test within specification and that the observed boil off rate is not consistent with the measured performance of the system.
The arithmetic is the part that stops me.
Boil off is heat. To boil a litre of liquid helium you have to put about two point six kilojoules into it. Four percent a month on that magnet is around seventy litres a year, which works out at about six watts arriving in that bath continuously.
Six watts is not much in a hospital. It is a small light bulb.
But every route for six watts to get in has been measured. The vacuum is good, so it is not conduction. The cold head is over performing, so it is not a failure to re-condense. The pipework is sound, so it is not a leak taking cold gas out.
Six watts is going in and there is no path for it, and I have had that conversation with two engineers who both went quiet and then said they would raise it internally.
What we do.
We fill it. Twice so far, and the trust has the cost in its equipment budget as a known issue on that scanner, and it is cheaper than replacing a magnet.
The service contract covers the engineer visits and not the helium, which was argued about for a year and settled.
The scanner works. The images are excellent, the field homogeneity is as good as any of the five, and there has never been a quench.
What I do.
I read the level every Monday and I plot it, and the plot is a straight line going down at a rate I could draw from memory.
The one thing I do that is not in the manual is that I read it in person rather than off the telemetry.
The gauge is on the front of the magnet in the scan room, and the scan room is the one with the five gauss line painted on the floor and the sign about pacemakers, and going in means the usual checks.
I did it off the console for a couple of years because it is the same number.
Since 2022 I have gone in, and the reason is not the number. A magnet at four kelvin makes a noise, a slow soft ticking from the cold head, about one a second, and I have listened to that sound in scan rooms for eleven years.
In scanner 2's room the ticking is the cold head, and there is a second one underneath it, at a slightly different rate, so the two go in and out of step across about a minute, and the cold head is the only moving part in that room, and I have stood there with the compressor isolated and heard the other one carry on.