The Level 2 method for MRI: history, map, test, boundary

Take any MR fault from the first complaint to a named subsystem with a fixed method - error history in time order, subsystem map, test scans and phantoms, a written boundary of what you may change - while keeping yourself and others safe in a field that never switches off.

Where Level 1 stopped

Level 1 gave you a scanner you can walk around safely, a set of readings you take on every visit, a weekly phantom trend and a report that holds up. That is enough to keep a healthy MR system healthy. Level 2 starts on the day the scanner stops a protocol halfway, refuses to start the RF, loses its centre frequency or produces images the radiologist will not read. From that moment you are not checking a list; you are building an argument that ends with one subsystem, one module or one cable.

MR is the hardest modality to fault-find for a simple reason: everything talks to everything. The gradients sit inside the magnet and heat it; the RF chain shares a room with the lights, the patient monitor and the injector; the cooling water serves the cold head compressor, the gradient amplifier and the RF amplifier at the same time; and the computers stand behind all of it. An error message names the module that noticed the problem, which is often not the module that caused it. The method in this lesson exists to stop you replacing the noticing module.

The method has four parts, always in the same order: the history, the map, the test and the boundary. None of them needs a screwdriver.

The history: errors in time order

Every MR system keeps logs: an error or event log for the operator, deeper service logs, test reports, and on many platforms a record of the magnet monitor - helium level, pressure, cold head and compressor status. Before you touch anything, read them in time order across at least several days before the first complaint. You are looking for three things: the first entry of the chain, the entries that always come together, and anything that changed - a power cut, a software update, a chiller service, a coil repair, a move.

The first entry matters most. When a gradient amplifier trips, the log may show a dozen messages within a second: amplifier fault, sequence aborted, communication timeout, reconstruction error. Only the first is close to the cause; the rest are consequences. Engineers who read the newest message first waste days on the last domino. A common trap is the operator-level message that sounds specific but is generic - 'table error', 'RF fault', 'hardware failure' - while the service log underneath names a stop button, a voltage dip or a flow switch.

Then put the log next to the outside world. Ask the staff when it happens: mornings only, after long diffusion protocols, on hot afternoons, when a certain coil is plugged in, when the lift next door runs. Ask what was done recently and by whom - building electricians and chiller contractors rarely tell the MR engineer. Write it all down as a timeline. A good timeline often names the subsystem before you open a single cover.

The map: which subsystem can produce this?

Level 1 split the scanner into four subsystems: magnet, gradients, RF and computer. At Level 2 you add the two that cause half of the real calls: cooling and cryogenics, and the room - power, earthing, RF cage, climate. For each complaint, draw the chain of subsystems that could produce it, and cross off those the evidence excludes.

Think in signal paths. An image is built from a magnet field that must be stable and uniform, gradient pulses that must be exact in shape and timing, an RF pulse at the right frequency and power, a tiny received signal that must travel clean from coil to receiver, and a reconstruction that must not lose data. A wrong value at any step leaves a recognisable fingerprint, and lesson 8 builds that atlas. A stop with an error is the same idea in time: the module that supervises a current, a temperature, a flow or a reflected power has seen something outside its window.

Use the map to choose the first measurement, not the first part. If the map says 'cooling, gradient amplifier or gradient coil', the first job is to read water temperatures and flow, not to order an amplifier.

• Magnet and shim: field drift, homogeneity, centre frequency, magnet monitor alarms.
• Cryogenics and cooling: helium level and pressure, cold head, compressor, chiller, secondary water loop.
• Gradients: amplifiers, power supplies, cables and filters, gradient coil, eddy current compensation.
• RF transmit: RF amplifier, transmit switching, body coil, SAR supervision, the RF room.
• RF receive: coils, preamplifiers, coil connectors and table sockets, receivers, optical links.
• Table, monitoring and peripherals; computers, reconstruction, disks, network and software.

The test: built-in tests, phantoms and moving the fault

A hypothesis is only as good as the test that can prove it wrong. MR scanners give you three kinds of test. First, the built-in service tests: gradient, RF, receiver, coil and spike checks, centre frequency determination, and on many platforms a quality test run on the manufacturer's phantom. Which of these an independent engineer may run depends on the access level the owner and manufacturer grant; never try to obtain a higher level than you were given. Second, the QA phantom and the weekly trend you set up at Level 1: if the trend moved before the complaint, it tells you when the fault began. Third, a test scan of a phantom with a protocol that reproduces the complaint - the same coil, the same sequence, the same gradient load.

The most powerful field technique is moving the fault. Swap a coil to another socket, a cable to an identical channel, a board between two identical racks or axes - only where the manufacturer's documentation treats them as interchangeable and only with the power safely off. If the fault follows the part, the part is guilty; if it stays, the slot, the cable or the supply is. Write down every swap and put everything back where the documentation says it belongs, because some boards carry addresses or calibration for their position.

Do not trust a single pass. Intermittent faults are the rule in MR, not the exception. A spike check that passes once out of ten runs has failed. A test that passes on an empty table may fail with the coils actually used clinically lying on it. Reproduce the clinical conditions, repeat the test, and count.

The boundary: what you may and may not change

Write the boundary down before you start, because MR offers many tempting shortcuts. The rule of this course: you may observe, measure, test, clean, reseat and replace like-for-like modules using the manufacturer's procedure and the access you were legitimately given. You do not change what the manufacturer controls.

If the honest answer is that the next step lies outside your boundary, the job is still yours to finish: collect the timeline, the logs, the test results and the swaps, and hand them to the authorised party in a form that saves them a day.

• Never ramp, de-ramp or shim the magnet, and never edit magnet, shim or field parameters: specialist tasks (lesson 2).
• Never defeat or bridge an interlock, flow switch, temperature sensor, stop button, door contact or SAR/PNS supervisor, even 'just for the test'.
• Never change SAR, stimulation or patient-safety limits, coil configuration files or calibration tables outside the documented procedure.
• Never install unlicensed software, copy licences between systems or use service keys and passwords you were not issued.
• Never open a helium circuit, cryostat port or quench line, and never run test scans with a person in the bore.

MR safety for the engineer doing Level 2 work

Level 2 work puts you in the magnet room for hours, with covers off, tools out and the staff's attention elsewhere. The rules from Level 1, lesson 2, still apply in full: the four-zone model, screening, gatekeeping and the 0.5 mT line. What changes is the exposure. You will be under the magnet covers, on a ladder by the service turret, at the cabinets with stored energy, and running test scans with the door open to watch.

The field is always on. Every tool, meter, ladder, trolley and spare part that enters Zone IV must be labelled MR Safe or MR Conditional for that field, or stay outside. Keep a dedicated set of non-magnetic tools, count them in and out, and never let a part of unknown material near the bore because the job is urgent. A small screw dropped into the bore or the gradient coil area can cause hours of searching or damage.

Respect stored energy in the technical room. Gradient amplifiers and their power supplies hold large capacitor banks; follow the manufacturer's lock-out and discharge procedure and measure before touching. The emergency power-off and the magnet stop are different buttons with different consequences, and in a quench the room atmosphere can lose oxygen: know where the oxygen monitor reads and where the exits are before you start. Wear hearing protection during test scans; the gradients are loud. Never attach yourself, a colleague or a volunteer to a coil to test it - phantoms only.

On-site checklist

• Confirm zone control, screen your tools and check the oxygen monitor before starting.
• Export or photograph the logs covering several days before the first complaint.
• Write a timeline: first entry, repeating groups, outside events, staff observations.
• Draw the subsystem map for the complaint and cross off what the evidence excludes.
• Reproduce on a phantom with the clinical coil and protocol; repeat and count.
• Record every swap and restore original positions.
• Write down your boundary and what you will hand over if the fault lies beyond it.

Check yourself