Bone Marrow Oedema

For Medical Students

What It Shows

Bone marrow oedema is high signal in marrow on a fluid sensitive sequence with low signal on a T1 weighted one. 5.2.17 covers the STIR imaging.

It is a signal pattern rather than a disease, and the tissue behind it is usually not oedema at all. 5.11.7 covers bone marrow lesions.

It matters because injury, overload, inflammation, infarction, infection and tumour all produce it, so the signal alone separates none of them. 5.14.7 covers septic arthritis.

How It Is Acquired

Use a fluid sensitive sequence with fat suppression, since marrow fat otherwise hides the change. 5.2.16 covers the fat suppression.

Pair it with a T1 weighted sequence, because marrow replacement and marrow oedema differ there and not on the fluid sensitive images. 5.2.13 covers the T1 imaging.

Record the location relative to the subchondral plate. 5.11.1 covers the normal cartilage.

Look for a fracture line inside the change rather than describing the change alone. 5.13.3 covers stress fractures.

Record what the patient had been doing, because the pattern is load dependent. 5.1.2 covers clinical examination versus imaging.

What Normal Looks Like

Marrow signal following fat on both sequences. 5.5.6 covers the bones.

Red marrow in the expected distribution, which is normal rather than infiltration. 5.2.13 covers the T1 imaging.

A continuous subchondral plate with no signal change beneath it. 5.3.19 covers subchondral sclerosis.

No oedema under a loaded compartment in a well aligned knee. 3.5.3 covers the load distribution.

What Abnormal Looks Like

Subchondral change under a worn compartment, which accompanies arthritis. 5.20.6 covers reporting osteoarthritis.

A focal change with a fracture line through it. 5.13.3 covers stress fractures.

A geographic change with a serpentine margin, which suggests infarction. 5.13.5 covers osteonecrosis.

Diffuse regional change crossing a whole bone end. 5.13.6 covers transient osteoporosis.

Change with cortical destruction or a soft tissue mass, which changes the pathway entirely. 5.13.7 covers bone tumours.

MRCS Level

How It Is Acquired

In a viva the expected answer refuses to call the finding a diagnosis.

The expected points are the fluid sensitive sequence with fat suppression, the T1 weighted pair, the location, the fracture line, and the clinical context.

Say that the term describes a signal and is used as though it named a disease. 5.11.7 covers bone marrow lesions.

Say that the differential is wide and is narrowed by pattern and history rather than by signal. 4.2.9 covers the onset.

Say that the finding is reversible, so the interval from the event matters. 5.13.2 covers stress reactions.

What Abnormal Looks Like

The reasoning expected is that marrow is a confined space inside stiff bone, so anything raising its content or its pressure produces the same signal whatever the cause. 5.5.6 covers the bones.

Histology of these lesions shows fibrosis, necrosis, trabecular microfracture and new vessels rather than free fluid, which is why the name misleads. 5.11.7 covers bone marrow lesions.

The pattern separates the causes: subchondral and focal with a line is mechanical, geographic with a margin is vascular, diffuse and regional is inflammatory or transient, and destructive is neoplastic. 5.13.7 covers bone tumours.

Pain tracks this finding better than it tracks cartilage loss, which is why it is read as significant in arthritis. 5.21.5 covers the osteoarthritis imaging literature.

How Good It Is

Magnetic resonance imaging is the most sensitive available test for marrow change and is close to useless at saying what caused it.

Its best contribution is detecting the change and locating it, which no other modality does. 5.3.1 covers the AP view.

It is limited by non specificity, by reversibility, and by a name that invites a diagnosis the images do not support. 5.1.6 covers imaging pitfalls.

No figures appear for accuracy on this page, under decision 028, and no lesion volumes, signal ratios or grade boundaries appear either: those are published quantities measured differently between studies and this environment cannot check them.

FRCS Level

How Good It Is

At decision level the finding changes a plan only once it has a pattern and a cause attached to it. 5.20.1 covers structured reporting.

Subchondral change under a worn compartment supports an alignment discussion rather than an arthroscopy. 3.13.8 covers alignment and OA.

Focal change with a line is a fracture and is managed as one. 5.13.4 covers insufficiency fractures.

Change crossing a compartment boundary raises the systemic causes and takes the problem out of the knee. 5.13.7 covers bone tumours.

A repeat scan after an interval resolves more of these than any additional sequence does. 5.13.6 covers transient osteoporosis.

Where It Misleads

The commonest error is reporting the finding as a diagnosis in its own right. 5.20.8 covers common reporting errors.

It is read as an injury when the mechanism was repetitive load. 5.13.2 covers stress reactions.

It is read as resolved when the signal fades and the fracture has not united. 5.13.3 covers stress fractures.

It is read as the explanation for the pain in a knee where it is incidental. 5.1.2 covers clinical examination versus imaging.

It is quantified and compared between scanners, which measures the scanners. 5.18.1 covers the quantitative MRI.

Fellowship Level

Where It Misleads

At fellowship level the scan arrives reported as marrow oedema and the referral asks what to do about it. 5.20.8 covers common reporting errors.

The useful next step is usually the history, an alignment film and an interval, rather than another sequence. 5.3.6 covers the long leg alignment films.

Serial scans are compared across scanners and protocols, so part of the change measured is technical. 5.18.1 covers the quantitative MRI.

Metabolic imaging can say whether a lesion is active, which structural imaging cannot, and is rarely available. 5.18.7 covers the PET-MRI.

Automated detection finds these lesions reliably and has the same differential problem a reader has, reported with more confidence. 5.18.8 covers artificial intelligence in imaging.

The honest position is that this is the finding most often converted into a diagnosis it cannot support, and the discipline is to report the pattern and the differential rather than a name. 5.20.1 covers structured reporting.