When your MRI looks fine but your cartilage isn't

Your knee has been grumbling for months, yet the MRI report lands on your GP's desk with the words 'mild degenerative change' — or sometimes nothing remarkable at all. That gap between what you feel and what the scan reports is not imaginary, and it is not simply a matter of pain tolerance. It reflects a genuine limitation in what standard MRI is designed to show.

Conventional MRI reads the architecture of a joint: the shape of cartilage, its thickness, the presence of tears, bone-marrow changes. What it cannot reliably detect is the earlier, chemical stage of deterioration — the point at which the internal structure of cartilage begins to break down before any surface damage is visible. Compounding this, cartilage itself carries no nerve supply. Pain in an arthritic knee originates from the synovium, the subchondral bone, and the joint capsule — structures that only become irritated once deterioration is already well under way. By the time a scan flags meaningful structural loss, the cartilage has often been quietly declining for some time.

T2 mapping works on a different principle. Rather than reading the shape of cartilage, it measures how water behaves inside the cartilage matrix — a distinction that matters considerably when the goal is to find change early enough to act on it.

What T2 mapping actually measures inside cartilage

Think of healthy articular cartilage as a firm sponge with a very tight internal scaffold. The scaffold — a dense, precisely organised network of collagen fibres — holds everything together and keeps water molecules constrained within it. Because that water cannot move freely, a T2 mapping sequence records a low relaxation time: typically around 32–35 milliseconds in healthy tissue.

When early osteoarthritis takes hold, two things happen simultaneously. Collagen fibres begin to lose their orderly arrangement and proteoglycans — the large molecules that help anchor the scaffold — start to deplete. With the scaffold loosening, water molecules gain mobility. The T2 signal rises accordingly, commonly reaching 40–47 milliseconds in the superficial cartilage layers during early degeneration. The sponge is getting wetter and looser, even though its outer surface still looks intact.

Critically, those numbers should be read as relative markers rather than fixed diagnostic thresholds: what matters clinically is the direction and pattern of change across cartilage subregions, not a single figure in isolation. A 2025 ex vivo study clarified the mechanism further — T2 elevation specifically tracks the anisotropic, structurally organised component of the collagen network, confirming T2 is a signal about matrix integrity rather than hydration alone.

All of this happens before any surface crumbling is visible. That is the clinical point: T2 mapping adds a quantitative biochemical layer to what a standard MRI already shows about shape and structure — providing an earlier signal, and therefore an earlier decision point — without replacing the radiologist's structural read.

The reversibility window — why timing is the whole point

Adult articular cartilage cannot repair itself in any meaningful sense. There are no cartilage stem cells waiting to replenish lost matrix; once full-thickness structural loss is established, the clinical options shift — from preserving what remains to repairing, augmenting, or ultimately replacing the joint surface. That biological reality is what makes timing matter so much.

The reversibility window describes a phase that precedes structural collapse. At this stage, the cartilage matrix has undergone biochemical change — loosening of the collagen scaffold, early proteoglycan depletion — but gross architecture is still intact. The question of whether biochemical deterioration can be modified at this point is not merely theoretical: a study of 111 obese adults using dGEMRIC imaging found that one year of weight loss produced a measurable increase in cartilage proteoglycan content. That is direct evidence that the matrix, under the right conditions, can respond positively — not that damage simply reverses on its own, but that the tissue retains some capacity for biochemical improvement when the mechanical and metabolic load is addressed.

Waiting for a standard MRI to flag the problem shortens this window considerably. Published data suggest conventional MRI misses somewhere between 30 and 55 per cent of early cartilage lesions that T2 mapping can detect — lesions where morphological changes are still subclinical and the biochemical signal is the only indicator that change is under way.

It would be misleading, however, to suggest that a raised T2 value guarantees intervention will succeed, or that there is a universally agreed threshold beyond which the window closes. The T2 cutoffs associated with reversible versus irreversible change vary across joint regions and patient populations and are not uniformly established in the literature. The reversibility window is best understood as a concept grounded in mechanism and supported by intervention evidence — a clinically useful frame for why earlier detection matters, rather than a precise line any individual scan can reliably draw.

How onMRI™ turns T2 mapping from research tool into a clinical metric

The science behind T2 mapping has been established for well over a decade. The reason it has not entered routine clinical practice is not scientific scepticism — it is practical friction.

Three barriers have historically limited uptake. Acquiring a T2 mapping sequence adds meaningful time to a scan appointment. There is no universally adopted acquisition protocol, which means T2 values recorded on one scanner in one centre are not automatically comparable with those from a different machine elsewhere — undermining the longitudinal tracking that makes biochemical imaging clinically useful. And cartilage segmentation has traditionally relied on experienced radiologists applying semi-quantitative systems such as MOAKS, which carries known inter-reader variability even between trained specialists reviewing the same scan. For a measurement meant to detect gradual change reliably over months or years, subjective grading is a structural weakness.

Software that automates this pipeline directly addresses each problem. Boundaries are traced algorithmically, removing reader variability and producing a consistent numerical output. Independent research is beginning to quantify what that shift delivers: a 2025 deep-learning study (DL CartiGram) achieved a 40% reduction in scan acquisition time while maintaining a coefficient of variation of 0.97% and inter-site reproducibility of CCC 99%. AI segmentation models published in May 2026 trace cartilage and bone boundaries to 0.17 mm precision — a level of consistency manual grading cannot reliably match.

onMRI™ (patent-pending) applies this approach within MSK Doctors consultations. It automates segmentation, integrates T2 quantification, and operates on low-field open-bore equipment — including the Open MRI in Sleaford — which matters for patients who cannot tolerate a conventional closed-bore scanner. The output is a set of reproducible numerical metrics that can be directly compared against a baseline scan taken months later. The consultant synthesises those numbers alongside the full clinical picture; the technology narrows the margin of measurement uncertainty, not the margin of clinical judgement.

What the evidence shows — and where the gaps remain

Three layers of evidence support T2 mapping as a clinically useful biomarker, alongside gaps the technique has not yet resolved.

What the evidence establishes. A 2025 study in early knee OA found T2 values at the medial femoral condyle, medial tibial plateau, patellar, and trochlear regions were significantly elevated versus healthy controls, achieving an area under the curve of 0.78 for early KOA diagnosis — meaningful discriminative ability, though not a definitive standalone test. At population scale, the Rotterdam Study (673 women, 1,332 knees) confirmed that T2 values are consistently elevated across cartilage segments in OA knees and correlate with BMI across multiple subregions, supporting T2 as an epidemiological biomarker rather than a laboratory curiosity. T1rho/T2 mapping can also stratify OARSI histopathological grade preoperatively — evidence that the signal maps to real tissue pathology, not just a statistical association.

Where gaps remain. Correlations between T2 values and patient-reported symptom scores such as KOOS are weak in some population studies. A patient with elevated T2 values may not have proportionally more pain; someone with significant pain may show modest T2 change. That decoupling is real and clinically important: T2 data describes the cartilage matrix, not the patient's pain experience. Additionally, T2 values are load-sensitive — a 2019 PMC study in young professional footballers found significant differences between early and late unloading conditions (P<0.001) — meaning standardised acquisition protocols are necessary for results to be meaningful across time points. Precise T2 cutoffs separating reversible from irreversible change also vary by joint region and have not been uniformly established in the literature.

Taken together, the evidence positions T2 mapping as a well-validated biochemical signal that informs clinical assessment rather than replacing it — one input, alongside symptoms, history, and examination, in a consultant-led decision.

What early detection actually changes for your OA trajectory

Knowing earlier does not just mean knowing sooner — it changes what remains possible.

A T2 finding within the reversibility window shifts the clinical conversation from symptom management to something more specific: a structured review of mechanical load, body composition, activity pattern, and whether a biologic or regenerative option is appropriate at this stage. The range of interventions a patient can realistically consider is broader here than it will be once collagen architecture is destroyed and structural erosion is established. That follows directly from the biology of adult cartilage, which cannot meaningfully rebuild once the extracellular matrix has been lost — not a motivational framing, but a constraint the tissue itself imposes.

For patients who already carry an OA diagnosis, serial T2 mapping adds something that symptom scores alone cannot provide: an objective, quantified measure of whether a treatment is slowing biochemical deterioration. Pain frequently lags structural change in both directions — improving before cartilage recovers, persisting after it stabilises. A reproducible numerical biomarker separates those threads, giving both clinician and patient a clearer basis for deciding whether to continue, escalate, or modify a plan.

Within the Regenerate | Repair | Replace framework, early T2 change sits at the Regenerate end — the stage before tissue-level restoration procedures such as MACI or autologous chondrocyte implantation, or eventual joint replacement, become the primary conversation. The point is not to avoid surgery indefinitely; it is to reach any future decision with the full biochemical picture in hand and the widest set of options still open.

At MSK Doctors' Sleaford site, Open MRI, onMRI™ analysis, and the Regeneration Hub sit within one clinical pathway — from scan to consultant-led plan, without GP referral or NHS waiting time. London-based patients can access equivalent consultant-led assessment through the London Cartilage Clinic. Appointments can be booked directly at mskdoctors.com.

1. [1] Orientation-Independent T2 Mapping Enhances MRI-Based Cartilage Characterization. (2025). https://doi.org/10.1007/s10439-025-03774-3 https://doi.org/10.1007/s10439-025-03774-3
2. [2] T2 mapping of the articular cartilage as a biomarker for knee osteoarthritis: An analysis of the population-based Rotterdam Study. (2025). https://doi.org/10.1016/j.joca.2025.09.009 https://doi.org/10.1016/j.joca.2025.09.009
3. [3] Exploring the asynchronous changes of articular cartilage and osteochondral junction in knee osteoarthritis based on T2 mapping and ultrashort echo time imaging. (2025). https://doi.org/10.21037/qims-24-1492 https://doi.org/10.21037/qims-24-1492
4. [4] Rapid and robust quantitative cartilage assessment for the clinical setting: deep learning-enhanced accelerated T2 mapping. (2025). https://doi.org/10.1007/s00256-025-05034-w https://doi.org/10.1007/s00256-025-05034-w