Essay 01 · Writing · The Missing Ledger series

The biomarkers already latent in the image

A position paper on opportunistic imaging and the economics of prevention

one scan, many signals

There is a version of the longevity conversation that is about buying more imaging, and a version that is about reading the imaging we already bought. Almost all of the public attention goes to the first. Almost all of the near-term value is in the second.

Consider what is physically present in a single abdominal CT scan ordered for an unrelated reason — a kidney stone, abdominal pain, a staging study. The same voxels that answered the clinical question also encode the patient's bone mineral density, the calcium burden in their aorta, the volume of their visceral fat, the mass and quality of their muscle, and the fat content of their liver [S11]. Each of these is a validated marker of future disease. None of them was the reason for the scan, and in routine practice none of them is extracted, priced, or acted upon. The information is acquired and then discarded — not through negligence, but because the system pays a radiologist to answer one question and then moves the queue along.

The case for the upside

The strength of opportunistic imaging is that its marginal cost is close to zero. The scan has been performed. The radiation has been delivered. The patient has been billed. Reading additional, validated risk information off that existing study adds no new imaging, no new appointment, and no new dose. This is why the framing matters so much: opportunistic biomarkers are not a proposal to scan more people, they are a proposal to stop throwing away what the scans already contain.

The individual biomarkers are not speculative. Coronary artery calcium scoring is one of the most thoroughly validated prognostic tools in cardiovascular medicine: it predicts ten-year events in asymptomatic people, reclassifies risk beyond traditional calculators, and — through a score of zero — identifies a genuinely low-risk group who can be spared intervention [S1][S3]. Automated bone-density measurement from routine CT has out-performed conventional DXA at identifying the people who actually go on to fracture (AUC 0.885 versus 0.668), which matters when fragility fractures already cost one health system £4.5 billion in a single year and are projected to rise by nearly a third within the decade [S8][S10]. A radiomic read of liver fat off a low-dose cardiac scan reaches an AUC of 0.91 against both MR spectroscopy and biopsy [S12], and visceral fat measured on cross-sectional imaging is an independent predictor of cardiovascular and metabolic mortality [S13]. Taken together, and automated, these measures can match or exceed the best clinical risk models we currently rely on [S11].

This is the same argument, in convergent form, that opportunistic screening advocates, health economists and imaging researchers have reached independently — a case of parallel play across disciplines that should give it more weight, not less.

The counter-case, which I hold at the same time

None of this licenses indiscriminate scanning, and the moment the argument slides from reading existing scans to ordering new ones for well people, the ledger changes. A meta-analysis of more than five thousand asymptomatic people found critical or indeterminate incidental findings in nearly a third of whole-body MRI studies, a false-positive proportion of 16%, and — the detail that should discipline everyone — not a single study that verified its negatives beyond five years [S4]. A separate synthesis of over nine thousand people put the confirmed-cancer yield of whole-body MRI at 1.57%, with no cost-effectiveness data and no protocol standardisation [S5]. At a base rate that low the positive predictive value of any flagged finding is poor, so most positives are not the cancer being sought — yet each still triggers a real cascade of follow-up scans, referrals and biopsies for a person who was well when they walked in.

The answer is not to abandon the modality but to govern it, because every one of those shortfalls is a governance gap rather than a property of the magnet. Breast screening faced the same low-prevalence problem and solved it with infrastructure, not argument: a structured reporting lexicon, mandated performance metrics — recall rate, cancer detection rate, interval-cancer rate, sensitivity and specificity tracked as ROC curves — and accredited quality standards that let an entire programme be measured and audited. Whole-body MRI screening needs the same scaffolding. ONCO-RADS already supplies the structured-reporting spine[S17]; what is missing is the surrounding clinical governance — explicit guardrails for who is scanned and how each class of finding is actioned, and the reporting systems that let clear performance metrics be developed in the first place. Governed that way, the low base rate stops being an indictment and becomes a design constraint the system is built to manage.

What follows for anyone allocating capital or policy

The practical conclusion is unglamorous and, I think, correct: the highest-return, lowest-risk move in preventative imaging is not a new scan. It is the infrastructure to extract validated value from the scans a population is already having — the workflow to surface a second read without slowing the first, the reimbursement line that pays for information already latent in the image, and the governance framework that assigns accountability when an opportunistic finding is raised. The algorithms are largely built and independently validated [S9][S11]. The bottleneck is organisational, not technical.

Australia is unusually well placed to reason about this, because it has already built bespoke assessment pathways when a technology did not fit the standard mould — the Medical Services Advisory Committee has used a dedicated Clinical Utility Card framework for genetic and genomic tests since 2016 [S18]. Opportunistic imaging deserves the same tailored treatment: assessed on its own terms, with the upside and the incidental-finding burden counted in the same model. That is the standard I argue for, and it is the standard I will keep arguing that we already know how to apply.

References

Dr Lisa Sorger is a consultant radiologist, healthcare executive, medical administrator, company director and founder of myradiologist.ai. She writes on the economics and ethics of preventative imaging and diagnostic radiology. Every claim in this paper is sourced to the public record.

All views expressed here are my own personal opinions and are not medical advice. General information only — not clinical or financial advice. myradiologist.ai · ABN 29 692 758 115 · ACN 692 758 115