Imaging Biomarkers and the Next Phase of Radiation Oncology

Radiation oncology is moving towards treatment decisions that reflect the biology of an individual tumour, rather than relying solely on anatomy, stage and population averages. Imaging biomarkers sit at the centre of this shift. They convert features seen on CT, MRI, PET or functional imaging into measurable indicators of tumour behaviour, treatment response and normal-tissue risk.

For medical physicists, this field connects image acquisition, reconstruction, quality assurance, data science and treatment planning. A biomarker is useful only when its measurement is repeatable, its biological meaning is credible and its result can influence a clinical decision. A visually impressive image has limited value if the signal changes with scanner settings, patient positioning or inconsistent analysis.

The subject is highly relevant to Australian services. Large centres in Sydney, Melbourne, Brisbane and Perth may have access to advanced PET/MRI, adaptive radiotherapy and research infrastructure, while regional and remote services often work with fewer scanners, longer referral pathways and limited specialist staffing. Any practical model must account for that uneven landscape.

The archived COMP 2014 programme offers a useful professional context for considering how imaging science develops through shared research, technical standards and clinical discussion. The same principles remain important as Australian departments assess radiomics, artificial intelligence and quantitative imaging for routine care.

Why Imaging Signals Matter

Traditional radiotherapy imaging answers important anatomical questions: where is the tumour, how large is it and which organs are nearby? Imaging biomarkers add another layer by asking whether a lesion is hypoxic, proliferating, perfused, metabolically active or responding to treatment. These measurements may support risk stratification, target definition or adaptation during a course of radiotherapy.

Potential biomarkers include standardised uptake values from FDG-PET, diffusion coefficients from MRI, dynamic contrast-enhanced parameters and texture features extracted from CT or PET. Their usefulness depends on context. A high uptake value may indicate aggressive disease, inflammation or technical variation, so it should not be treated as an isolated truth.

From Pixels to Patient Biology

The translation from image intensity to biology involves several stages. Acquisition protocols must be stable, scanners must be calibrated and segmentation must be reproducible. Pre-processing choices such as resampling, denoising, intensity normalisation and motion correction can materially alter a radiomic result.

Medical physicists are well placed to test this chain. Phantom measurements, repeat scans and inter-observer studies can reveal whether a proposed feature reflects the patient or the equipment. A biomarker that performs well in one Melbourne hospital may lose reliability when applied to a different scanner in Adelaide or a regional Queensland service.

Where Biomarkers Can Change Care

Imaging biomarkers may help identify patients who need intensified treatment, closer surveillance or an alternative strategy. In head and neck cancer, metabolic response during or after treatment could support earlier recognition of residual disease. In prostate cancer, multiparametric MRI features may refine target definition and help distinguish clinically significant lesions.

They may also contribute to adaptive radiotherapy. Daily cone-beam CT, repeated MRI or interval PET can show anatomical and biological change, allowing clinicians to consider replanning. The strongest applications will be those linked to a defined action, such as changing margins, escalating review, modifying dose distribution or arranging additional diagnostic assessment.

Evidence must remain stronger than enthusiasm. A predictive biomarker indicates what treatment is likely to work; a prognostic biomarker describes expected outcome regardless of treatment. Confusing these categories can lead to inappropriate escalation, especially when access to advanced imaging differs between Australia’s private and public sectors.

Biomarker Families to Prioritise

A department assessing new quantitative imaging methods can begin with applications that have a clear clinical question and a manageable technical burden. The following groups are commonly relevant to radiation oncology research and service planning:

• Metabolic markers: FDG-PET uptake, tumour-to-background ratios and other measures of glucose or tracer activity.
• Diffusion markers: Apparent diffusion coefficient and related MRI measures that may reflect cellular density or treatment response.
• Perfusion markers: Contrast-enhanced CT or MRI parameters associated with blood flow, permeability and vascular behaviour.
• Radiomic descriptors: Shape, intensity and texture features extracted through controlled, reproducible software pipelines.

Each group requires careful interpretation. Radiomic features can be sensitive to voxel size, reconstruction kernel and segmentation method, while diffusion measures may vary with acquisition sequence and magnetic field strength. A short list of robust features is generally more useful than a large catalogue with uncertain repeatability.

The Australian market adds practical constraints. New imaging software may involve subscription costs, integration work and cybersecurity review, while procurement teams must weigh those costs against patient benefit and staff capacity. If a platform functions as regulated clinical software, its status with the Therapeutic Goods Administration should be checked; radiation use and practice requirements also vary across state and territory legislation.

Validation in Australian Practice

Validation should progress from technical repeatability to clinical association and then to clinical utility. A multi-site study is valuable because it tests scanner differences, population variation and workflow realities. Australian centres can strengthen evidence by including both metropolitan hospitals and regional networks rather than treating tertiary datasets as universally representative.

Data governance must be designed from the beginning. Imaging linked with pathology, genomics or treatment outcomes may constitute sensitive health information under the Privacy Act 1988 and applicable state or territory rules. Secure storage, controlled access, de-identification and transparent consent arrangements are essential, particularly when commercial vendors or cloud processing are involved.

Local workflow matters as much as statistical performance. A biomarker requiring a specialised scan may be difficult to implement for patients travelling from the Northern Territory or western New South Wales. Appointment availability, Medicare funding arrangements, transport, interpreter access and the division of responsibilities between radiology, nuclear medicine and radiation oncology all affect whether a research result can become a dependable service.

Questions for a Safe Implementation

Before adopting an imaging biomarker, a multidisciplinary team can test its readiness against practical questions:

• Measurement: Is the acquisition protocol standardised across scanners, sites and time points?
• Reproducibility: Do repeated scans and independent observers produce comparable results?
• Clinical action: What treatment or review decision will change when the biomarker crosses a defined threshold?
• Governance: Are privacy, software validation, cybersecurity and regulatory responsibilities documented?

The team should also define failure conditions. Missing scans, motion artefact, incompatible protocols and incomplete clinical records can create misleading certainty. Results should be visible with appropriate limitations, rather than presented as definitive scores without confidence intervals or quality flags.

Education supports safer uptake. The COMP student chapter activities illustrate how professional communities can connect early-career physicists with technical learning and collaborative exchange. For Australian trainees, that culture can extend through hospital research groups, the Australasian College of Physical Scientists and Engineers in Medicine, university partnerships and supervised work with data scientists.

Building a Translational Culture

Imaging biomarker projects work best when physicists, radiation oncologists, radiologists, nuclear medicine specialists, MRI scientists, therapists, statisticians and patients define the problem together. Patient perspectives are especially important when extra scans, longer appointments or uncertain findings may create anxiety or financial pressure.

The field also needs transparent reporting. Studies should describe patient selection, scanner models, acquisition settings, segmentation methods, missing data, preprocessing and external validation. Open methods and carefully governed data sharing make it easier to distinguish a reliable signal from a result shaped by local practice.

For Australian services, staged implementation is sensible. A department might begin with a quality-assured research protocol, audit technical stability, compare outcomes across population groups and assess workload before embedding a biomarker in routine planning. This approach respects the realities of public hospitals, private providers, regional referrals and competing demands on imaging equipment.

The most valuable biomarker will be one that improves a decision without creating unnecessary complexity. It should be measurable in the intended setting, explainable to clinicians and patients, and supported by evidence that extends beyond a single institution. Medical physicists can lead that discipline by treating biological imaging as a measurement science as much as an innovation programme.

Map one clinically important decision, document the imaging pathway that supports it and build a validation team around the result. That practical starting point can turn promising quantitative images into safer, fairer and more effective radiation oncology care across Australia.