Founded in Wellington and now used in clinics worldwide, Volpara provides physics-based, FDA-cleared breast-health software that helps clinicians assess cancer risk, optimise image quality, and support more personalised screening decisions. Over more than a decade, the company scaled from early grants to public listing, U.S. expansion, and two acquisitions.
From Oxford labs to an ASX listing, Volpara’s story is a playbook for deep-tech founders: standardise before you “do AI,” sell direct before you sign exclusives, and ship when it’s “good enough” to unlock the next gate.
Key Takeaways
<aside>
Quick Navigation
In the 1990s basement of Oxford’s Robotics Research Group, Ralph Highnam and colleagues taped mammography films to lightboxes and photographed them with robot cameras. On a screen, the same breast image looked starkly different depending on the lighting.
“We realised if you just throw machine learning at raw images, it won’t be reliable,” Highnam recalls.
The fix would be heretical for the AI zeitgeist for decades: standardise the physics of the image first, then learn from it. That decision—normalise before you optimise—became the technical spine of Volpara Health.
By 2009, digital mammography was finally mainstream. Highnam—by then in Wellington—re-formed an old alliance with Sir Mike Brady (Oxford), Nico Karssemeijer (Nijmegen), and Martin Yaffe (Toronto). The idea was simple and audacious: turn decades of breast-density research into robust, clinic-ready software that would help personalise screening for millions of women.
Volpara’s first spark was funded not by VCs but by micro-grants: a NZ$10k seed from Grow Wellington, support from Callaghan Innovation, and donated IP legal work.
That scaffolding let Highnam translate physics into product.
<aside>
Ralph x Angus: The Volpara Health Top 10
Top 10 takeaways from Volpara founder Ralph Highnam, in conversation with Angus Blair, General Partner at Outset Ventures.
</aside>
Breast density is clinically tricky: dense tissue and tumours both appear bright on a mammogram, making cancers harder to see. The concept had circulated for years; what didn’t exist was a standardised, automated way to measure it at scale and make it legible to clinicians in the moment.
Volpara’s bet was that physics-grounded algorithms—quantifying breast composition—would travel better across vendors, hardware quirks, and environmental artefacts than end-to-end black-box models. The market inflection—digital imaging everywhere, clinics hungry for workflow-ready tools, and a growing regulatory consciousness—made 2009 a now-or-never window.