Putting Harrison.rad.1 to the Test

“Foundation models need to be tested, not just promised,” Dr. Jarrel Seah highlighted.

When we introduced it, we benchmarked it to some common industry standards. Harrison.rad.1 achieved a score of 51.40 on the FRCR 2B Rapids exam, the same credentialing exam taken by radiologists in the UK. Jarrel noted, “That exceeds the average performance of human radiologists retaking that exam within a year of passing.”

But real-world results matter most.

In May 2025, Mass General Brigham AI Arena and the American College of Radiology ran an independent challenge at the ACR Annual Meeting.

The question: could board-certified radiologists tell the difference between AI-generated and human-written reports?

• 113 radiologist
• 2,840 blinded evaluations

The results: reports powered by Harrison.rad.1 achieved a 65.4% acceptability rate, compared to 79.6% for radiologist-written reports.

Reflecting on the real-world challenge, Jarrel noted, “Mass General Brigham AI Arena said that these results show that draft reporting AI is improving at breakneck speed and is closer than ever to meaningfully enhancing radiologist efficiency.”

Why Harrison.rad.1 Performs So Well

The key lies in the data. Harrison.rad.1 is built on millions of DICOM images, radiologist annotations, and structured reports spanning all X-ray modalities.

Jarrel explained, “Why does Harrison.rad.1 perform at this level? Well, it comes down to comprehensive training data. This model leverages the same high-quality radiology data sets that power our Harrison.ai clinical decision support tools.”

“The most comprehensive detection algorithms on the market are now powering, the most promising foundation model for radiology reporting. And that’s not a coincidence,” Jarrel added.

When the Clinical Case Isn’t Enough

A few months ago, a large radiology group processing 150,000 chest X-rays and 35,000 CT brain studies annually, came to Harrison.ai with a familiar challenge. As Caitlin explained:

“Their backlogs were rising and their outsourcing costs were climbing, and their clinical team was convinced that AI could help them solve both issues. The problem was they didn’t know how to justify the technology investment to their board.”

“This is a common story in healthcare, even when the clinical case is strong,” Caitlin added. “Clinical teams understand the value, but CFOs, boards, payers, they need numbers. They need a value analysis, and they need a defensible business case.”

That’s where the Harrison.ai Value Calculator comes in.

Built on Real Deployments and Real Outcomes

To tackle this challenge, the team brought forward a capability shaped by years of real-world deployments and measured outcomes. Mark introduced it clearly, “The Harrison.ai Value Calculator. This tool is built from years of partnering with healthcare organisations that actually measure health impact. Real deployments, real data, real results.”

The process begins by grounding the model entirely in the organisation’s own operational metrics – an approach that avoids assumptions and keeps the analysis anchored in reality. As Mark explained, “First, we collect data from you – key metrics that drive your operations: annual scan volumes by modality, current outsourcing numbers, average reporting times, your patient demographics, and reimbursement rates.”

For this radiology group, those numbers surfaced the underlying strain almost immediately; they were outsourcing 50% of their CT brain volume and had a significant backlog problem.

With this operational picture established, the calculator then incorporates real-world evidence from comparable deployments. As Mark explained, “We combine this with industry benchmarks and demonstrable results that we’ve seen in similar organisations who’ve deployed Harrison.ai.”

Together, these layers create a coherent, organisation-specific model, grounded in evidence and built to clearly demonstrate the economic value of AI in driving clinical improvements.

The Findings: A Compelling Business Case for Radiology AI

When the numbers were pulled together, the value story became undeniable.

1. Reduced Outsourcing

• This healthcare provider was able to bring up to 35% of outsourced brain CT studies back in-house, as Harrison.ai had already identified these as non-urgent.
• $460,000 of annual savings from reduced outsourcing costs alone.

2. Faster Reporting and Backlog Reduction

With AI-assisted prioritisation and comprehensive detection, the average report turnaround time was significantly reduced. Urgent cases were flagged immediately, meaning that radiologists could tackle work lists in priority order instead of chronologically.

“So this really is about workflow transformation. It’s not just making radiologists faster, but it’s making them more effective, focusing their clinical expertise where it really matters the most,” Caitlin added.

3. Sustainability Wins

An unexpected value stream emerged when unnecessary follow-up imaging was reduced: “They reduced their carbon footprint significantly, the equivalent of over 300,000 kilometres driven,” Caitlin explained.

For health systems with emission reduction targets, this has become a powerful additional lever.

A Business Case That Secured Board Approval

Armed with quantifiable clinical, operational, financial and sustainability benefits, Mark shared how this became the business case that the client took to their board. “Fast forward to today, we’re now rolling out multiple Harrison.ai products with this client: chest X-ray, CT brain and CT chest.”

Just as importantly, Mark explains how the real-world impact is continuously measured against the original model, “We’re tracking real value against those same KPIs, measuring the actual impact and showing the value that we projected.”

Built on Clinical Excellence

Draft report generation isn’t magic, Jarrel emphasised. “It simply requires two things done exceptionally well. First, comprehensive detection. Our chest X-ray model identifies 124 findings with accuracy, leveraging over a million images trusted by radiologists in over one thousand sites globally.”

The second requirement is clinical understanding, delivered by Harrison.rad.1, a radiology-specific foundation model. Jarrel highlighted its validation in a rigorous independent healthcare AI challenge, noting that:

“113 board-certified radiologists evaluated over 2,800 reports. And they couldn’t reliably distinguish between AI-generated and human-written reports.”

These combined technologies power Harrison.ai’s Draft Report Generation solution. The solution complies with Australian clinical and regulatory standards and is already being piloted at early adopter sites. 

Radiologists are leveraging the solution as an intelligent AI assistant to draft preliminary reports. The report is subsequently reviewed by the clinician, who can make small edits or additions if required, before signing-off the report. 

AI Assisted Report Generation in Action

Dr. Sajith Karunasena, Clinical AI Consultant at Harrison.ai, took the stage:

“I’m demonstrating our products today integrated with Sectra PACS and Nuance reporting software, a workflow many radiologists use every day. Our AI triage system flags cases requiring urgent attention, so radiologists can prioritise the patients who need them most.”

In the demo, Sajith showed a routine outpatient chest X-ray with findings suggestive of emphysema. The diagnostic assistant pre-populates the report, summarising positive findings upfront and permanent negative findings afterward. “If I wanted to add a recommendation to the referrer to correlate with lung function tests at the end, it would be a simple one-line dictation. Then I can hit sign to send the report to the referrer,” Sajith explains.

Seeing Comprehensive AI in Action

Ben explains, “Harrison.ai CT Chest is the comprehensive CT chest solution clinicians have been asking for. 167 radiological features, including life-threatening conditions, tumours, and chronic diseases, are covered in a single seamless workflow. All findings are detected, measured, and characterised automatically using the same intuitive interface you already trust with our CXR and CTB solutions.”

To demonstrate Harrison.ai CT Chest in practice, Dr. Jennifer Tang, Clinical AI Consultant at Harrison.ai, walks through real clinical cases:

• Case 1: A 67-year-old female presenting with acute shortness of breath. The AI immediately identifies a central and peripheral pulmonary embolism, along with airspace opacification, pericardial and pleural effusion. Even more impressive is the ability to go further, picking up the sclerotic spine lesion, potentially representing metastasis.
• Case 2: A 50-year-old patient with fever undergoing a routine chest-abdomen-pelvis scan. The AI detects a subtle right lower lobe pulmonary embolus, which is easily missed during a standard infection workup, alongside coronary calcification.
• Case 3: A 73-year-old with chest pain and prior aortic surgery. The AI flags a dissection, bilateral pleural effusions, an underlying aortic aneurysm, and a right subclavian pseudo aneurysm.

Taken together, these cases illustrate what comprehensive coverage looks like with CT Chest – a single solution surfacing both the expected findings and those that are easy to overlook. As Dr. Tang described, it means detecting “the obvious critical findings, but also the hidden time bombs.”

The Challenge: Fragmented AI in Healthcare

Across hospitals and radiology practices, teams are struggling with disconnected systems and multiple vendors, each typically requiring its own integration, contract, security review, and performance monitoring. As Josh Duncan explains:

“Let me tell you what we hear every single day. From health systems and radiology practices across the country, often they’re managing 2, 3, 5, sometimes 10 different AI vendors.”

Traditional platform providers have tried to simplify this by becoming “one-stop shops,” but they often impose markups of 30–60%, diverting resources that could otherwise support patient care and radiology workflow optimisation.

This fragmented approach not only drives up costs but also results in subpar workflows and inefficiencies. As Dimitry Tran put it:

“So you are paying more, getting locked in, and worst of all these platform companies, they don’t build clinical applications themselves. They’ve never felt the pressure of delivering AI at scale… It’s hard to build a track if you’ve never built a train.”

The Solution: Harrison.ai Open Platform

Harrison.ai’s approach to overcoming fragmented AI is built on three guiding principles: zero markup, radically open architecture, and customer ROI first.

The platform has launched with partners including Lucida Medical, AZmed, Nicolab, Radiobotics, CoLumbo, and Us2.AI. These organisations highlighted how a zero-markup, open, and scalable platform can simplify adoption, streamline workflows, and deliver real clinical impact. Their perspectives provide early validation of the platform’s alignment with the needs of healthcare organisations.

Zero Markup

“The platform layer is zero markup. You pay for the AI you use from the algorithm provider and not a penny more. No hidden charges. No revenue sharing that gets passed to you,” Josh Duncan explained.

By eliminating standard platform markup, which can range from 30–60%, hospitals and developers can invest directly in AI that improves diagnostic accuracy, streamlines radiology workflows, and enhances patient care.

“For specialised applications like stroke clinical decision-support algorithms, the traditional platform markup model was prohibitive,” said Michael Macilquham, CEO at Nicolab. “Healthcare systems need the best AI for their patients, not just what’s economically viable after platform fees. Harrison.ai’s approach allows us to focus our resources on clinical validation and customer support rather than funding a middleman’s margin.”

Radically Open Architecture

The platform supports Harrison.ai native applications, third-party algorithms, and even custom in-house AI tools — all within a single, unified system. As Dimitry emphasised:

“We will never block an app on Harrison Open Platform, even if it competes directly with one of our own. No exclusivity deals, no vendor lock-in, and no access barriers.”

Michel Krambousanos, Director of Strategic Alliances at AZmed, welcomed the platform’s open approach, noting that it “supports our shared goal of enabling clinicians to access proven technologies that improve diagnostic accuracy and workflow efficiency” by removing unnecessary restrictions.

This openness gives healthcare organisations the freedom to choose algorithms based on clinical merit, ROI, and patient outcomes, not artificial distribution barriers.

Customer ROI First

At Harrison.ai, customer ROI and clinical impact come first.

“Success is only real when you achieve outcomes,” Josh said. “AI only works when it delivers value in your environment.”

With over 1,000 deployments across 18 countries, as Josh Duncan explains, Harrison.ai has experience integrating with every major PACS, RIS, and EMR. “We ensure AI adoption is seamless within your actual clinical workflows,” he says. Backed by a robust infrastructure and optimised for real-world healthcare delivery, the platform ensures that AI deployment is efficient and reliable across diverse clinical environments.

Reflecting on the platform’s potential impact, Dr. Antony Rix, CEO of Lucida Medical, noted that the Harrison.ai Open Platform “aligns perfectly with our vision of making prostate cancer detection more accessible, accurate, and cost-effective,” adding that “AI should empower clinicians and simplify workflows, not add complexity.”

A Growing Ecosystem

The platform launches with native applications included as standard: Harrison.ai Chest X-ray, Harrison.ai CT Brain, and Harrison.ai CT Chest, detecting a combined 300+ radiological findings.

Partners including Lucida Medical, AZmed, Therapixel, Nicolab, Radiobotics, CoLumbo, and Us2.ai are joining to expand capabilities.

“Some of these algorithms compete with ours – and that’s good. Competition makes for a stronger ecosystem,” Dimitry noted.

“You integrate once… one connection, and you’re done,” says Josh Duncan. Harrison.ai’s vendor-neutral approach ensures a single security review, and seamless access to native, partner, and in-house AI solutions, reducing complexity and supporting smoother clinical workflows.

Reflecting on the ecosystem, Jonathan Whitmore, Director of Global Partnerships at Radiobotics, noted that the platform’s zero-markup model “could be a turning point in how widely AI is deployed in patient care. This shifts AI adoption from a privilege of well-resourced hospitals to a tool that lifts clinical quality everywhere.”