Parkway Radiology’s secret to handling 300 chest X-rays a day
To stay at the forefront of patient care, Parkway Radiology needed to manage rising imaging demands while maintaining exceptional accuracy standards. After trialling various AI options, they chose to go with Harrison.ai. Two of their leaders explain why.
Singapore’s leading private radiology provider
Parkway Radiology is part of the IHH Healthcare family, which operates 140 healthcare facilities across 10 countries in Southeast Asia. As Singapore’s largest private provider of radiology services, Parkway Radiology serves four IHH hospitals (Gleneagles, Mount Elizabeth, Parkway East and Mount Elizabeth Novena) and seven clinics – conducting up to half a million radiology procedures annually.
Growing demand for accurate, reliable chest imaging
Chest X-rays comprise the largest volume of their imaging, with 85,000 studies performed each year – or about 300 per working day. This modality is pivotal in numerous settings, including diagnosis and management of heart and lung conditions, acute and emergency care, and in statutory health and pre-employment screenings.
Parkway Radiology employs about 8% of Singapore’s radiologists, but the volume of radiology studies keeps mounting year on year, placing more pressure on the team.
The demand for faster turnaround times (without sacrificing the accuracy of results) and patient expectations also continue to grow.
Faced with these challenges, Parkway Radiology looked to artificial intelligence (AI) to enhance their chest X-ray workflow.
Seeking a solution to satisfy their needs
They had key criteria for choosing a solution, with accuracy being first and foremost, explains Yujuan Tan, Parkway Radiology’s CEO.
It would also have to meet the clinical needs of their radiologists.
Dr Tham Seng Choe, Clinical Director of Parkway Radiology’s Radiologic Clinic at Mount Elizabeth Novena Hospital, says every chest X-ray requires a timely and accurate report – but this task is not always as simple as it sounds.
“We have to compress a 3D body part onto a flattened 2D image which is sometimes difficult to interpret and difficult to read, and diseases can potentially hide in those areas,” he says.
“Sometimes I wish I had a copilot, another pair of eyes, to help me with my daily work. Missing a sub-centimetre faint opacity can change a patient’s life.”
Ms Tan says their ideal solution would weave seamlessly into clinical workflows.
Chest X-rays are reported in under a minute, she explains, and additional clicks or steps are not acceptable when every second counts.
Furthermore, it would need to:
- satisfy all regulatory requirements – from national licensing to cybersecurity
- help mitigate the rising costs of healthcare delivery
- support better patient health outcomes.
Harrison.ai Chest X-ray, the standout solution
They started trialling different tools in 2021, settling on Harrison.ai Chest X-ray – a decision Ms Tan describes as a “no-brainer.”
“The Harrison.ai solution checked all the boxes,” she says.
Dr Tham says they needed a comprehensive tool that could do more than just flag signs of infection or cancer.
“Harrison’s chest X-ray solution delivers more than 120 diagnoses across the board, and this fits the needs of Parkway Radiology since we have such diverse X-ray clients,” he says.
“What’s more, Harrison’s X-ray solution comes in a very nice hanging widget, so this keeps our interface very clean.”
“Harrison.ai Chest X-ray delivers more than 120 diagnoses across the board, and this fits the needs of Parkway Radiology since we have such diverse X-ray clients,” he says. “What’s more, the CXR solution comes in a very nice hanging widget, so this keeps our interface very clean.”
Strategies to support successful implementation
To help implementation go smoothly, Harrison.ai’s solution was deployed in phases, starting in November 2024 with a pilot on one inference server at Mount Elizabeth Novena Hospital, supported by a radiologist-led validation panel.
Full roll-out was completed on a clinic-by-clinic basis, with an average downtime of less than 30 minutes.
Stakeholder buy-in was key to successful deployment, and included on the ground discussions with radiologists, the IT team and RIS/PACS admins. Engagement of the senior management team helped drive the business case.
User training included a two-hour CME accredited session plus on-screen hint overlays.
Faster reporting and better patient care
Ms Tan says Harrison.ai’s solution has bolstered their mission to maintain a “consistently high standard of care,” with the deployment expected to benefit over 100,000 patients annually.
In addition to supporting in-house diagnostics, the technology is being used to analyse chest X-rays referred from other medical institutions, reinforcing Parkway Radiology’s reputation for delivering innovative healthcare.
Dr Tham says having a “second pair of eyes” is helping their radiologists deliver more accurate and timely reports, contributing to reduced turnaround times.
Accelerating diagnosis of metastasised cancer
In a 48-year-old woman who presented to Mount Novena with lower back pain, MRI showed a single destructive lesion in the lumbar spine. A biopsy was planned.
On chest X-ray, the Harrison.ai solution picked up a possible right lung nodule which was hidden by the right lung hilum, which was still equivocal to the reporting radiologist in hindsight.
However, due to this AI finding, a PET CT was performed, clinching the diagnosis of a lung tumour metastasised to the spine.
This shortened the overall workup tremendously into a single two-day admission and enabled faster treatment.
Key lessons from Parkway Radiology’s AI implementation
1. Look for a technology partner who listens and adapts to your needs
“Finding the right AI solution partner is very important because this is not a one time IT deployment – it’s a long-term partnership,” Ms Tan says. “The Harrison team was very supportive from the start, and they listened to our needs and addressed them in a very responsive and collaborative manner.”
2. Engagement is key to successful deployment and acceptance of new solutions
“During the process of deployment, we heavily engaged all stakeholders – for example our radiologists, our radiographers, our hospital IT staff and even management,” Dr Tham says.
3. Plan your approach
“We decided to roll it out in phases to allow our users time to learn and adapt. Doing that has enabled us to have a very successful deployment,” Ms Tan says.
4. Don’t settle for the first solution
Parkway Radiology tried various solutions before choosing Harrison.ai. Do a thorough evaluation and document your learnings. And remember different institutions have different needs, so there’s no one-size-fits-all answer.
Looking ahead
Parkway Radiology is excited for things to come, Ms Tan says.
Their next step will involve seeking solutions to support complex sub-specialised reporting, and AI that improves the technologists’ workflow.