AI Axial3D

Axial3D

Nominated Award:
Best Application of AI in Healthcare and
Best Application of AI in an SME

Website of Company:
https://www.axial3d.com/

 

Axial3D is an innovative medical imaging technology company, based in Belfast, Northern Ireland. The company was founded in 2016 and has since grown to a team of 46 employees. We create patient-specific 3D visuals and printed models of anatomy based on 2D medical scans.

Our vision is to make patient-specific care routine – these 3D anatomical visuals and printed models accomplish precisely that by making surgical planning and treatment bespoke to each patient’s case, transforming outcomes for surgeries worldwide. It’s our mission to make these patient-specific 3D visuals and printed models easily accessible and affordable to physicians globally.

To achieve this mission, Axial3D develops artificial intelligence-based software solutions that automatically convert hard-to-understand 2D medical scans into incredibly precise, patient-specific 3D models at scale and at a price point that is accessible for all using a process called segmentation. Our cloud-based Segmentation-as-a-Service portal enables surgeons or hospitals to upload their patient’s 2D medical scan data where it is then automatically converted into a 3D digital model or printable file using proprietary machine learning algorithms.

We offer the creation of 3D printable files within minutes that can be output in a hospital’s 3D print lab or full production and shipment of a 3D-printed model within 48 hours. We are currently the only company in the world capable of this turnaround time, which means that our offerings can be used for time-sensitive surgical cases.

The insights provided by our medical 3D models help surgeons with pre-operative planning, which saves surgery time and costs and ultimately improves clinical outcomes as a result. Surgeons also use the 3D models to brief surgical teams and improve the patient consent process.

Reason for Nomination:

We believe that the future of surgery is patient specific. Patient-specific surgical planning drastically improves patient outcomes and reduces healthcare costs. In order to achieve this tailored way of treatment planning, clinicians need  access to a three dimensional visualisation of the patient’s anatomy, something that’s not possible with conventional 2D imaging, such as CT scans.

2D images alone often do not fully convey the necessary precision and insight for physicians to confidently plan a surgical procedure on 3D anatomy. This can lead to a lack of clarity among surgical teams and longer operating room times, resulting in higher costs to the healthcare provider and heightened risk for the  patient. Conversely, 3D visualisations and 3D-printed models of patient anatomy help physicians to more quickly and clearly understand the patient’s condition before surgery while supporting enhanced surgical team and patient communication. Ultimately, 3D anatomical modelling helps physicians to make faster decisions with much greater confidence to help them perform safer, and often faster and cheaper surgical procedures with reduced risk of complications.

In order to create a 3D visual or 3D-printed model of a patient’s anatomy, their 2D medical scan needs to be converted into a 3D file, a process known as segmentation. This process is currently performed manually by skilled engineers using specialty software, with the 3D images taking hours to produce. This is labour intensive, costly, and doesn’t scale, and restricts the types of surgeries that the 3D visuals and printed models can be used for. To overcome these limitations, Axial3D has been developing innovative supervised machine learning algorithms to be able to perform this process automatically at scale and at a price point that is accessible for all. Clinicians can easily upload a 2D medical scan to our online platform. These scans are then automatically segmented by the software which is then validated by our medical visualisation engineers. The creation of 3D visuals and printable files takes only minutes. The printable files can be output in a hospital’s 3D print lab or can be printed on site at Axial3D. Full production and delivery of a physical 3D-printed model is fulfilled within 48 hours. We are currently the only company in the world capable of these turnaround times.

Our medical 3D visualisations and printed models have transformed the pre-operative planning process. Surgeons are able to confirm their plan and even practise the approach before the surgery using the 3D-printed model. As a result, the time spent in the operating room can be reduced significantly, which shortens the patient’s recovery time and benefits the overall wellbeing of the patient. Additionally, the surgical team is able to pre-select the appropriate equipment, which saves time and costs sterilising unnecessary equipment.

Our mission is to make medical 3D models affordable and accessible for everyone. There have been some cases in which our models have had life-changing results, such as baby Pippa’s case. Pippa was suffering from a complex malformation of her heart and had to undergo life-saving surgery. To reduce the risk of complications, the surgical team ordered a 3D-printed model that was an exact replica of Pippa’s heart, enabling them to plan the procedure more precisely, tailoring it specifically to Pippa’s complex case. Pippa was finally able to go home after spending the first six months of her life in hospital.

Having access to a tangible model not only benefits the surgeon, but also the patient. It can be difficult to understand what a surgical procedure entails and the associated risks. Using the 3D model enables the surgeon to show and explain in great detail what the procedure will look like and how this is necessary for the patient’s treatment. This improves the patient’s confidence going into surgery, and it enables informed consent. This can be especially useful when parents have to make decisions about their children. It allows them to make these decisions with more confidence, and trust that the procedure is the best option for their child.

Lastly, it’s no secret that COVID-19 has wreaked havoc on surgical capacity in many hospitals across the country, leading to the cancellation or deferral of millions of elective procedures and resulting in a significant backlog of surgeries. It has left hospital systems across the world wondering how they can quickly and safely increase operating room throughput to reverse the backlog. Thanks to the additional insights our patient-specific 3D modelling provides, surgeons are able to better optimise the time in the operating room, which enables them to improve operating throughput and help reduce surgical backlogs.

Additional Information:
A recent ground-breaking study, undertaken by the University of Ulster showed that use of our 3D-printed models for surgical planning fundamentally reduced surgical time and cost, and improved the resulting clinical outcomes.

Between July 2020 and July 2021, Axial3D provided 72 patient-specific 3D-printed models to 33 surgeons from orthopaedic, cardiothoracic, maxillofacial and neurosurgical specialties. The surgeons were asked to assess the impact the models had on surgical planning, communication, cost and patient safety.

The convincing findings clearly show a wide range of benefits to surgeons, hospitals and patients:

• 42.7% of responses reported a reduction in pre-operative planning time.
• 92.6% of surgeons said the model helped surgeon-to-surgeon communication.
• 91% of surgeons said the model helped trainee communication and education.
• 59.1% of surgeons said the 3D model positively affected patient consent.
• Surgical time was reduced in 41.5% of cases.
• 86.6% of responses agreed that the model improved surgical outcomes.
• 92% of surgeons agreed that 3D models were a better method for surgical planning than traditional 2D methods alone.
• 94% of surgeons agreed that 3D models either saved money or were cost neutral.
• 52.2% of responses agreed that the model provided a cost saving.

The surgeon testimonials below show that our models met the needs of our customers and improve the patient journey:

“[The 3D model] was incredibly helpful, when we were viewing the CT the model had been reconstructed from the ventricular side so everything was a mirror of what it should be. The model instantly highlighted our mistake and helped us plan the operation better highlighting areas that we need to be aware of and actually how close the structures are when the valve is being inserted”
– Cardiac surgeon, Belfast Health and Social Care Trust

“The quality of the model is far superior than that of leading brands in the industry. Dedicated approachable and quite helpful team. Innovative thinking and always available to explore new requests.”                                                                                                      – Craniomaxillofacial surgeon, South Eastern Health and Social Care Trust

“The model was extremely useful in pre-op planning amongst the surgical team and explaining the options to the patient and her family. It allowed us to plan the bone tumour excision and reconstruction using a periarticular locking plate into the proximal tibial epiphysis rather than an Ilizarov frame and fibular strut graft harvested from the non-involved limb. This reduced the operative time, patient morbidity and hospital stay and allowed for a quicker rehabilitation for the young lady. We would definitely use 3D modelling for future complex bone or soft tissue tumour cases. Many thanks.”
– Oncologist, Belfast Health and Social Care Trust

Axial3D project 2021

Nominated Company: Axial3D

Nominated Award: Best Use of AI in Sector

Medicine is moving rapidly towards a patient-specific model in which surgical planning, diagnosis and treatment is tailored to the unique characteristics of each individual patient. Such a model drastically improves patients’ medical outcomes and reduces healthcare costs. Axial3D is an innovative medical imaging technology company, based in Belfast, Northern Ireland with a mission to transform the accuracy and speed of patient-specific surgical planning, diagnosis and treatment. Our vision is to make patient-specific surgery routine and accessible for all patients.

We started out 3D printing patient-specific anatomical models to improve surgical planning following research carried out by our founder, Daniel Crawford at the University of Glasgow Medical School. In 2016, this research enabled our company to be born out of skills attained in turning complex 2D medical scans manually into patient-specific anatomical models that could be 3D-printed.

After our initial seed funding round, we were able to hire biomedical engineers and start to service the public healthcare networks in the UK and Ireland (NHS & HSE). It was quickly apparent that there was a high demand for the 3D-printed models given their clinical use case and ability to transform surgical outcomes. When assessing how to scale into the wider healthcare network locally as well as our next target market of the US it was evident that we would be unable to address the market through using manual and labour intensive processing to convert the 2D medical scans into the 3D models. It was clear that automation of this process was required to create these models at scale. A key step in the creation of the 3D models from the 2D medical scans is a process known as segmentation, which identifies and labels the anatomical features within the scans. This is a highly repetitive image processing step, which lends itself well to being automated by machine learning.

Over the course of the last 5 years, we have been developing machine learning algorithms to perform this process automatically, allowing us to scale our production of 3D models and transform the accuracy and speed of patient-specific surgical planning, diagnosis and treatment globally. These algorithms drive our cloud-based software-as-a-service solutions in which clinicians can easily upload a 2D medical scan to our web portal and have a patient-specific 3D model available to them within minutes.

Reason for Nomination:

Patient-specific surgical planning, diagnosis and treatment can only be realised through giving clinicians access to a precise visual of a patient’s anatomy. The current best
practise is to use 2D images of a patient’s anatomy (CT, MRI and X-ray scans). These images are inherently difficult to understand and highly limited in accuracy due to a human’s multi-layered three-dimensional arrangement. This makes surgical planning, diagnosis and treatment challenging, particularly in complex cases, leading to misdiagnosis, surgeries being misplanned and patients spending millions of unnecessary hours in surgery.

Clinicians can move towards providing patient-specific surgery with access to 3D digital images or printed models that are an exact replica of a patient’s anatomy. These are derived from the routinely taken 2D medical images. There are as many as 10 million complex surgeries worldwide that could be transformed using patient -specific 3D images and models.

However, transforming a 2D medical image into a 3D image that can be used as is or 3D printed involves a crucial step known as segmentation in which the anatomical features on the scans are identified and labelled. This is typically performed manually by skilled engineers using software. With currently available software, 3D images take hours to produce. This process is labour intensive, costly, and doesn’t scale, restricting the types of surgeries that the images and models can be used for.

Axial3D addresses these limitations, enabling every surgery to be patient-specific. We do this through developing artificial intelligence (AI)-based software solutions that use supervised machine learning to automate this segmentation process, enabling hard-to-understand 2D medical scans of patient anatomy to be turned into incredibly precise, patient-specific 3D models, within minutes, at scale and at a price point that is accessible for all. Our solutions are cloud-based so that clinicians can easily upload a patient’s scan to our web portal and have the 3D model delivered to them without any need for installation of on-premise software within the hospital.

These solutions enable clinical teams to more accurately and quickly plan, diagnose and treat a patient in a tailored manner, drastically improving their surgical outcome. When used in surgical planning particularly, clinicians report that use of our 3D models provide them with new and crucial insights about their patient’s condition that changes their surgery plan for the better in 53% of cases, reduces surgical time by 62 minutes on average, and reduces patient recovery time in hospital by 16%. This equates to a reduction in the cost of each surgery of €2000 on average.

Additional Information:

Our main competitors are Materialise and 3D Systems, which enable an expert user to convert a 2D medical scan into a 3D model in a matter of 4-5 hours, one at a time. However, their software solutions are not powered by machine learning, which means that the segmentation step has to be performed in a manual fashion by skilled personnel. This lack of automation means that the production of 3D models by these companies typically takes weeks to months (due to order backlogs even at a modest scale) and each one costs on the order of £1000s. Furthermore, they are not scalable either from a capacity or a cost point of view, and are not cloud-based, which means that the (licenced) software has to be deployed on-site at hospitals. Approval of new on-site software by hospitals takes time and can be costly.

We differentiate ourselves by producing thousands of models in parallel, with the segmentation process conducted automatically by our machine learning-based software in minutes not hours. This allows 3D models to be delivered to clinicians much more quickly (within hours for digital files to 2-3 days for a print file) and at a much lower cost on the order of €100s not €1,000s per model. Furthermore, the requirement for specialist staffing is removed. In essence, we move this capability from being low scale, costly and highly limited by available expert staff capacity, to massively scalable, highly cost effective and with no staffing (or other resource) constraints.