The interview of Jarkko Niemelä, CEO, and Co-founder of MVision AI, for IntelligentHealth.tech magazine. October 2023.
What specific challenges does Contour+ address in the field of radiotherapy, and how does its AI-powered automation contribute to overcoming these challenges?
The largest challenge in cancer radiotherapy is the shortage of experts, including oncologists. For this reason they have too little time to create a thorough treatment plan for each individual patient, which translates into two challenges. First, is the increase in the patient waiting times and therefore increasing patient backlog, and second, is the non-optimal quality of work. Contour+ addresses both of these challenges. First challenge is addressed by the significant time-savings obtained, and the second challenge is addressed by a data-centric AI development process, whereby the AI guides the experts towards best practice industry standards.
How does Contour+ leverage AI technology to automate organ-at-risk and lymph node volume delineation in radiotherapy planning?
We use and train deep learning algorithms that are very powerful in image processing. The key element is to select the most appropriate model architecture for the task, combined with the data-centric approach. Every AI algorithm is only as good as the data it is given on which to learn from. At MVision AI, we have built a comprehensive pipeline to generate training data that complies with the industry standards pertaining to the approved contouring guidelines.
Could you elaborate on the standardization aspect of Contour+ and how it aligns with industry contouring guidelines such as ESTRO, EPTN, RTOG, and the UK SABR Consortium?
Contour+ has been developed to follow and comply with the industry standards which are the contouring guidelines. This is not possible with clinical datasets obtained from clinics due to the inherent human or expert variation contained within them. With MVision’s combined data-centric and peer-review approach by Oncologists, we enforce our AI to be guideline compliant.
Contour+ generates guideline compliant contours for all patients including the mentioned ESTRO, EPTN, RTOG, and the UK SABR Consortium. Consequently, we believe, our approach will enhance the quality of future conclusions relating to best practice due to the variation from these standards being minimized. Development of any AI based tools carries a responsibility for the companies who develop them.
The increased awareness of AI in the media is driving Governments around the World to form governing bodies and/or departments who will be overseeing this industry, such as the EU AI Act. From our perspective, considering what we are developing and who we are developing for, there has to be a standardization which acts as the ground truth – and we believe this to be the Guidelines.
What specific time-saving benefits does Contour+ offer in terms of 3D organ model generation, and how does this impact radiotherapy workflow efficiency?
Contour+ automates the vast majority of the contouring tasks undertaken when creating a treatment plan, translating into a time saving of up to 95%. It’s true that the mean time savings will vary between different anatomies. An example of these time savings can be clearly seen when delineating a head and neck case, which consists of not only the anatomy, but also the significant lymph node volumes. Typically our customers are seeing a saving of up to 2 hours of time for this kind of patient.
These time savings not only deliver benefits for the clinical teams performing these tasks, but they also aid in operational efficiencies for resource constrained departments. These efficiencies can result in patient waiting times being reduced to the start of their treatment, as has been witnessed by our customers (from 9 days to 6 days). What is really exciting is that one of our customers has leveraged these efficiencies to offer their patients same day treatments.
What are the key advantages of Contour+ as a cloud-based AI solution for radiotherapy, and how does it contribute to maintaining high-quality treatment standards?
Contour+ is extremely user friendly; there are no new user interfaces needed. The cloud setup is complimented with our automated configuration that sends CT or MRI images that are anonymized to our cloud service. Contours are applied and the structure sets are automatically sent to the storage location of choice ready for importation into their treatment planning system. Here the Dr’s will review the returned contours and delineate the target volume. Job done!
There is no need for powerful computers to be installed within the hospital, which saves on cost. When we release a new version of the AI models, everyone who is on our Cloud service immediately has access to them. It really makes life very simple for our customers. It has to be said, some customers, due to their IT compliance policies, are not permitted to use Cloud services in general. That’s OK. We have the ability to provide local hardware installations of the system as well as access via a Web browser. There really are no limitations for any customer who wishes to use our solution offering.
Could you share insights into the clinical impact of Contour+ in reducing inter-observer variation and promoting standardized clinical practice?
As shown in studies with Contour+, the inter-observer, expert oncologist variation is dramatically reduced on average by 20%. Put the other way around, consistency increases by 20%. This includes both organ-at-risk and treated target volumes with lymph nodes. It’s important to highlight that the consistency witnessed is not only as a result of our AI models, but also how we inadvertently maintain the clinical teams expertise and knowledge as it relates to the industry standard contouring Guidelines.
Can you explain the Return On Investment (ROI) potential associated with Contour+ and how it translates into both revenue and resource optimization?
There is tremendous ROI potential for any and all radiotherapy departments when deploying Contour+ within their service. As an example, in the UK we are processing 2750 patient scans across 27 NHS Trusts in each month. This is about a quarter of all UK radiotherapy cancer patients. This yields an estimated 7 working-month returns on investment each month based on time saving! From an economic perspective, time is money! But more importantly in our industry, time is a commodity that is precious and can be spent to address many common issues such as lack of resources, ability to treat more patients, and thereby reduce waiting lists etc.
How does Contour+ ensure GDPR compliance and seamless integration with various hospital systems, including cloud-based solutions?
Our cloud-based solution has been developed with the Safety-by-Design principle. This basically means that all data is de-identified by our Daemon software, which is installed on a low powered server or a virtual machine in the hospital, before being sent to our Cloud for processing. All data transfer, and data at-rest, are encrypted. This data is then permanently deleted within 24 hours of its receipt. This means that in any given moment there is no large database of de-identified images in our maintained cloud-servers to be breached.
Looking ahead, what are your expectations for the future of AI in radiotherapy, and how does MVision AI plan to further enhance its offerings in this field?
It is obvious that AI has the potential to dramatically increase the efficiency and quality of radiotherapy. We will see more and more AI automation to undertake repetitive tasks thereby enabling clinicians to do what they want, and need to do, and that is to focus on their patients. This does not mean we will no longer need the experts, on the contrary! Actually, it is imperative that we train our clinicians to be competent in their ability to firstly understand, and secondly, how to monitor the AI generated results. Similar to that of airline pilots who still have to undertake extensive training to fly planes, despite planes having an autopilot function! I believe this is the direction of travel for radiotherapy as well. As a consequence, MVision is offering new training tools with new products Guide and Verify to assist with this evolution of AI contouring tools. It is our responsibility to enable the experts to use AI safely and responsibly, while AI manufacturers are responsible for the safe and transparent development of AI.