Artificial Intelligence (AI) offers numerous opportunities and potential in the healthcare sector. It is therefore not surprising that numerous Swiss digital health startups are also applying this technology in their products and services. One of these startups is b-rayZ, which we introduce to you in this blog post.
Author: Jonas Probst
AI has emerged as a transformative force in healthcare, revolutionizing the field of diagnostics. With its ability to process vast amounts of data and identify complex patterns, AI algorithms are enhancing the accuracy and speed of medical diagnoses. From medical imaging analysis to genomic sequencing, AI-powered systems are assisting healthcare professionals in detecting diseases, predicting outcomes, and tailoring personalized treatment plans. By leveraging machine learning and deep learning techniques, AI algorithms continuously learn and improve, leading to more precise diagnoses and better patient outcomes. The integration of AI in healthcare diagnostics holds immense promise for early detection, improved efficiency, and ultimately, saving lives.
The Swiss digital health ecosystem is actively engaged in the realm of AI-driven diagnostics as well. Enter b-rayZ, a Schlieren-based (ZH) digital health startup that leverages AI-assisted mammography to enhance breast cancer diagnosis. This innovative solution supports not only radiographers but also various stakeholders throughout the entire imaging process, including examination, as well as the reading and reporting stages. By promptly assessing the quality of each mammographic image right after acquisition, b-rayZ can offer personalized recommendations for optimizing patient positioning, thereby improving image quality. Moreover, the platform automates the classification of breast density, enabling informed decisions regarding the need for further testing. In this manner, a more precise and reliable diagnosis is achieved compared to traditional assessment methods.
But what is the story behind this company? Why is quality in mammography such a pressing issue and what does it take to successfully scale such a solution in a rather conservative environment? To shed light on these and many other topics, I had the pleasure to talk with Prof. Dr. Cristina Rossi, the CEO and co-founder of b-rayZ.
Hi Cristina, b-rayZ has been officially founded in 2019: What is the story behind the company? When did you meet and thought you want to enhance mammography as it is today?
Officially the company was incorporated in 2019. But of course, there were some signs of potential commercial success already before that. I think the founding team came together in 2017. At this time I was working as a data analyst in magnetic resonance imaging in the radiology department at the University Hospital of Zurich – so not really a core mammography or breast imaging field. At this point my two co-founders and I formed the first steps of the company: My husband Andreas Boss specialized in breast imaging as a radiologist and leader of the breast unit at USZ and Alexander Ciritsis, our CTO, was also working as a data scientist in the hospital.
Before the new possibilities of computer vision and deep learning, we were always attempting to do a model-driven analysis of acquired images. This means we assumed that there is a model that describes the images of a pathology or disease, applied this model to the acquired data, and then tried to make a prediction. But actually, this model-driven approach has a big limitation. First of all, the model needs to be robust and reliable which is not always the case. Secondly, people working in radiology do not look at the images as detailed and reliable with a model in mind as a computer would do.
New AI-driven technologies are able to look for patterns in the images and recognize degrees of grey, shadowing, and shapes and for this reason, we thought “Okay, let’s see, how this technology performs!”. And already with a very preliminary prototype, we found out that actually, if you ask the right question, the technology was quite robust in the answer. So we thought, we know the questions that the people need to answer in the real clinical routine. We have this technology in the end, then let’s make this big step and bring this technology together with the answer to the right question into a global market.
Why did you focus on breast imaging?
To start the commercialization of this product we focused on breast imaging because breast imaging was not only a pain point in our institution at that time, but is also an area of high volume of examinations on a global level. However, there is a lack of specialized radiologists and the screening programs are maybe the highest regulated area in radiology with a lot of administrative burden. So we thought, there is a huge potential for our AI to not only take over the clinical task but to also support non-clinical tasks which still require the understanding of images. Furthermore, for me personally, the topic is a very important one since I felt an intrinsic need and drive to tackle this threatening disease and save women’s lives with the leveraging power of our idea.
And this is how it started and now we are commercializing the product in Switzerland, where we are quite present today. We have several big customers in Switzerland, but we are also active outside of Switzerland and scaling our activities in the sweet spots of the European market.
Can you give me a number of hospitals in which your solution is already in use?
We have many hospital chains as customers. Currently, we have more than 30 customers and today we are particularly proud of having started our first activities with the screening program in Basel. There are six centres that are involved in the screenings and we think that our solution is of especially high value for institutions that collaborate closely and need to have standardized procedures. Our customers need to have a high quality of performance because they get monitored and wish to exceed in the market by showing that they offer the highest quality in the screening results.
Would you mind elaborating on the product and its functioning? Could you explain how it operates and outline the necessary steps for customers to implement it?
The solution reaches the user exactly where and when he is working. So we offer a digital platform that covers the whole journey of the breast cancer patient from the early diagnosis over biopsy to follow-up surgery planning and so on. The users are healthcare professionals with different profiles which we all consider in our platform. So it can be the radiologist that wants to have a second opinion in reading the images, the radiographer that wants to have a quality check on the image she just acquired or it can be the person that is responsible for the breast unit in general. So this could be a manager, who needs to create reports for the authorities, for the accreditation or for an audit. We reach the different users through dedicated channels and this is really a strong point of our solution because in some institutions they screen three different patients in just 10 minutes. So they cannot stop the whole process just to use a solution – the results need to reach them where they are.
When it comes to the implementation, the workflow of the radiologists doesn’t get disturbed in any way from our integration. I think radiology is a special area of medicine where the integration can be quite standardized and in which you really need to have a plug and play solution – something that doesn’t stop the flow of information in the hospital. If you start requiring the use of different interfaces, it gets complicated for the customer and it gets difficult to be scaled. Every hospital has a different patchwork of IT solutions and for this reason, the engineering of the installation and the integration is a key part of the product.
In terms of technical implementation, could you shed some light on how the solution is integrated? Do you collaborate with other service providers within the hospital industry to ensure such a smooth implementation?
Of course, the solution needs some interfaces between key points: Of course, it needs a connection with the device, which is doing the acquisition, and an interface to the PAX system. So this kind of integration is necessary, but it is a standalone solution, a plug-and-play system. We are independent from the mammography devices and from the IT of the clients: It works with all mammography vendors and all IT settings at the customers’ site. So it’s very plastic.
Commercially, it can of course be integrated into other IT solutions, if we believe that in some regions this may accelerate the penetration of the market. These are definitely aspects that we take into account.
Currently, do you have any existing partnerships or collaborative arrangements with other vendors for commercial purposes?
At the moment we see a trend in radiology to offer whole AI platforms. These are sort of an app store for AI solutions in radiology in which the customers can easily buy for example b-rayZ for breast cancer imaging. As a customer, I then don’t need to care about the integration of many different solutions. Some of those platform providers are more successful than others. So from our side, there is a screening of the possible partners that fit best to us and our philosophy.
It’s intriguing to hear about the advantages your solution offers in both clinical and non-patient-related tasks, including administration. Could you provide any specific figures or data regarding the impact it has on quality improvement and cost reduction?
Let’s start with quality assurance, which is not always an attractive topic when you approach it – it’s hard but work that needs to be done. Here our solution is checking if the positioning of the breast into the device was done properly during the acquisition process. Mistakes in the acquisition process may lead to incomplete images and therefore undetected cancer but false positioning may also create artefacts. These are parts of the images, which look like cancer but result from a mistake. So you may not only recall the patient and cause a lot of stress for her but you may also request additional unnecessary diagnostic which leads to additional costs. So the recalling due to technical mistakes really is a large problem for many institutions. Especially in the screening program, it is well known that if women get recalled because of a mistake in the acquisition of the images, they lose trust in the screening program as a whole and drop out of it.
At the time at which the images have been acquired, we immediately check the result and flag technical mistakes within seconds. Then the radiograph knows she needs to retake this image, when the patient is still there. This dramatically reduces the number of recalls to something between 5% to 8%.
Additionally, we also immediately give a warning to the radiographer if the woman needs additional diagnostics. There is one very important aspect in mammography, which is the density of the glandular tissue in the breast. Some women have a very dense breast, which means that the risk for breast cancer is higher and the mammography less sensitive. So on the image the glandular tissue looks white but the breast cancer is also white. In this case the patient needs additional tests, for example through additional ultrasound or MRI scans.
We immediately give a warning to the radiographer and say, look, this woman needs for example an ultrasound. Before the introduction of our software, there was always a consultation between the radiographer and the radiologist which of course created waiting time for the patient because the radiologist is not always there. In one study we performed in Switzerland, we saw that the waiting time of the patient was shortened from 25 minutes to six minutes through our technology.
Based on what you mentioned earlier, it appears that your solution offers numerous advantages, with the issue of false positives being particularly significant. However, in your experience, what would you say is the primary pain point of your customers that your solution addresses?
I think it’s the concept of having a solution that is creating a digital workspace that helps them a lot. Even if, for example, an institution wants to improve its competencies in reading mammography, undertaking the investment for a software that only solves one very specific problem is unfavorable. But if you have a platform which is highly specialized for the whole workflow in breast imaging, you cover different needs.
And I have to say that the main problem is really the lack of specialized radiographers in mammography. There are no radiographers or they are aged above 50/55. The institutions need to have the new people coming in faster, ready to reach high-quality performances. The use of our software is accelerating the education and specialization of radiographers in a short time and this is a huge advantage for them.
How does your solution impact the roles of radiographers and physicians in this field?
I would say it’s making the radiographer more independent from the radiologist which is good because their work is not just executing the instruction of the radiologist, but it’s also growing in the responsibility of the handling of the patient for the time of the image acquisition.
I think what is also nice and exciting in breast imaging is that it is still very clinical. It’s very close to the patient and you have a lot of contact. I think through our technology, we give the specialists more time for those tasks in which they cannot be replaced. We get rid of all this administrative work and we give real-time instructions that make people more independent. We give the possibility to split the work among the people of the team according to their education or to the time capacities they have for the daily work. And this is really bringing a lot of value to the customers.
While your solution appears promising from the perspective of technicians and physicians, are there any concerns or reservations that you encounter from them regarding its implementation or usage? Could you elaborate on any potential challenges or hesitations that they may have expressed?
Among the breast imaging specialists, there are those who are sceptical but this is really a minority. I have to say that they often have a misconception of the integration of AI in the workflow and miss a clear idea, but this is a real minority. Actually, they are calling for new technology.
For some radiographers, the situation was different at the beginning. Generally the radiographers are not the target users of AI solutions, as the first solutions have been really developed for radiologists or for doctors. So the doctors have been educated, but the radiographers were left a bit alone. And when you come with an AI also for them, they may be a little bit confused at first but I think the solution and its value is so easy to grasp that we didn’t get negative feedback so far.
So change resistance is not that main barrier for the implementation in the hospital?
No, absolutely not. A limitation is the fact that the budgeting of the solution is often difficult because the hospitals don’t always have a pool of money dedicated to digitalization. Many of the solutions are not reimbursed by the health insurances, so it’s missing a bit of an incentive, which would make the budgeting of the new solutions much easier. I think this issue especially concerns the health insurances because they may shape a completely new market, but not only because they feed a new industry, but also because they would save a lot of money by standardizing the decision making in radiology. I really don’t understand why these solutions are still not in the catalogue of tools for which the doctor can get a reimbursement.
To broaden the scope a little bit: From an outside perspective it sometimes seems as if there is some sort of gap between the many new and innovative solutions in the health space and the large-scale application of those in the day to day treatment of patients. Would you agree with this?
If you develop a product just because you think you have the technological power, don’t do it. You don’t need to solve a problem that no one has. This is a golden rule independent of the technology you are using. But if you have a good idea to solve a problem, which is a burning issue in the clinics, then the story is different.
Healthcare or MedTech is a conservative industry in which you are bringing a new product generation and you need to be aware that this industry is resistant to change. You need to take account of this status quo, so you cannot think that a commercial model that works very well for iPhones, will work in the same way in medicine – it’s just not true. I give you an example: At the beginning we were talking with a lot of business people and they told us “Oh, why don’t you charge per use? Why don’t you charge per patient?” – I understand that this model works for Netflix or when you want to buy a movie, but it doesn’t work for the medical industry. They have such a burden in budgeting the solutions, so they want to have that done once and in the best case not to think about that for the next two or three years. Your business model, your go-to-market strategy and your pricing need to be adapted to an industry with existing rules. This may be also a reason why a good technology that maybe also addresses meaningful problems is not commercially successful. A part of this business idea is not adapted to the acceptability for the market.
Do you observe differences regarding those unique characteristics of the industry among various markets? Can you provide insights into how these variations impact your business operations?
Absolutely, they’re very different. Take for example Switzerland and Germany: Even though they mostly speak the same language, they are completely different. Switzerland still has quite a large component of private insurances while Germany is mostly dominated by a public healthcare system. This has a huge impact on the way healthcare is managed and on the pricing and time saving performances, which are expected from solutions. Another interesting aspect is how regulated the activities are. Germany has one of the strongest requirements in Europe regarding the quality assurance in mammography screening programs. But in another country, they may not have any quality assurance. So there are still many regional differences.
Looking ahead five years into the future, what would you identify as the primary trends within the industry that present both opportunities and challenges for your business?
That’s a good question. I think for sure the growing maturity of the markets is offering a huge opportunity. Population is ageing and breast cancer awareness is constantly growing, while there is still no constant growth in healthcare professionals. So these are points that for sure push the solution and create new opportunities for us.
I’m a bit critical in looking at the previously mentioned AI platforms. Not because I think that they are in principle a bad idea, but because I see that some of them think that the work is done when the platform is there. I think that this is a big mistake. Those platforms are now fulfilling a first need of the customers, but they need to be able to grow over time. For this reason, I’m also critical in the selection of the platform partners. I would like to have a platform partner or multiple platform partners that see this as a journey, as a product that grows in maturity along the time.
To wrap up, I’d like to inquire about your vision for b-rayZ. What are your goals and where do you see the company heading in the future?
The vision is to have this complete end-to-end platform. We want to go deep into the challenges and needs of breast diagnostics. Now we have the focus on diagnostic and radiology, but with the possibility of covering the neighbouring fields as well. Commercially, our goal is of course to reach the main global markets Europe, North America and Asia. So I think this will be the goal for the company in the next years but I also have some other dreams in the drawers, which I prefer not to share yet.
So they remain a secret for now! Thank you very much for the interview and the interesting insights.
0 Comments