In the previous blog post you read how deep learning is the backbone of BoneMRI and how its machine learning algorithm is trained with data, data and more data.

Deep learning is the technical side of BoneMRI, but what about the added value we bring to our patients? How do we demonstrate the performance of our product? And how do we assure it is safe to use? The answer is clinical evidence.

How is clinical data and evidence collected? The magical answer is: clinical investigations! The purpose of clinical investigations is to collect imaging data used for training. But it has a bigger purpose: the generation of clinical evidence. In this blog post you will learn about the need for clinical evaluation and evidence, how it all begins with data which is later used to support market releases.

Let’s be real, AI is taking over the world! We’re still not sure what the future holds and how AI will be used in the medical world, but one thing is for sure: medical devices based on AI need to be evaluated for their safety and effectiveness before they can be unleashed to the public.

Clinical evaluation and evidence

Let’s kick off with the most important, but also stuffy, part of medical devices. The Medical Device Regulation (MDR) are strict guidelines applied in the European Union that ensure the safety and performance of medical devices. The MDR outlines the clinical evaluation required for all medical devices before they can be placed on the market. This evaluation uses clinical evidence to assess the safety and performance of a device, including information from clinical investigations, post-market surveillance, and scientific literature. Clinical evidence is the way to demonstrate that a device meets the essential requirements of the MDR.

Clinical evidence and investigations require filling in all sorts of paperwork, dealing with ethical and government regulations, and working out logistics with people from different backgrounds and education levels such as study nurses, investigators, doctors, MRI lab technicians, etc. This might not sound like the most thrilling job in the world, but the final outcome is totally worth it because clinical evaluations are actually pretty important and valuable. Think about it this way: would you want to use a medical device that hasn’t been properly tested? No, you wouldn’t. That’s like driving a car with no brakes or using a toaster in the bathtub. It’s just not a good idea. Clinical evidence helps ensure medical devices are safe and effective to use.

Now, the million dollar question is “how do we collect clinical evidence?” With BoneMRI we perform clinical investigations to gather clinical data and use it as clinical evidence. The clinical investigations we set up often have two purposes, namely the collection of data for training our magical BoneMRI algorithm and the generation of clinical data that supports the intended use of BoneMRI.

First, we will dive into the collection of clinical data for the algorithm’s training. Afterwards, we will explain the most important and fun part: the data collection process to prove the performance and safety of BoneMRI.

Data is where it all begins

Let’s start at the beginning. BoneMRI is a software solution that transforms an MRI image into a CT image, based on AI. In order to do so, data is required from both imaging modalities. We need an MRI scan and a CT scan of the same patient, but we can not just start collecting data from patients that get both scans. Why not? Because the MRI image used in training has a specific type of contrast. If you want to learn more about this read our previous blog post BoneMRI explained: How deep learning is the backbone of our product

Soft tissue, BoneMRI and CT images

Due to this specific set of parameters, we have to proactively work with hospitals so that this MRI sequence is added into the MRI protocol of the patient. Since data is being used for development purposes, this is considered research. Therefore, clinical investigations are arranged to collect data for the development of BoneMRI. This is where the Clinical Team of MRIguidance gets involved. We are responsible for designing, setting up and monitoring these clinical investigations.

The fun part: clinical investigations

In clinical investigations we collect MRI and CT images of the same patient. This data is used to train our software model, and once the model is fully trained we use the MRI image to generate a BoneMRI and the CT of the same patient to validate the performance of BoneMRI. Since we want BoneMRI to be the smartest model, it should be exposed to as many data variations as possible. This means we want to collect data from different MRI vendors, magnetic field strengths, anatomies and patient populations. In order to do this, we work together with many hospitals.

The biggest clinical investigation that we are currently running is called Clever Bones! Why that name? Because we want BoneMRI to be clever! (Also because within the clinical studies world there’s a fun challenge to come up with the best acronyms.) Clever Bones stands for CLinical Evaluation and VERification of BONEMRI in the Spine. The ‘S’ is not for multiple boneS, the ‘S’ indicates the anatomical region we are focussing on within this study: the spine.

Clever Bones is a multicenter clinical investigation in which 9 hospitals all over Europe participate. Patients suspected of spine disorders that receive an MRI and CT as part of their clinical care are asked to join the study. As part of the study procedure, the patient just needs to lie still for another 5 minutes in the MRI scanner to run our boneMRI sequence. This is enough to help us improve our product.

Together with the investigators we validate BoneMRI for different endpoints. How these endpoints are validated is based on the study design and they are defined to establish and verify the clinical benefits, safety and performance of the device. An example is the accuracy in terms of geometrical morphology; the accuracy of BoneMRI for different patient populations like oncology or trauma patients; how BoneMRI can be used for the planning and navigation in spine surgery or the impact of BoneMRI on the clinical workflow.

The investigation of endpoints can be performed in different ways, with qualitative and quantitative methods. Quantitative aspects on the accuracy are validated internally using in-house developed tooling focussing on geometrical morphology and voxelwise Hounsfield Unit accuracy. Other quantitative methods can be the scoring of the diagnostic performance on detecting pathologies including erosions, sclerosis and ankylosis. Qualitative aspects on clinical usability are validated in more subjective ways, for example by using questionnaires that are filled in by the end users. In order to ensure we validate our product in an adequate way, we work with investigators that could be potential end users, such as radiologists and orthopaedic/neuro- surgeons. These physicians guide us in the right direction and help us investigate the relevant clinical validation questions.

Besides the 9 European hospitals participating in the Clever Bones study, we have collaborations with many more hospitals all over the world. We are really grateful for their support. It is also beneficial for them because they are in the front row witnessing our developments and helping make BoneMRI available for different types of applications and patient populations. They are heroes that help bring BoneMRI to the patient.

Conclusion

So, long story short, clinical evidence and investigations are a big deal. As the medical device industry continues to evolve, their importance will only increase. It is essential for manufacturers to understand their role in providing information to support clinical evaluation for their devices. Ultimately, the goal is to ensure that patients have access to safer and more effective medical devices, which is essential for improving healthcare and quality of life. We at MRIguidance really do care!