Automated Device Positioning: In our in-silico medical device testing practice what often happens is that we are presented with extensive patient cohort data. The first step often involves segmenting out all the relevant anatomies using a range of automatic to semi-automatic algorithms. What we end up with at this stage are 3D meshes (or 4D in case of time-series data). Then, to assess how a medical device fits within a patient cohort, careful positioning in 3D space must be performed for each patient, respecting their unique anatomical differences and idiosyncrasies. This process helps in simulation stages and allows us to perform measurements or iterate on the device design accordingly.
Our Data Team is working on a method to optimize medical device fit with digital twins, making it easier to achieve the optimal fit quickly and accurately across multiple patients using a technique called automated device positioning.
Image 2: Mesh of a segmented human heart with left ventricle highlighted. This is where we will position our device.
Image 3: A component of a left ventricular assisting device (LVAD). This is what we want to position correctly.
Challenges of 3D Device Positioning
Positioning devices in 3D space is hard. Although technologies such as AR/VR go a long way in facilitating much needed spatial awareness to the user, they still have not been widely adopted. Plain old 2D screens are what most of us still use daily. They are more than sufficient for most of the common tasks, like reading text, editing a video, or programming. However, they struggle as soon as the third dimension comes into play.
In 3D space, we generally have 6 degrees of freedom for a rigid object placement: 3 for translation, and 3 for rotation. Careful iterative tinkering is required to move around such an object and to correctly place it. In most industry software, one must have experience working with 3D geometries to do this seemingly straightforward with ease.
Image 3: 3 degrees of freedom for translation.
Image 4: 3 degrees of freedom for rotation.
Having to perform this task for the whole cohort of patients might also host potential headaches down the road, such as:
- What if we change our mind about what is a fitting placement for a given device?
- What if a new or modified device design needs testing?
- What if we want to increase our cohort size?
In all these cases, we would ultimately need to go back to our 3D software of choice and spend a considerable amount of time positioning the device yet again.
Image 5: Possible (here overexaggerate) design improvements that might require repeated placement of the device.
Simplifying the Process with Automated Device Positioning
Our solution aims to eliminate these challenges by enabling users, regardless of their experience with 3D software, to accurately position devices in minutes. Using automated fitting of medical devices in 3D patient models, users can fine-tune the device fit, swap to a new version, or apply it to a different patient—all with just a few clicks. Our Data Team can help you set it up and ensure this process is easy and straightforward for you.
How would this work?
We are developing a method where users only need to position a device once in a template patient. This positioning will then propagate automatically to all other patients in the cohort, saving considerable time and effort.
Image 6: LVAD positioned in the template patient.
Key Steps of Automatic Device Positioning:
- Obtain point correspondences between the target patient’s anatomy and the template patient’s anatomy.
2. Anchor the device to the template patient’s anatomy at key points.
3. Optimize the device position relative to the target patient anatomy to ensure the fit is consistent across patients.
Image 9: LVAD automatically positioned in other patients.
Benefits of automated device positioning
- Significant time savings
- In traditional 3D environments, manually positioning devices for each patient is time-consuming. Automated 3D placement of medical devices in patient models reduces this time by allowing users to position a device once in a template patient and replicate it across the entire cohort.
- Consistency across multiple patients
- When multiple users manually position devices, variations are inevitable. Automated device positioning ensures consistent placement across all patients, leading to more accurate and reliable simulation results.
- Quick iteration on design changes
- Whether it’s a new device version or an expanding patient cohort, automated positioning lets design teams quickly adjust. Instead of repositioning devices for each individual patient, a simple update can propagate across the entire cohort, enabling faster design iterations and testing.
Conclusion
Automated device positioning simplifies the process of placing medical devices in patient models, saving time and ensuring accuracy across multiple patients. By making the process more accessible, even for users without extensive 3D experience, this feature allows teams to focus on improving device designs and achieving faster, more reliable outcomes. It’s a game-changer in optimizing medical device fit and performance within digital twins, leading to better, more efficient in-silico medical device testing.
This is, of course, just a high-level overview of the algorithm. For more information on how we can apply this to your specific device and project, feel free to send us an e-mail. For more technical details about the method, you can contact Miro from Data Team directly at brezik@virtonomy.io.
