The Physics-Based Simulation Approach to De-Risking Imaging Innovation

Executive summary

• We propose a relatively novel approach to gathering real, defensible and personalised data

• Physics-based simulations like Becquerel provide a cheap way to derisk and iterate your device early on

• Now backed by the FDA, it is a great way to accelerate innovation for medical imaging and MedTech developers

  
  

The Data Bottleneck

If you are attempting to bring a new medical imaging device to market, you might find that your team can iterate quickly in software and hardware but obtaining high-quality data and moving towards validation may be lagging behind. This is a major bottleneck nowadays due to data being expensive and hard to obtain. Every month spent waiting for data has implications on burn rate and delayed evidence. Meaning you may end up losing your competitive edge. Our proposal is this: physics-based simulations provide an alternative to the industry’s three default options for gathering data, synthetic data is a means of obtaining quicker, cheaper and specific data 

The Three Default Options

Methods of gathering data are typically viewed through three main methods. Based on your specific needs, one may serve to be better than the other based on your own cost considerations, clinical specificity needs and time-to-evidence approach. 

Real Clinical Data (Expensive but Generic)

This is seen as the most realistic baseline and it is a defensible starting point. However, acquisition is costly and the governance is heavy: IRB processes, de-identification pipelines, data use agreements, site contracting etc… Including as well that timelines for proceeding with this approach are usually unpredictable and may set you behind schedule. But still, when considering validating your edge-case indication or device behaviour, the data you obtain may be more generic than you want. Especially if you are addressing a novel pathology pattern, you might find yourself stuck because you can’t conjure what hasn’t been captured.

Model Zoo - Generative Data (Fast, Cheap but Low Verifiability)

This method will provide volume quickly and help with early prototyping but the core limitation is traceability. Due to the nature of generative data, it is speculative and often it is regressed to “vibe imaging”. The production of plausible-looking images without the aspect of ground-truth will certainly produce drawbacks. Validation is limited without traceability and makes the data hard to defend. Regulators and investors will look at this with a grain of salt and your team may build false confidence in data that is not true to real problems. 

Custom Clinical Trials (Gold Standard but Late-Stage Expensive)

These are custom, purpose built studies designed to produce the most relevant evidence for your product. Unfortunately, the cost that comes along with obtaining this evidence restricts it from being an early phase tool. Costs start around $200k for a clinic-ready prototype design and can scale quickly depending on sites, protocols and endpoints. They are also lengthy and best reserved for confirmatory phases; at that point you already know what you are looking for.

But what about when you need solid verifiable early stage validation?

The Missing Fourth Path; Physics-Based, High-Fidelity Simulation

A physics-based simulation offers an alternate route. One which is capable of generating image data from anatomically grounded models and the known physics of image formation.One simple differentiator of this approach is that you can prove the pathology you defined is present. Plus any data obtained is fundamentally backed by ground-truth physics. Being able to easily introduce variability in the system also provides an advantage. For example, one can alter autonomy, pathology severity, acquisition parameters and device settings to cater to your needs. You are no longer limited to real human patients and instead can form as many virtual patients, with your specific data needs, as you wish. This allows your team to learn what truly drives your device’s performance and gather lots of useful early data.The proposition of this method from a business perspective:

• Before expensive clinical commitments you are able to de-risk feasibility

• Improved efficiency in R&D by reducing reworks

• Early evidence to incentivise investors and progress your regulatory strategy

Ultimately, synthetic data of this nature doesn’t serve to replace real data per se but if used intelligently, it can improve velocity and set you apart from competitors on a budget.

Imaging Alone is not Enough

Another advantage of physics-based simulations is that it can be coupled with multi-physics. To achieve decision-grade validation you need more context than just imaging. One needs to understand why something appears, shifts or fails under certain physiological conditions.Imaging simulation can be connected to:

• FEM (finite element methods): understand tissue mechanics, deformation and device-tissue interaction

• CFD (computational fluid dynamics): model fluid flow, hemodynamics and perfusion related effects

These techniques will be able to create a richer virtual environment so that you can further confirm whether your device is valid and allow your team to gather the data you need. Through this you can understand issues of your tech earlier and know what to work on. Very useful for stress testing your device at an early stage and support preclinical rationales and device design validation.  

Introducing Becquerel

Becquerel accelerates imaging-based medtech development by giving teams instant access to realistic, multimodal synthetic data without the delays and operational difficulties of sourcing clinical imaging. Earlier testing and refinement of your medical imaging device will allow your team to iterate and validate quickly saving you time and money when it counts.

Features:

Seamless Multi-Modal Simulation

Generate realistic CT/MRI/ultrasound/radiography/nuclear imaging from a single anatomical model.

Benefit: consistent comparisons across modalities and pathologies

FEM + CFD Software Integration

Bridge imaging with physics by connecting to FEM and CFD tools to simulate tissue mechanics, fluid flow, and physiological response alongside imaging.

Benefit: accurate and diverse methods of testing your device

Advanced Anatomical Shape Models

Anatomically accurate parametric models built from high-fidelity data.

Benefit: explore case/patient and parameter variability -> generate custom anatomies and pathological changes

With Becquerel, the data you generate will be specific, controllable and defensible and tailored to your needs.

Where this fits in the Development and Evidence Roadmap 

You might ask what does the FDA think about this synthetic data?

Let's have a look at the Virtual Imaging Clinical Trial for Regulatory Evaluation (VICTRE) trial.

In this landmark study by the FDA themselves, using almost 3000 patients, they found that the virtual trial reached the same conclusion as the human trial. Meaning that the industry views this high-fidelity synthetic data as regulatory-grade evidence.  

In this way, Becquerel, or physics-based simulations in general, are a very solid pre-clinical validation accelerator. It is an excellent way to de-risk and iterate your product.

Use cases:

• AI and algorithm training/stress-testing

• Virtual trial design

• Stress-test edge cases 

• Device design iteration 

• Device-tissue interaction validation

• Pre-clinical evidence packages

• Enrichment strategy to maximise signal

More and more companies are adopting this approach as well as regulatory agencies like the FDA. If you think you may benefit from this technology it is worth looking into.

A Practical Bridge to Earlier Confidence

Real clinical data and custom clinical trials are slow and expensive and more tailored towards late-stage development. In addition, generative data is fast but hard to validate. 

Physics-based simulation could be your early-stage method of obtaining solid data to back your product. 

If you want a second opinion on your data strategy, contact us and we can help you figure out what approach is best for you.