How can engineers test an imaging system before trying every possible patient, position, or device setting? They build a model: a mathematical representation that predicts how parts of a system interact. Models can reveal promising configurations and expose weak assumptions, but their answers remain predictions that need comparison with measurements and real-world performance.
A virtual test bench
A computational model translates a physical question into equations and data that a computer can work through. For an imaging system, the model might represent a device, the body region being scanned, and the signals that pass between them. Researchers can vary conditions systematically in a virtual environment, exploring combinations that would be expensive or impractical to test one by one. This can help focus physical experiments on the most informative cases. [1]
A model is not a miniature scanner. It is a simplified description, built around assumptions about materials, anatomy, motion, or signal behavior. Mechanistic models use known scientific principles; data-driven models learn patterns from datasets; some combine both. Each can be useful, but each may miss effects that were not represented in its equations or training data. [1]
What gets represented
For example, imagine testing whether a sensor’s position changes the strength of a signal. A model could represent the sensor, tissue layers, and relevant physical interactions, then estimate how a shift alters the recorded signal. If the prediction suggests a sharp drop in one position, engineers can investigate that arrangement with a physical prototype. The causal chain is explicit: assumptions define the virtual setup, equations calculate expected behavior, and measurements test whether that behavior is realistic.
Medical device technologies span tools that monitor, modulate, or supplement bodily functions, from external wearable sensors to implanted pacemakers. Modeling can support development across such devices, but it does not make every wearable a diagnostic tool. The purpose and performance of each device depend on its design and evidence. A model may estimate a signal or help refine a system; clinical claims require validation beyond a simulation. [2]
Validation against reality
To validate a model, researchers compare its predictions with independent observations, such as bench measurements, experimental data, or clinical data appropriate to the question. Agreement in one setup gives confidence for that range of conditions; it does not guarantee accuracy everywhere. If the body moves differently than assumed, or tissue properties vary, the prediction may drift. Sensitivity analysis—changing assumptions to see which ones strongly affect results—can identify where more data are needed.
The same principle applies to imaging. A model that predicts how a system behaves under a fixed geometry may not capture patient movement, device variation, or the full diversity of human anatomy. Testing must be matched to the intended use. The better the correspondence between model inputs and the actual device and setting, the more meaningful its predictions become.
Models guide questions, not certainty
Computational modeling can run many simulated experiments and help select a smaller number of laboratory tests that are likely to teach researchers more. It can also link levels, from molecular processes to tissues and whole organisms, in multiscale models. But each added scale introduces assumptions and uncertainty. A detailed output is not automatically a reliable one; the underlying evidence and validation matter. [1]
The most useful way to read a model result is as a conditional statement: if the represented physics, inputs, and boundaries are adequate, then the predicted behavior follows. That makes models powerful design tools. They help researchers ask sharper questions, compare configurations, and decide what to measure next, while the imaging system’s actual performance must still be established through appropriate testing.
Sources & further reading
Prepared for the independent PSI Center reading library. About edition dates and source checking.
This explainer describes scientific principles. It does not assess individual treatment choices.
