"How much realism is needed?" - the wrong question in silico imagers have been asking

Area of Science:

• Medical Imaging
• Computational Science
• Biomedical Engineering

Background:

• In silico methods are promising surrogates for physical experimentation in device development.
• Acceptance of computational imaging methods is challenging.
• Realism is often used as a first approximation for assessing these methods.

Purpose of the Study:

• To discuss the limitations of using realism for assessing computational imaging methods.
• To propose an alternative approach for evaluating these techniques.

Main Methods:

• Critique of realism as a sole criterion for in silico imaging assessment.
• Emphasis on the subjectivity and potential irrelevance of realism.
• Advocacy for objective and measurable evaluation metrics.

Main Results:

• Realism does not always guarantee that key features reflect relevant aspects for stakeholders.
• In silico image realism can be irrelevant or misleading in certain applications like automated image analysis.
• A rationale for objective and measurable evaluation of in silico imaging methods is provided.

Conclusions:

• Improved in silico imaging evaluation approaches are crucial.
• These advancements will accelerate the adoption of computational techniques in medical imaging.
• This is particularly important for the regulatory evaluation of new imaging products.