The impact of registration accuracy on imaging validation study design: A novel statistical power calculation.

摘要: Abstract Novel imaging modalities are pushing the boundaries of what is possible in medical imaging, but their signal properties not always well understood. The evaluation these novel critical to achieving research and clinical potential. Image registration accepted reference standard an important part characterizing elucidating effect underlying focal disease on signal. strengths conclusions drawn from analyses limited by statistical power. Based observation that this context, power depends uncertainty arising error, we derive a calculation formula relating number subjects, minimum detectable difference between normal pathologic regions for validation study design accommodates correlations within image regions. Monte Carlo simulations were used evaluate derived models test strength assumptions, showing model yielded predictions power, simulated experiments accurate maximum error 1% when assumptions derivation met, sensitivities violations assumptions. use formulae illustrated through subjects required case study, modeled closely after prostate cancer currently taking place at our institution. address three central questions studies: (1) What acceptable error? (2) How many needed? (3) regions?