Uncertainty Assessment of Quantitative Measurements in Medical Imaging: Applications in Cross-Sectional and Longitudinal Diffusion MRI

Medical imaging measurements have been increasingly investigated as imaging biomarkers for detecting physiological and pathological changes. It is critical to evaluate the reliability of the changes in measurements observed in an individual patient for any clinical decision making. The purpose of this article is to review the utility of uncertainty analysis in medical imaging measurements for individual patients undergoing any pathology or medical therapy. We specifically show how to assess whether the observed alterations in measurements are true changes, i.e. the changes due to medical conditions alone, beyond the uncertainty associated with the imaging, image analysis, or natural physiological occurrences. In order to elaborate the uncertainty concept and the potential applications of uncertainty analysis in the field medical imaging, we reopen two of our previous articles published in the “journal of Physics in Medicine and Biology” [1], and the “Journal of neurological sciences” [2], respectively. While the first study delineates the uncertainty analysis in the context of longitudinal imaging for evaluating the validity of serial measurements, the second work intricates such an application for evaluating the laterality indices and interhemispheric changes in the pathologic brains in cross-sectional studies. Assessment of the repeatability and confidence interval of medical measurements of an endpoint of interest allows us to determine the true changes in measurements, beyond the uncertainty. We have demonstrated how to use estimated repeatability coefficients to evaluate imaging biomarkers for assessing radiation-induced neurotoxicity in patients who had low-grade or benign tumors and were treated by partial brain radiation therapy. We also showed how to use estimated interhemispheric variation uncertainty to identify the epileptogenic side in the patients with temporal lobe epilepsy. In conclusion, we demonstrated that the analysis of uncertainty for the changes in imaging attributes is absolutely crucial, yet beneficial in the context of accurate and reliable diagnostic medical imaging and customizing appropriate treatments for individual patients. The concept can be applied to other imaging modalities and biomarkers in diagnostic and therapeutic assessments.