Jasmin Annica Kuhn-Keller

166 Chapter 8 clinical patient data to create subgroups of patients with specific phenotypes. Such software needs to be validated in bigger populations and with data from different MRI scanners before it can be used in clinical practice. It should be noted that the possible clinical translation described may only be adopted in the future, when more treatment options for selected patients are available as this could lead to a cost-effective approach. Frequent MRI scanning as a population screening tool is unlikely to be feasible due to high costs and logistic constraints. Therefore, general practitioners could make a first assessment of risk factors for dementia and preselect patients for an MRI scan. Software-based brain image analysis results, including WMH shape markers, could subsequently be integrated leading to individualized risk assessment into the radiological report. With all novel data-driven or machine learning methods emerging these days, the question should not be if we should use them, but how.

RkJQdWJsaXNoZXIy MTk4NDMw