Although several studies demonstrated high performances for classification of dementia based on atrophy derived from structural MRI, these computer-aided diagnosis methods are not yet used in clinical practice. A major reason for this is that these methods are not well enough validated for clinical use. Firstly, although these studies mostly perform a thorough validation, the data set and evaluation methods vary, which makes it hard to compare performances between different methods. Secondly, for clinical implementation, the generalizability of the methods should be evaluated on previously unseen multicenter data, which is not often performed. In this challenge, the ground truth diagnoses are blinded to the participants to avoid that the methods are overtrained. Thirdly, a requirement for clinical applicability is multi-class classification of Alzheimer's disease (AD), mild cognitive impairment (MCI) and healthy controls. In contrast to most previously published work, this challenge addresses the multi-class diagnosis.
By addressing the above mentioned issues on comparability, generalizability, and clinical applicability, we aim to take a step forwards to clinical use of computer-aided diagnosis (CAD) methods for dementia.
Web site: caddementia.bigr.nl