Affiliations: [a] VTT Technical Research Centre of Finland, Tampere, Finland | [b] Department of Neurology, University of Eastern Finland, Kuopio University Hospital, Kuopio, Finland | [c] Department of Computing, Imperial College London, London, UK | [d] Department of Neurology, Memory Disorders Research Group, Rigshospitalet, Copenhagen University Hospital, Copenhagen, Denmark
Correspondence:
[*]
Correspondence to: Jussi Mattila, VTT Technical Research Centre of Finland, Sinitaival 6, 33101 Tampere, Finland. Tel.: +358 40 592 7979; Fax: +358 20 722 3499; E-mail: [email protected].
Note: [1] 1Data used in preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (http://adni.loni.ucla.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report. A complete listing of ADNI investigators can be found at: http://adni.loni.ucla.edu/wp-content/uploads/how to apply/ADNI_Acknowledgement_List.pdf
Abstract: In the diagnostic process of Alzheimer's disease (AD), there may be considerable delays between first contact to outpatient services and a final, definitive diagnosis. In Europe the average delay is 20 months. Nevertheless, patient data preceding clinical AD diagnoses often contains early signs of the disease. Several studies have analyzed data of mild cognitive impairment (MCI) subjects, showing that conversion from MCI to AD can be predicted with a classification accuracy of 60–80%. This accuracy may not be high enough for influencing diagnostic decisions. In this work, the prediction problem is approached differently; a target prediction accuracy is defined first and is then used for identifying MCI patients for whom the required accuracy can be reached. The process uses a novel disease state index method in which patient data are statistically compared to a high number of previously diagnosed cases. It is shown that the disease index values derived from heterogeneous patient data can be used for identifying groups of patients for whom the prediction accuracy reaches the previously set target level. The results also show that 12 months before receiving clinical AD diagnoses, approximately half (51.5%, 95% confidence interval: 48.6–54.2%) of MCI subjects who progressed to AD can be classified with a high accuracy of 87.7%, possibly enough to support earlier diagnostic decisions.
