Background: On diagnosing Alzheimer's disease (AD), most existing imaging-based schemes have relied on analyzing the hippocampus and its peripheral structures. Recent studies have confirmed that volumetric variations are one of the primary indicators in differentiating symptomatic AD from healthy aging. In this study, we focused on deriving discriminative shape-based parameters that could effectively identify early AD from volumetric computerized tomography (VCT) delineation, which was previously almost intangible. Methods: Participants were 63 volunteers of Thai nationality, whose ages were between 40 and 90 years old. Thirty subjects (age 68.51 ± 5.5) were diagnosed with early AD, by using Diagnostic and Statistical Manual of Mental Disorders IV (DSM-IV) criteria and the National Institute of Neurological and Communicative Disorders and the Stroke and the Alzheimer's disease and Related Disorders Association (NINCDS-ADRDA) criteria, while the remaining 33 were in the healthy control group (age 67.93 ± 5.5). The structural imaging study was conducted by using VCT. Three uninformed readers were asked to draw left and right hippocampal outlines on a coronal section. The resultant shapes were aligned and then analyzed with statistical shape analysis to obtain the first few dominant variational parameters, residing in hyperplanes. A supervised machine learning, i.e., support vector machine (SVM) was then employed to elucidate the proposed scheme. Results: Provided trivial delineations, relatively as low as 5 to 7 implicit model parameters could be extracted and used as discriminants. Clinical verification showed that the model could differentiate early AD from aging, with high sensitivity, specificity, accuracy and F-measure of 0.970, 0.968, 0.983 and 0.983, respectively, with no apparent effect of left-right asymmetry. Thanks to a less laborious task required, yet high discriminating capability, the proposed scheme is expected to be applicable in a typical clinical setting, equipped with only a moderate-specs VCT.
Keywords: early Alzheimer’s disease; shape analysis; support vector machine; volumetric computerized tomography.