Classification of early and severe Alzheimer Disease. Differences in accuracy basing the analysis of SPECT CBF data on either hippocampus, temporo-parietal lobes or factorial analysis

Background: Hippocampus and temporo-parietal regional cerebral blood flow
(rCBF) reductions are well known to be specific of early (eAD) and severe
Alzheimer Disease (AD). The specificity and the sensitivity of SPECT rCBF
in discriminating the two stages of the disease vary across different
investigations. The aim of this study is to develop a new method to
increase the accuracy in the classification of eAD and AD.
Methods: Twenty eAD and 21 controls (CTR), and 15 AD and 13 further CTR
were included in the study. The two group pairs were age-matched and rCBF
was assessed by 99m-Tc-HMPAO and using a three-headed gamma camera. Regions
were identified and signal intensity was evaluated by a standardised brain
atlas. Hippocampus, temporal and parietal lobes and four factors, derived
from a previous principal components analysis, and whose rCBF was proven to
significantly differ between groups, were considered for data analysis.
These four factors were functionally connected clusters of Brodmann areas
belonging to the temporo-parietal lobes (n=3) and to central structures
(n=1) and were analysed together. The accuracy of the classification of
eAD, AD and CTR utilising the K-means clustering method was separately
tested for each group pair and for each region.
Results: In AD/CTR evaluation, hippocampus uptake could correctly classify
the 82.1% of the subjects, while the accuracy of both temporo-parietal
lobes and the four joint factors was 96.4%. When the correct classification
to eAD/CTR groups was tested, the accuracy of 99m-Tc-HMPAO uptake intensity
in discriminating the groups was 85.4% for hippocampus, 80.5% for
temporo-parietal lobes and 87.8% for the four joint factors.
Discussion: Utilising separately sensitive regions (hippocampus), lobes
(temporo-parietal) and functionally connected regions in classifying eAD
and AD yielded different results. Severe AD was better classified by data
from lobes and from the functionally connected regions affected by the
disease. Early AD was classified with higher accuracy by hippocampus and
the functionally connected regions in temporo-parietal lobes and central
structures. Such differences are consistent with the progression of the
disease which is supposed to start in the medial temporal lobe and spread
through the temporo-parietal cortex. The proposed method adds information
mainly in the early stage of the disease.