Screening of Alzheimer's disease by facial complexion using artificial intelligence.

Despite the increasing incidence and high morbidity associated with dementia, a simple, non-invasive, and inexpensive method of screening for dementia is yet to be discovered. This study aimed to examine whether artificial intelligence (AI) could distinguish between the faces of people with cognitive impairment and those without dementia.121 patients with cognitive impairment and 117 cognitively sound participants were recruited for the study. 5 deep learning models with 2 optimizers were tested. The binary differentiation of dementia / non-dementia facial image was expressed as a "Face AI score". Xception with Adam was the model that showed the best performance. Overall sensitivity, specificity, and accuracy by the Xception AI system and AUC of the ROC curve were 87.31%, 94.57%, 92.56%, and 0.9717, respectively. Close and significant correlations were found between Face AI score and MMSE (r = -0.599, p < 0.0001). Significant correlation between Face AI score and chronological age was also found (r = 0.321, p < 0.0001). However, MMSE score showed significantly stronger correlation with Face AI score than chronological age (p < 0.0001). The study showed that deep learning programs such as Xception have the ability to differentiate the faces of patients with mild dementia from that of patients without dementia, paving the way for future studies into the development of a facial biomarker for dementia.

Spatial metabolic profiles to discriminate dementia with Lewy bodies from Alzheimer disease.

BACKGROUND: To differentiate dementia with Lewy bodies (DLB) from Alzheimer disease (AD) using a single imaging modality is challenging, because of their common hypometabolic findings. Scaled subprofile modeling/principal component analysis (SSM/PCA), an unsupervised artificial intelligence, has the potential to offer an alternative to image analysis. OBJECTIVE: We aimed to produce spatial metabolic profiles to discriminate DLB from AD and to identify the characteristics of the profiles. METHODS: Fifty individuals each with DLB, AD, and normal cognition (NL) underwent 18F-FDG-PET and MRI. The spatial metabolic profile to differentiate DLB from AD (DLB-AD discrimination profile) was determined using SSM/PCA with tenfold cross validation. For comparison, we also produced disease-related profiles that can discriminate AD and DLB from NL (AD- and DLB-related profiles, respectively). RESULTS: The DLB-AD discrimination profile significantly differentiated DLB from AD with comparable accuracy to that of discriminating DLB and AD from NL. The AD- and DLB-related profiles comprised metabolic imaging features typical of each pathology. In contrast, the DLB-AD discrimination profile emphasized preservation in the posterior cingulate cortex (cingulate island sign) and medial temporal lobe, and occipital hypometabolism. Common hypometabolic findings between DLB and AD were less noticeable in the profile. The DLB-related profile significantly correlated with cognitive function and three core features of DLB, whereas the DLB-AD discrimination profile did not. CONCLUSIONS: Spatial metabolic profile that could discriminate DLB from AD emphasized different imaging features and eliminated common findings between DLB and AD. Neither cognitive function nor core features were associated with the profile.

Deep-learning-based imaging-classification identified cingulate island sign in dementia with Lewy bodies.

The differentiation of dementia with Lewy bodies (DLB) from Alzheimer's disease (AD) using brain perfusion single photon emission tomography is important but is challenging because these conditions exhibit typical features. The cingulate island sign (CIS) is the most recently identified specific feature of DLB for a differential diagnosis. The current study aimed to examine the usefulness of deep-learning-based imaging classification for the diagnoses of DLB and AD. Furthermore, we investigated whether CIS was emphasized by a deep convolutional neural network (CNN) during differentiation. Brain perfusion single photon emission tomography images from 80 patients, each with DLB and AD, and 80 individuals with normal cognition (NL) were used for training and 20 each for final testing. The CNN was trained on brain surface perfusion images. Gradient-weighted class activation mapping (Grad-CAM) was applied to the CNN to visualize the features that was emphasized by the trained CNN. The binary classifications between DLB and NL, DLB and AD, and AD and NL were 93.1%, 89.3%, and 92.4% accurate, respectively. The CIS ratios closely correlated with the output scores before softmax for DLB-AD discrimination (DLB/AD scores). The Grad-CAM highlighted CIS in the DLB discrimination. Visualization of learning process by guided Grad-CAM revealed that CIS became more focused by the CNN as the training progressed. The DLB/AD score was significantly associated with the three core features of DLB. Deep-learning-based imaging classification was useful for an objective and accurate differentiation of DLB from AD and for predicting clinical features of DLB. The CIS was identified as a specific feature during DLB classification. The visualization of specific features and learning processes could be critical in deep learning to discover new imaging features.

Cingulate island sign temporally changes in dementia with Lewy bodies.

The cingulate island sign (CIS) that reflects sparing of the posterior cingulate cortex (PCC) relative to the precuneus plus cuneus on FDG-PET and brain perfusion SPECT, has been proposed as a feature of dementia with Lewy bodies (DLB). As the CIS is influenced by concomitant Alzheimer's disease (AD)-type neurofibrillary tangle (NFT) pathology, we postulated that the CIS gradually disappears as DLB progresses. To determine temporal changes in the CIS, 24 patients with mild DLB and 7 with prodromal DLB underwent 123I-IMP-SPECT and MMSE twice at an interval of two years. The CIS was evaluated as a ratio that was derived by dividing IMP accumulation in the PCC with that in the precuneus plus cuneus. We found that the CIS changed over time and that the relationship between CIS ratios and MMSE scores was inverted U-shaped. Thus, the CIS was most obvious in the vicinity of an MMSE score of 22 and it gradually diminished as the MMSE score decreased. Moreover, a lower CIS ratio in mild DLB was associated with a worse prognosis for cognitive decline, presumably due to concomitant AD-type NFT pathology. Our findings would provide a foundation for the appropriate usage of CIS as a biomarker.

Metabolic correlate of memory-deficit awareness in dementia with Lewy bodies: Implication in cortical midline structure.

The neural substrate of memory-deficit awareness has been studied in mild cognitive impairment and Alzheimer's disease (AD). However, little is known about that in dementia with Lewy bodies (DLB). To determine the neural substrates of memory-deficit awareness in DLB, we investigated the relationship between awareness of memory-deficit and glucose metabolism in DLB. Thirty-four patients with DLB were assessed by 18F-FDG-PET and dopamine transporter (DAT)-SPECT. The awareness was evaluated using an awareness index that represents the discrepancy between objective and subjective memory scores. The association between awareness index and FDG uptake was analyzed using SPM12. Awareness index was significantly lower in DLB than in individuals with normal cognition and was associated with glucose metabolism in the bilateral posterior cingulate cortex and right orbitofrontal cortex. Moreover, the awareness index positively correlated with the cingulate island sign ratio but not with striatal DAT density. The findings suggested that the awareness of memory-deficit in DLB was to some extent impaired and its neural substrate was located in cortical midline structure. The concomitant AD-type pathology might have influenced memory-deficit awareness in DLB.

Preventive Effect of Rifampicin on Alzheimer Disease Needs at Least 450 mg Daily for 1 Year: An FDG-PET Follow-Up Study.

BACKGROUND: Rifampicin was reported to inhibit amyloid-ß oligomerization and tau hyperphosphorylation in mouse models and could serve as a promising available medicine for the prevention of Alzheimer disease (AD). To examine whether rifampicin has such preventive effects in humans, we retrospectively reviewed 18F-FDG-PET findings of elderly patients with mycobacterium infection treated with rifampicin. METHODS: Forty nondemented elderly patients treated with rifampicin for mycobacterium infections who showed AD-type hypometabolism were enrolled. The hypometabolic patterns were evaluated with stereotaxic statistical analysis and region of interest analysis. RESULTS: Before treatment, AD-type hypometa bolism was observed in 12 patients. The FDG uptake in the posterior cingulate cortex (PCC) was improved or stabilized in 6 patients after 12-month therapy (450 mg/day), whereas another 6 patients with 6-month therapy showed a decreased FDG uptake in the PCC. In patients who underwent FDG-PET only after treatment, the metabolic decline in the PCC was significantly milder in patients with ≥12 months of rifampicin treatment than in those with 6 months of treatment. Multiple regression analysis revealed that the dose of rifampicin and treatment duration significantly influenced FDG uptake in the PCC. CONCLUSION: The preventive effect of rifampicin depended on the dose and the treatment duration, and the effect needs at least 450 mg daily for 1 year.

Cholinergic enhancement increases regional cerebral blood flow to the posterior cingulate cortex in mild Alzheimer's disease.

AIM: The brain region that shows reductions in regional cerebral blood flow (rCBF) earliest is the posterior cingulate cortex (PCC), which is thought to have a relationship with cognitive function. We made a hypothesis that the PCC hypoperfusion is a result of cholinergic dysfunction and can be restored by cholinergic enhancement. This present longitudinal study aimed to detect the restoration of PCC rCBF in response to donepezil, an acetylcholine esterase inhibitor. METHODS: We evaluated rCBF changes in the PCC, precuneus and anterior cingulate cortex using perfusion single-photon emission computed tomography (SPECT), statistical analysis and region of interest analysis, prospectively. We allocated 36 patients with mild AD to either the responder or non-responder groups based on changes in Mini-Mental State Examination scores. The patients were followed up for 18 months. RESULTS: The PCC rCBF significantly increased in responders after 6 months of donepezil therapy. Statistical maps at baseline showed a typical decreased pattern of mild AD and obvious rCBF restoration in the bilateral PCC at 6 months in responders. Changes in Mini-Mental State Examination scores and the AD assessment scale cognitive scores significantly correlated with rCBF changes in the PCC of responders. CONCLUSIONS: Cholinergic enhancement restored PCC rCBF under the three conditions of mild AD, responders and short follow-up interval, and that increase correlated with improved cognitive function. These findings support our hypothesis that PCC rCBF reflects cholinergic function in AD patients. Geriatr Gerontol Int 2017; 17: 951-958.

Cingulate island sign on FDG-PET is associated with medial temporal lobe atrophy in dementia with Lewy bodies.

OBJECTIVE: The cingulate island sign (CIS), which refers to sparing of the posterior cingulate relative to the precuneus and cuneus, has been proposed as an FDG-PET imaging feature of dementia with Lewy bodies (DLB). The sign is reportedly associated with Alzheimer's disease (AD) type neurofibrillary tangle (NFT) pathology in autopsy cases. To confirm this relationship using neuroimaging modalities in vivo, we investigated associations between CIS and the medial temporal lobe (MTL) atrophy in DLB. METHODS: Twenty-four patients each of DLB and AD underwent both (18)F-FDG-PET and MRI with voxel-based morphometry. Dopamine transporter (DAT) density was also measured by DAT-SPECT in all those with DLB and in five with AD. The accumulation of FDG in the posterior cingulate ROI was divided by that in the precuneus plus cuneus ROI to derive the CIS ratio from the FDG-PET images. Values for cognitive function of Mini-Mental State Examination (MMSE), Frontal Assessment Battery (FAB) and Ray Auditory Verbal Learning Test (RAVLT) and scores for the core-feature triad of fluctuation, hallucination and parkinsonism were also statistically analyzed. RESULTS: The CIS ratio was higher in DLB than in AD (p < 0.001). The degree of MTL atrophy was lower in DLB than in AD (p < 0.001). The CIS ratio and the degree of MTL atrophy were inversely correlated with DLB (p < 0.001) and with AD (p < 0.05). The CIS ratio did not significantly correlate with DAT density in DLB or with MMSE, FAB, fluctuation score and parkinsonism score. However, the CIS ratio significantly correlated with RAVLT and hallucination scores (both, p < 0.05). CONCLUSIONS: The CIS on FDG-PET in DLB was associated with MTL atrophy but not with striatal DAT density, suggesting that the CIS is a useful neuroimaging biomarker to evaluate coexisting AD-type NFT pathology in vivo. The CIS was also associated with memory impairment and visual hallucination in DLB.

An easy Z-score imaging system for discrimination between very early Alzheimer's disease and controls using brain perfusion SPECT in a multicentre study.

BACKGROUND AND AIM: In Alzheimer's disease (AD), regional cerebral blood flow (rCBF) in the posterior cingulate gyrus and precuneus has been reported to decrease even at a very early stage. We performed a multicentre SPECT study to evaluate the discrimination ability of an easy Z-score imaging system (eZIS) by detecting an rCBF decrease in this area with a common normal database between very early AD patients at the stage of mild cognitive impairment and age-matched healthy volunteers. METHODS: Brain perfusion SPECT images of 40 Alzheimer's disease patients and 40 healthy volunteers were acquired from four gamma camera systems in different institutions. Systematic differences of SPECT images between different gamma cameras were corrected using conversion maps calculated from the SPECT images of the same brain phantom. Ten observers with various degrees of expertise graded eZIS results for receiver operating characteristic (ROC) curves. ROC curves for a positive Z-score in the volume of interest (VOI) of the posterior cingulate gyrus and precuneus were also analysed. RESULTS: An area under the ROC curve value (AZ) for ten observers showed the highest value of 0.866 on average with the smallest standard deviation of 0.027 in the condition of the lower threshold of a Z-score map of 2 without superimposition of VOI. Automated analysis of a Z-score in the VOI showed an AZ value of 0.895. CONCLUSION: Since the degree of expertise of the observers with respect to reading eZIS did not influence the performance and an eZIS can use a common normal database by converting site-specific SPECT data to the core data, the eZIS was considered to be very useful for diagnosing early AD in routine studies in many institutions.